Colour Measurement and Print Quality Assessment in a Colour Managed Printing Workflow

Transcription

1 Colour Measurement and Print Quality Assessment in a Colour Managed Printing Workflow Doctoral Dissertation by Peter Nussbaum The Norwegian Color Research Laboratory Faculty of Computer Science and Media Technology Gjøvik University College P. O. Box 191 N-2802 Gjøvik, Norway Submitted to the Faculty of Mathematics and Natural Science at the University of Oslo in partial fulfillment of the requirements for the degree of Philosophiae Doctor (PhD) in Imaging Science Department of Informatics University of Oslo P. O. Box 1080 Blindern N-0316 Oslo, Norway December 2010

2 Peter Nussbaum, 2011 Series of dissertations submitted to the Faculty of Mathematics and Natural Sciences, University of Oslo No ISSN All rights reserved. No part of this publication may be reproduced or transmitted, in any form or by any means, without permission. Cover: Inger Sandved Anfinsen. Printed in Norway: AIT Oslo AS. Produced in co-operation with Unipub. The thesis is produced by Unipub merely in connection with the thesis defence. Kindly direct all inquiries regarding the thesis to the copyright holder or the unit which grants the doctorate.

3 Abstract When digital image data are reproduced and printed, the corresponding visual stimulus, i.e., a print viewed in a certain way, can be highly dynamic. There is a great potential for variations in the final perceived print quality due to various factors involved in the printing process, as well as in the viewing of the resulting print. A number of studies and research have been done in the field of print quality and print assessment, and repeatedly it has been concluded that it is a very complex issue. Often, printed matter, such as magazines and newspapers show variations in their print quality caused by a range of factors, and the competitive situation in the graphic arts and printing industry is an invitation to develop methods and procedures to limit these variations in the printing process. The purpose of this study is to develop methodologies and procedures for print quality assessment in a colour managed printing workflow. Even though multiple factors can affect the appearance of the print, the aspect "colour" is paid most attention to in this work. In order to achieve this goal, the work requires collection of relevant data gathered from visual observations using psychophysical experiments and quantitative measurement methods, applying state of the art measurement instruments. An important part of the thesis is the examination of typical measurement devices used in the graphic arts and printing industry in terms of their accuracy and reproducibility. The performance of measurement instruments can severely affect the process control and finally the judgement of the print and proof quality. Hence, the inherent uncertainty of measurement instruments is investigated and the consequences discussed. Considering the comparison of densitometric and planimetric measurement technique for newspaper printing the outcome showed that the proposed model does not accurately describe the relation between the two measurement technologies due to the large uncertainty for dot meters. This can be explained by the poor repeatability performance for dot meters on newspaper print. Regarding the identified measurement variations between different colour measurement instruments, a solution is required to reduce the measurement uncertainty between them. Therefore another method is proposed to reduce these variations by applying a regression technique directly on the measurement output values in the CIELAB colour space. Simultaneously, the aim is to improve the colorimetric performance and inter-instrument and inter-model agreement. V

4 The main contribution of this thesis is to identify methodologies and routines to implement standardization procedures in a colour managed printing workflow to ensure consistent and predictable print quality. In particular, important parts of the result is the demonstration of a method for soft proofing in a standardized printing workflow and pointing out procedures for applied colour management in a standardized heat-set web offset and newspaper printing. The analysis of measurement uncertainty and the proposed solution to correct the instrument s systematic errors is also one of the important achievements in this work. In conclusion the results can be used as an application independent framework for those involved in the process of print quality assessment. VI

5 The Publications The seven papers listed below constitute the core work of the present thesis. Further, four publications and their contribution are related to the present research work, but are not included in the full text in this thesis. List of Included Papers Paper A P. Nussbaum, J. Y. Hardeberg, and S. E. Skarsbø Print Quality Evaluation for Governmental Purchase Decisions In Advances in Printing Science and Technology: Proceedings of the 31 st International iarigai Research Conference, Volume 31, pp M. Lovreček, Ed., Acta Graphica Publishers, Paper B P. Nussbaum and J. Y. Hardeberg Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset In Advances in Printing and Media Technology: Proceedings of the 33 rd International Research Conference of iarigai, Volume 33, pp N. Enlund and M. Lovreček, Ed., Acta Graphica Publishers, Paper C A. Sole, P. Nussbaum, and J. Y. Hardeberg Implementing ISO Standards for soft Proofing in a Standardized Printing Workflow according to PSO In Advances in Printing and Media Technology: Proceedings of the 37 th International Research Conference of iarigai, Volume 37, pp N. Enlund and M. Lovreček, Ed., International Association of Research Organizations for the Information, Media and Graphic Arts Industries, IX

6 Paper D P. Nussbaum and J. Y. Hardeberg Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print Color Research & Application, Wiley. Article first published online: March 8 th 2011, DOI: /col Paper E M.S. Wroldsen, P. Nussbaum, and J. Y. Hardeberg A Comparison of Densitometric and Planimetric Measurement Techniques for Newspaper Printing In TAGA Journal of Graphic Technology, Technical Association of the Graphic Arts, Vol. 4, pp , Paper F P. Nussbaum, A. Sole, and J. Y. Hardeberg Analysis of Colour Measurement Uncertainty in a Colour Managed Printing Workflow Accepted for publication in Journal of Print and Media Technology Research, The International Association of Research Organizations for the Information, Media and Graphic Arts Industries (iarigai). Paper G P. Nussbaum, J. Y. Hardeberg, and F. Albregtsen Regression based Characterization of Colour Measurement Instruments in Printing Applications In Electronic Imaging: Color Imaging XVI: Displaying, Processing, Hardcopy, and Applications, SPIE Proceedings, 7866, San Francisco, CA, X

7 List of Related Papers P. Nussbaum, A. Sole, and J. Y. Hardeberg Consequences of using a number of different color measurement instruments in a color managed printing workflow In Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 61 st Annual Meeting, New Orleans, LA, P. Nussbaum and J. Y. Hardeberg Print quality evaluation and applied colour management in coldset offset newspaper print In Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 60 th Annual Meeting, San Francisco, CA, S. Roch, J. Y. Hardeberg, and P. Nussbaum Effect of time spacing on the perceived color In Proc. SPIE: Color Imaging XII: Processing, Hardcopy, and Applications, 2007, 6493, pp J. Y. Hardeberg, P. Nussbaum, S. Roch, and O. Panak Time matters in soft proofing In ACTA GRAPHICA, Journal for Printing Science and Graphic Communication, vol. 19, pp. 1-10, XI

8 Contents Abstract...V Acknowledgements...VII The Publications... IX List of included Papers... IX List of related Papers... XI PART I INTRODUCTION 1 Introduction Motivation Aim of the Study Research Methodology Outline of Thesis Background Colour Fundamentals Human Visual System Colorimetry CIE Standard Observer CIE Standard Illuminants Metamerism CIE 1931 XYZ Colour Space Uniform Colour Spaces CIELAB CIELUV CIELAB Colour Difference Colour Measurement Measurement Instruments and their Application Sources of Error and Measurement Uncertainty...37 XIII

9 2.3 Image Reproduction and Colour Management Additive or Subtractive Colour Mixing Image Reproduction Colour Gamut Mapping Principles of ICC Colour Management Calibration and Characterization Standardization in Offset Printing and Prepress Print Quality and Print Assessment What is (Print) Quality? Factors Contributing to Print Quality Print Quality Evaluation by Visual Observation Quantitative Print Quality Evaluation Summary of included Papers Paper A: Print Quality Evaluation for Governmental Purchase Decisions Paper B: Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset Paper C: Implementing ISO Standards for soft Proofing in a Standardized Printing Workflow according to PSO Paper D: Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print Paper E: A Comparison of Densitometric and Planimetric Measurement Techniques for Newspaper Printing Paper F: Analysis of Colour Measurement Uncertainty in a Colour Managed Printing Workflow Paper G: Regression based Characterization of Colour Measurement Instruments in Printing Applications XIV

10 4 Discussion and Conclusion Discussion of Papers in Context Summary of Contributions Conclusion Perspectives References PART II INCLUDED PAPERS Paper A Paper B Paper C Paper D Paper E Paper F Paper G..289 XV

11 XVI

12 PART I INTRODUCTION 1

13 2

14 Introduction 1 Introduction This chapter provides a brief introduction considering the aim of and the methodology used in the present work, and also describes the structure of the thesis. 1.1 Motivation Although the electronic media market is in rapid development, it is still considered that most people prefer reading on paper rather than on displays [75]. From an environmental point of view recent research results demonstrate that paper based media are contributing less to global warming compared to electronic media, taking a life cycle perspective on the products into account [40, 119, 120]. These, among many other reasons, are why paper based material still has a notable presence in the communication market, and printed matter such as books, textbooks, magazines and newspapers still play an important role in our daily life. A recent study by Intergraf 1 [76] considering the future of the European print industry confirms this statement. However, the study concludes that the effect of external factors, such as expansion in communication technology, market shifts, and changes in production technologies, will force traditional printing companies to be more agile and to respond more swiftly to future changes. The competitive situation has become very dynamic in the sense that not only the competition with other media has grown but also the requirements for more rapid, predictable and cost efficient print production has increased. To meet these demands, standardization and the adoption of international specifications in the printing workflow and process control are essential and an important key to success. We recognized the potential in facing the opportunities to embrace these challenges and decided to approach this research field. Print quality, print assessment, process control, colour management, and colour measurement are the main issues involved and discussed in this work. 1.2 Aim of the Study The aim of this study is to develop methodologies and procedures for print quality assessment in a colour managed printing workflow according to objective evaluation and perceptual judgement. Furthermore, due to the quantitative evaluation involved, an 1 International confederation for printing and allied industries (Intergraf). 3

15 Introduction important part of the study is the investigation of measurement devices in terms of accuracy and reproducibility for process control and print and proof quality assessment. Hence, the measurement uncertainties of measurement instruments used in the graphic arts and printing industry will be investigated and the consequences discussed. The challenges identified during the initial research period in using different measurement instruments and the reliability of the obtained measurement data have led to finding a solution to reduce the measurement uncertainty among different instruments. Hence, a method is proposed to reduce the colour differences among different measurement instruments, and simultaneously improve the colorimetric performance and inter-instrument and inter-model agreement. In this study print quality is seen in the context of a colour managed printing workflow and the factors affecting the appearance. Although there are a number of factors in the workflow influencing the appearance of the print, it is in particular the factor colour which is the most important focus of attention in this work. Our aim has from the outset of the project been to develop methodology and procedures that could be used as an application independent framework for those involved in the process of print quality assessment. 1.3 Research Methodology The research area of the thesis is in the field of colour measurement and print evaluation in a colour managed printing workflow. In order to obtain the stated aim of the study which includes the development of methodologies and procedures for print quality assessment, our research methodology requires gathering relevant data from visual observations using psychophysical experiments and quantitative measurement methods applying state of the art measurement instruments. It has been of pivotal importance to obtain measurements from different types of instruments used in the graphic arts and printing industry, and to study inter-instrument and inter-model performance in terms of accuracy and reproducibility. In the research period three different types of psychophysical methods, presenting the printed samples to the observers and collecting the observer s judgement were used in Paper A [134], Paper B [132], and Paper D [131]. On the other hand considering the instrumental measurements, different types of measurement instruments were used to gather the data set in all seven Papers A-G [ , 136, 172, 194]. 4

16 Introduction 1.4 Outline of Thesis The thesis is intended to provide the reader with the understanding needed to determine print quality in a colour managed printing workflow using both psychophysical experiment and quantitative evaluation. In order to structure the content the thesis is divided into two parts and four chapters. PART I provides in Chapter 2 an overview of the research area and related fields including an introduction to colour fundamentals, colour measurement, colour reproduction and colour management, and print quality and print assessment which are essential topics of the thesis. Chapter 3 highlights the research work by summarizing the contribution from the individual papers. Furthermore, a discussion, summary, and conclusion of the entire thesis are presented in Chapter 4 and finally suggestions for further research are proposed. PART II presents the main contribution of the thesis including the seven published papers describing the research work of the thesis. 5

17 Introduction 6

18 Background Colour Fundamentals 2 Background The focus of the present study is in the field of print quality evaluation in a colour managed printing workflow. Therefore some knowledge of fundamental colour principles is important and the subject has to be addressed. Hence, the purpose of this chapter is to give an overview of the research area and provide a concise introduction to some of the fundamentals of colour science, colorimetry, colour measurement, colour management and print quality. 2.1 Colour Fundamentals In this chapter we recall some of the basic concepts of measuring colours, in particular the relationship between light sources, reflective objects and observers. For instance we explain how to compute the tristimulus values CIEXYZ and CIELAB of an object, knowing its spectral reflectance property, illuminant and observer conditions. However, a wide range of comprehensive literature exists describing in depth the different topics addressed in this section. For instance Roberts [154], and Wandell [188] review the properties of the human visual system, and Valberg [185] explores the fundamentals of human colour vision. A complete treatise on colour science can be found in the book by Wyszecki and Stiles [198]. Further, refer for example to Hunt [70], [71] for measuring colour and for colour reproduction, respectively, and Otha and Robertson [139] for colorimetry, Schanda [159] for understanding the CIE system, and Sharma [167] for colour management Human Visual System The human visual system is sensitive to electromagnetic radiation in the range of wavelengths of approximately 380 nm-780 nm, usually referred to as light. When we observe light reflected from an object as illustrated in Figure 1, the light enters the eye and is focused onto the retina, which consists of a mosaic of specialized cells called rods and cones containing pigments that respond to light. When the visual pigments absorb light chemical changes take place, which initiate electrical impulses. The stimuli are processed by a neural network of brain cells and eventually lead to the excitation of other cells in various specialized areas of the outer region of the brain known as the visual cortex where the stimulus is interpreted as a sensation of colour [191]. In the human retina there are 7

19 Colour Fundamentals Background approximately 130 million photoreceptors (120 x 10 6 rods and 6 x 10 6 cones). The rod cells are responsible for low-intensity night vision referred to as scotopic vision, while at a higher or photopic level of illumination three types of cone receptor cells have maximum sensitivity to light in three different parts of the visible spectrum at 420 nm (short wavelengths corresponding to violet), 530 nm (medium wavelengths corresponding to yellowish-green) and 560 nm (long wavelengths corresponding to greenish-yellow) named as S, M, and L cones respectively [185]. It is important to note that the three cone types are not distributed evenly throughout the retina. The ratio of their relative abundances is approximately twice as many L cones as there are M cones. The S cones are rare throughout the retina and are almost entirely absent in the fovea itself [67]. Figure 1: Schematic representation of the principles of viewing colour. The perceived colour stimulus depends on the spectral power distribution (SPD) of the light source, the spectral reflectance properties of the object, the spectral sensitivities of the L, M, and S cones in the human eye and the resulting interpretation of the sensation by the visual system Colorimetry To quantify and to model the human colour perception, colorimetry has been introduced, which is the science of measuring colour. This takes into account the physical characteristics of the light source, the object s spectral properties and the physiological aspects of human vision [138]. In colorimetry this is also referred to as psychophysical interaction, which is the relationship between the physical properties and the resulting colour sensation. Colorimetry is founded on a classic series of colour matching experiments that allowed the trichromatic (three-channel) properties of human vision to be studied and characterized. Hunt [70] gives a very good overview of CIE colorimetry [25], 8

20 Background Colour Fundamentals and most of the present summary is based on that source (all formulae are taken from there, unless stated otherwise). In 1931 the international Commission on Illumination (Commission Internationale de l Eclairage, CIE) introduced a system for the specification of colour measurement based on additive colour mixing. Thus, by choosing three appropriate primaries all colour stimuli can be matched by the additive mixture. In conducting a colour matching experiment, the observer regulates the amount (intensities) of the three primaries until their mixture appears to match the test colour, and the amount of the primaries used to produce the match are commonly known as the tristimulus values. A set of three curves called colourmatching functions (CMF) are obtained by using a series of test colours of monochromatic light throughout the visible spectrum and recording the amounts of the three primaries required, to match each individual test colour CIE Standard Observer The 1931 CIE system is based upon a linear transformation of the original colour-matching functions averaged from matching experiments performed by Guild [62] and Wright [192] using particular real primaries R, G, B to a set of imaginary primary stimuli known as X, Y and Z. The CIE experiment for deriving the colour-matching functions were conducted at 2º viewing angle, and the functions are denoted by x (λ), y (λ), and z (λ). In 1964 CIE defined Supplementary Standard Colorimetric Observer colour- matching functions for samples viewed under 10, denoted by x 10 (λ), y 10 (λ), and z 10 (λ). The viewing angle refers to the angle formed at the eye by an object at normal viewing distance (approximately 45 cm) from the viewer. When viewing images either on the monitor or on print the viewing angle created is considered as 2 degree. Therefore, computing colour measurement data for print quality evaluation or colour management the CIE 2º (1931) Standard Observer is required. Figure 2 to the left illustrates the colour-matching functions for the CIE 1931 Standard Colorimetric Observer, expressed in terms of matching stimuli R, G, and B, with monochromatic primaries of 435,8 nm, 546,1 nm, and 700 nm respectively. Since R, G, and B were real primaries some of the values were negative. To ensure that tristimulus values for all real colours are always positive, the CIE adopted three unreal primaries X, Y, and Z. The transformed system is called the X, Y, Z system with colour-matching functions x (λ), y (λ), and z (λ) as shown in Figure 2 to the right. This system is often referred to as the CIE colour-matching functions for the CIE 1931 Standard Colorimetric Observer or the 2 Observer. 9

21 Colour Fundamentals Background Spectral tristimulus values b (λ) r (λ) g (λ) Wavelength Spectral tristimulus values z (λ) y (λ) x (λ) Wavelength Figure 2: To the left: Colour-matching functions for the CIE 1931 Standard Colorimetric Observer, expressed in terms of matching stimuli R, G, and B, with monochromatic primaries of 435,8 nm, 546,1 nm, and 700 nm respectively (Data set available by Rijgersberg [153]). To the right: CIE colour-matching functions for the CIE 1931 Standard Colorimetric Observer (full lines), and for the 1964 CIE Supplementary Standard Colorimetric Observer (dashed lines), (Data set available by CIE [31]). The tristimulus values X, Y, and Z for a given object, which is illuminated by a certain light source, can be calculated for the CIE Standard Observer by summing the products of these distributions over the wavelengths range typically from 380 nm to 780 nm, usually at 5 nm intervals. Figure 3 shows a schematic of the computing process. Light source Object CIE Standard Observer Tristimulus values Relative power Reflectance X X = Wavelength Wavelength Wavelength Tristimulus value X Y Z Figure 3: Data for the light source, the object, and the human observer are used to derive the tristimulus values CIEXYZ. These three numbers constitute the units of the first CIE colour space CIEXYZ, whose coordinates are referred to as tristimulus values. The calculation of the CIE 1931 tristimulus values X, Y, and Z is shown in the following equations: 10

22 Background Colour Fundamentals X = k Y = k Z = k k = x (λ)r(λ)s(λ)δλ y (λ)r(λ)s(λ)δλ z (λ)r(λ)s(λ)δλ 100 y (λ)s(λ)δλ (1) where X, Y, and Z are the CIE tristimulus values; x (λ), y (λ), and z (λ) are CIE 1931 Standard Observer colour-matching functions; S(λ) is the spectral power distribution of the light source; R(λ) is the spectral reflectance of an object and k is a normalising constant. Δ(λ) is the wavelength interval (e.g. 5, 10 or 20 nm). Note that by convention, k is determined in such a way that Y=100 when the object is a perfect diffuse reflector (for which R(λ) =1 for all λ), which implies that the absolute spectral power distribution for the light source (illuminant) is not required. Further, note that the X, Y, Z colour matching functions do not correspond to a set of physical primaries nor represent the actual response properties of the cones, but are linear transformations of the R, G, B colour matching experiment using real primaries. The replacement of the R, G, B primaries by the three imaginary primaries X, Y, Z was completed to avoid negative numbers in the colour specification and normalized to yield equal tristimulus values for the equi-energy spectrum. Furthermore, y (λ) was chosen to be equivalent to the luminous efficiency function for photopic vision, V(λ) CIE Standard Illuminants As seen above the perceived colour by a human observer is the result of the interaction between the colour-matching functions, the spectral reflectance characteristic of the object, and very importantly by the spectral power distribution (SPD) of the light source under which the sample is viewed. Therefore CIE has specified a range of standard illuminants (light sources) in terms of their spectral power distribution for industrial applications. In 1931 CIE defined the illuminant A (tungsten light) and in 1963 CIE introduced D50 (representing daylight) with a correlated colour temperature (CCT) of 5000K, and D65 (with a CCT of 6500K) [70]. Most important for the graphic arts and printing industry D50 is the common illuminant required for the appropriate viewing conditions mentioned in ISO 3664 [79]. 11

23 Colour Fundamentals Background Metamerism As discussed previously, the trichromatic properties (three colour receptors) of the human visual system means that the detection of all colours is reduced to three sensory quantities. The consequence of that characteristic is the fact that, although two colour stimuli have different spectral power distribution, the visual appearance can be identical. A visual colour match can for example be obtained between a colour stimulus on a printed substrate and a corresponding colour stimulus on a monitor display (e.g. soft proofing), although their spectral power distributions obviously are very different. This phenomenon, called metamerism, can be explained by the individual cone responses to the cumulative power energy across a broad range of wavelengths so that different mixtures of light across all wavelengths can produce an equivalent stimulation on the receptor and eventually result in the same tristimulus values or colour sensation. For further information on metamerism, see e.g. [25, 42, 71, 198]. In other words, when two samples with different spectra appear to match under one light source but are different under another light source, this phenomenon is known as metamerism, and the samples are known as a metameric pair. This can lead to unwanted practical problems, e.g. in the graphic arts industry, where image reproduction issues such as print and proof quality may be evaluated under inappropriate viewing conditions. Typically, a visual comparison between a print and a proof under office light condition can result in an unacceptable mismatch whereas under standard viewing conditions using D50 the comparison results in a visual match. However, even though the phenomenon may cause some problems sometimes, metamerism makes colour image reproduction practically possible by using three primary colours to obtain a colorimetric match rather than recreating the spectral reflectance of the original colour. In print quality and soft proofing the appropriate light source is required to predict the appearance. The CIE has published standards for assessing quality of light sources and has recommended the use of metamerism indices which quantify the degree of metamerism by changing the conditions [25]. Firstly, the illuminant metamerism index considers the colour difference between two samples, which match under one illuminant but will mismatch under another illuminant. CIE recommends a D65 illumination as the reference illuminant and the test illuminant can be chosen depending on the application. Typically ISO 3664 [79] specifies the viewing conditions for printed images as well as images displayed in isolation on colour monitors according to the metamerism index in accordance with ISO/CIE [95]. 12

24 Background Colour Fundamentals Furthermore, there are three other types of metamerism to consider. The observer metamerism index quantifies the colour difference degree in which the CIE standard observers and standard deviate observer do not agree in matching. Field-size metamerism is to consider in order quantifying the degree of colour difference of a sample pair by changing the viewing field size. For instance, two samples can be deemed identical in 2 degree viewing condition, but different under 10 degree viewing field. Finally, when the colour matching disappears by changing the illumination and viewing conditions the geometrical metamerism has to be considered, which usually is seen in metallic paint where two samples are identical in a given viewing geometry while different in other conditions [70] CIE 1931 XYZ Colour Space As seen above, the CIE X, Y, Z tristimulus values are fundamental measures of colour. Because they do not express immediately an obvious representation of colours nor do they have a direct perceptual correlation, the X, Y, Z tristimulus values can be transformed into a graphical system known as the CIE x, y chromaticity diagram, providing a twodimensional representation. In this system a colour is represented by its coordinates x, y, known as chromaticity coordinates, derived from the tristimulus values as follows: X x = X +Y + Z Y y = X +Y + Z Z z = X +Y + Z (2) The derived x and y values are normalized tristimulus values such that x + y + z = 1. The variables x and y are projective coordinates forming the chromaticity diagram, representing a visualization of the three-dimensional XYZ colour space onto a two-dimensional plane. The chromaticity of all colours can be plotted in the chromaticity diagram. But stimuli of identical chromaticity but different luminance are collapsed onto the same point in the twodimensional plane of the chromaticity diagram. Consequently, the luminance value must be quoted as a separate value (Y). The CIE xyy colour space is widely used to specify colours in practical application. In photometry, the tristimulus value Y has units either of lux (lx) which is lumens per square meter, or candelas per square meter, cd/m 2. The unit lux is in particular used to describe the concentration of luminous flux in a given direction, at a given point of a real surface (e.g lx in pressroom for critical comparison, P1) 13

25 Colour Fundamentals Background [79]. The unit cd/m 2 is intended to express how much light a device can generate (e.g. monitor display) [83] nm Spectral locus nm Purple boundary nm Figure 4: CIE x, y chromaticity diagram including the chromaticity coordinates of the real RGB primaries (forming triangle of dashed lines) corresponding to the colour-matching functions of r (λ),g(λ), and b (λ). Figure 4 illustrates the CIE x, y chromaticity diagram with the horseshoe-shaped outline, representing the chromaticity of monochromatic light called the spectral locus; with wavelengths shown in nanometers along the curve. Colours, which are less saturated appear in the interior of the diagram with white at the centre. The corners of the triangle (dashed lines) shows the location of the chromaticity coordinates of the real RGB primaries corresponding to the colour-matching functions of Figure 2. An important characteristic of the CIE x, y chromaticity diagram is the additive colour mixture calculation. If any two points on the chromaticity diagram are chosen, then all colours which lie on a straight line between the points, can be obtained by mixing these two colours. The straight line which connects the two end points of the spectrum locus is known as the purple boundary and presents the additive colour mixing of the monochromatic lights red and blue (violet). The diagram represents the gamut of all the chromaticities visible to the average person and all colours can be formed by mixing three sources and are found inside the area enclosed by the spectral locus and the purple boundary. 14

26 Background Colour Fundamentals Although the CIE x, y chromaticity diagram is often used in colour management, e.g. to represent certain primary colours (e.g. chromaticity coordinates of srgb) and to visualize the corresponding gamut, it suffers from a serious disadvantage. In fact, the distribution of the colours in the chromaticity diagram is not uniform. The consequence of this is that colour differences, represented by equal distances in the CIE x, y chromaticity diagram, are not perceived as being equal, as shown in Wright s and MacAdam s results. In Wright s experiment performed 1941, he shows that each line in the CIE x, y diagram represents a colour difference of equal proportion. However, the length of lines varies depending on the region of the diagram as seen in Figure 5 a). In a perceptual uniform colour space they are expected to have the same length. In 1942 MacAdam conducted visual experiments to prove the non-uniformity of the CIE x, y chromaticity diagram and to determine the perceptual distance within the CIE x, y chromaticity diagram [113]. He performed additive colour mixing experiments for 25 different test colours, establishing elliptical regions on the chromaticity diagram which contain all colours that are perceptual indistinguishable from the colour at the centre of the ellipse. According to his results, the size and shape of these regions vary over the chromaticity diagram. Although there are very small perceptible differences in the blue-purple colours region, in the green area of the CIE x, y chromaticity diagram, where colour differences are not distinguishable, the regions are approximately 10 times larger. Figure 5 b) demonstrates the different ellipses, which are called the MacAdam ellipses. Ideally, the MacAdam ellipses should all be circles of the same radius if the chromaticity diagram was perceptually uniform. To overcome these shortcomings a number of different diagrams have been proposed by squeezing or stretching the CIE x, y chromaticity diagram to provide a better correlation between the perceived colour difference and the corresponding distances in the diagram. 15

27 Colour Fundamentals Background a.) b.) Figure 5: a) CIE x, y chromaticity diagram with lines representing small colour differences. b) MacAdam (1942) ellipses plotted in CIE x, y chromaticity diagram (axes of the plotted ellipses are 10 times their actual lengths) Uniform Colour Spaces To correct the non-uniformity in the x, y chromaticity diagram, in 1976 the CIE standardized the u, v diagram as an approximately uniform chromaticity diagram whereby its axes are defined as follows: u'= v'= 4x 2x +12y + 3 9y 2x +12y + 3 (3) Figure 6 demonstrates the derived CIE standardized uniform CIE u, v chromaticity diagram in which a selection of lines seen in Figure 5 a) are shown again. It can be immediately observed, that the variation in the length of the lines has been much reduced. In fact, the ratio of the longest line to the shortest line is only about four to one, compared to the CIE x, y chromaticity diagram where the ratio is approximately twenty to one. 16

28 Background Colour Fundamentals Figure 6: CIE standardized uniform CIE u, v chromaticity diagram. Although the improvements have increased, the perceptual uniformity of the CIE u, v chromaticity diagram it is still only applicable to colours having the same luminance. In general, however, colours vary in both the chromaticity and luminance domain. Therefore CIE has introduced a further model combining these variables in another colour space CIELAB In 1976 a major development was made by the CIE introducing the CIELAB uniform colour space and colour specification intended to be used for surface colours. The CIELAB colour space is established as a colour-opponent three-dimensional space with the axes L*, a*, and b* forming a rectangular or Cartesian coordinate space. L* shows the dimension of lightness, while a*, b* refer to the colour-opponent dimensions red-green and yellow-blue, respectively [71]. The scale of lightness is from 0 to 100, with L* of 0 representing black and L* of 100 representing white as demonstrated in Figure 7. The L*a*b* coordinates are calculated using a non-linear transformation from the XYZ tristimulus values with the intention to provide a partial solution to both problems of colour appearance and colour difference in the new space. Hence, the aim of the non-linear transform of the tristimulus values in the CIELAB formula is to allow the Euclidean distance between two points to enhance the prediction of the visual colour difference between the colour stimuli represented by these two points [191]. 17

29 Colour Fundamentals Background L* = 100 = White E* ab +b* = Yellowish -a* = Greenish +a* = Reddish -b* = Blueish L* = 0 = Black Figure 7: Schematic diagram of CIELAB axis L*, a*, and b* (after Berns, page 71 [16]). From the obtained tristimulus values X, Y, and Z the L*a*b* coordinates can be calculated based on the equations: L*= 116(Y / Y n ) 1/3-16, for Y / Y n > , L*= 903.3(Y / Y n ), for Y / Y n , (4) a*=500[f(x / X n ) 1/3 - f(y / Y n ) 1/3 ], b*=200[f(y / Y n ) 1/3 - f(z / Z n ) 1/3 ], where f(i)= (I) 1/3, if I > , f(i)=7.787(i) + 16 / 116, if I , where X n, Y n, and Z n refer to the tristimulus values of the reference white point. Although the reference white can be specified according to the white point of the medium on which a colour is presented (relative colorimetry), for surface colours the values of X n, Y n, and Z n are usually calculated in accordance with the perfect reflecting diffuser and are therefore equivalent to the light source itself (absolute colorimetry). A further variant of the CIELAB colour representation is to calculate the polar coordinates of the perceptual attributes chroma C* ab and hue-angle h ab : * C ab = a *2 + b *2 (5) 18

30 Background Colour Fundamentals h ab = arctan b* a * (6) The CIE 1976 chroma C* ab predicts the chromatic content of a colour increasing from 0 (achromatic colour in the centre), in other words the distance from the L* axis as shown in Figure 8. Whereas the CIE 1976 hue angle h ab describes a colour s hue, which ranges from 0 to 360 following the colours of the spectrum. +b* C* ab -a* h ab +a* Figure 8: CIELAB a*, b* plane including chroma C* ab as the length of the vector, and the angle as the hue angle h ab CIELUV A further colour space was defined by the CIE at the same time as CIELAB and for the same purpose of uniformity, the CIE 1976 (L*u*v*) uniform colour space. CIELUV colour space is currently used for applications such as lighting, computer graphics, and other applications involving additive colour mixing. Its coordinates are defined by the transformation: L*= 116(Y / Y n ) 1/3-16, for Y / Y n > , L*= 903.3(Y / Y n ), for Y / Y n , (7) u*=13l*(u - u n ), v*=13l*(v - v n ), where u' and v' are the coordinates of the CIE standardized uniform CIE u, v chromaticity diagram, and the index n denotes the coordinates of the reference white. In the last couple of decades CIELAB has become almost exclusively used for colour specification especially for the graphic arts and printing industry. Furthermore, CIELAB is the source for the vast majority of work on the prediction of colour difference and also colour appearance. However, it is important to note that both the CIELAB and -b* 19

31 Colour Fundamentals Background CIELUV colour spaces are only approximately perceptually uniform and neither of them is generally considered as superior to the other. Moreover, there are still significant differences in the correspondence between the perceived colour difference and the Euclidean distance between two points, especially in the blue region of the CIELAB space [169]. Nonetheless, compared to the CIE XYZ tristimulus colour space, both of them offer significant improvements for print quality evaluation CIELAB Colour Difference In the context of print quality or process control it is required to measure and compare colours. The colour difference between two colour samples can be quantified by plotting their coordinates in the three-dimensional CIELAB colour space. The Euclidean distance of two points is defined by ΔE* ab as seen in Figure 7 and defines how well two samples match or how much they differ. Hence, two main points can be considered which a colour difference equation is aiming for. Firstly, a colour difference of unit one ΔE (ΔE=1) represents approximately a just noticeable difference (JND) for samples seen side by side under defined viewing conditions, i.e. the smallest colour difference which is possible to distinguish by the observer. Secondly, the magnitude of difference between the numerical values representing two colour samples should be proportional to the perceived colour difference between them [121] (see Figure 9). E=1 E=1 E=1 E=1 = E=2 = E=2 = = E=4 = Figure 9: Schematic illustration expressing colour differences (after Morovic s Figure on page 27 [121]). 20

32 Background Colour Fundamentals The equation describing the Euclidean distance between two sample points in the CIELAB colour space is, ΔE* ab = [(ΔL*) 2 + (Δa*) 2 + (Δb*) 2 ] 1/2, (8) where ΔL* = L* sample - L* target Δa* = a* sample - a* target Δb* = b* sample - b* target ΔL* > 0 means sample is lighter than target ΔL* < 0 means sample is darker than target Δa* > 0 means sample is redder than target Δa* < 0 means sample is greener than target Δb* > 0 means sample is yellower than target Δb* < 0 means sample is bluer than target Considering industrial applications of colour science it is common that one of the samples is a target and the other one is a sample that is supposed to be a visual match to the standard, as discussed in the publications in PART II of the present thesis. An alternative formulation of the above CIE 1976 colour difference is the computation of the polar coordinates, writing an equivalent formula in terms of lightness difference, ΔL*, chroma difference, ΔC* ab, and hue difference, ΔH* ab, ΔE* ab = [(ΔL*) 2 + (ΔC*) 2 + (ΔH*) 2 ] 1/2, (9) where ΔC* = C* sample - C* target and ΔH* ab is know as the CIE 1976 a, b hue-difference, expressed as, ΔH* ab = [(ΔE*) 2 - (ΔL*) 2 - (ΔC*) 2 ] 1/2, (10) The total colour difference, ΔE* ab can be divided up into parts of ΔL* ab, ΔH* ab, and ΔC* ab. Essentially these components correlate with the perceptual attributes, lightness, hue, and chroma whose squares sum to the square of ΔE* ab. Note that although the simple difference between two hue angles can be calculated, Δh* ab does not have the property of being a perceptual attribute. ΔH* ab is computed as what is left over once the lightness and chroma differences are subtracted from the total colour difference. 21

33 Colour Fundamentals Background These simple equations have been and are still used effectively to quantify colour difference in a wide range of applications, in particular the graphic arts and printing industry where process tolerances in international standards are still mostly defined as CIE 1976 a, b colour differences [85-87]. It is important to note that these measures only correlate approximately with the corresponding attributes. Furthermore, the perceived colour is always depending on viewing conditions, including illumination intensity, the colour of the illumination, and the characteristics of the surroundings. However, how do we interpret or classify the ΔE* ab scale? When considering color differences two types of visual assessments are most prevalent perceptibility and acceptability [16]. First, perceptibility thresholds indicate what magnitude of colour difference is a just noticeable difference (JND). Colour difference equations are set such that their units correspond to JND s hence it is commonly stated that any colour difference below 1 unit is predicted as not being perceptible for samples viewed side by side [16, 38, 71, 101]. However, in a study by Mahy et al. [114] the authors found a JND of ΔE* ab =2.3. A rule of thumb for practical classification of a perceptibility threshold which corresponds with the ΔE* ab scale, when two colour samples are viewed side by side is proposed by Schläpfer [161] where the author classifies a ΔE* ab < 0.2 as "Not visible", ΔE* ab between 0.2 and 1.0 as "Very small", ΔE* ab between 1.0 and 3.0 as "Small", ΔE* ab between 3.0 and 6.0 as "Medium" and ΔE* ab > 6.0 as "Large". Second, the acceptability threshold can be seen as a less concrete concept and one that depends strongly on application and industry. For example in the work of Hardeberg [64] a rule of thumb for practical interpretation of a ΔE* ab is proposed where errors of ΔE* ab <3 are classified as Hardly perceptible, 3 ΔE* ab < 6 is defined as perceptual, but acceptable and ΔE* ab > 6 as Not acceptable. Sharma [167] states that ΔE* ab between 4 and 8 is generally deemed acceptable in prepress and colour imaging. In another study by Stokes et al. [179] an acceptability threshold of approximately ΔE* ab = 6 was found for their experimental images and observers. We note the disagreement between these classifications because the evaluation of quality and acceptability is highly subjective and depends greatly on the experiences and expectations of observers as well as the application for which the colour stimuli are intended. A further quality metric, which could be more suitable for the printer market, would be a tolerance threshold and this issue could be the subject of future research. 22

34 Background Colour Fundamentals As mentioned previously, the perceived magnitude of the colour difference does not entirely agree with the Euclidean distance in CIELAB colour space. Therefore several attempts have been made to improve the inconsistency between the perceived colour difference and the Euclidean distance and hence several alternative colour difference equations have been proposed. Because CIELAB has become a very common and widely used colour space most of the alternative equations are based on that colour system. However, the calculation of the distance between two colours is not the Euclidean distance itself. To correct the non-uniformity in parts of the CIELAB colour space the colour difference equations firstly decompose the Euclidean distance into components corresponding to the differences in the colour attributes of lightness ΔL*, hue, ΔH*, and chroma ΔC* respectively. Essentially, the colour attributes are weighted differently to define a new colour difference which corresponds better to the perceived colour difference. In 1984, Clark et al. [32] (members of the Colour Measurement Committee of the society of Dyers and Colourists) published one of the first advanced difference formula known as CMC(l:c) providing among other parameters scaling functions for lightness, L*, and chroma, C*. This equation has been extensively used but it was never adopted as a CIE standard [191]. The complexity of the CMC formula has been criticised and after analyzing a large set of psychophysical data it was suggested that simple weighting functions, S L, S C, and S H, would be sufficient to improve the perceived colour difference. Hence, CIE proposed a new formula, which is known as CIE94 [26], and the colour difference ΔE* 94 is calculated by ΔE * 94 = ΔLab * klsl 2 + ΔCab * kcsc 2 + ΔHab * 2 khsh 1/2 (11) where k L = k C = k H = 1 S L = 1 for reference conditions S C = C* ab,standard S H = C* ab,standard The variables k L, k C, and k H are parametric factors to correct for experimental viewing and illumination conditions. Except for the textile industry, CIE 94 (2:1:1), a value of one is recommended for all parametric factors [70]. 23

35 Colour Fundamentals Background The weighting functions S C and S H are calculated using the chroma value of the standard sample specimen C* ab, Standard. When neither sample can be deemed as standard, the geometric mean chroma of the two samples should be used [191]. Traditionally ΔE has been measured as a spherical distance. A more realistic measure is based on an elliptical volume to improve the correlation between the visual and numerical colour difference. Therefore, the goal of the weighting functions S C and S H in ΔE* 94 is to scale down hue and chroma differences for higher chroma colours in comparison to the traditional ΔE* ab and thus correct for the predominant deficiency in CIELAB. For a neutral colour along the L* axis the scaling factor is 1, which means that the colour differences ΔE* 94 are identical to the ΔE* ab Euclidean colour difference. On the other hand, the weighting functions S C and S H are greater than unity for chromatic colours. Hence, the results of ΔE* 94 become smaller than the results of ΔE* ab Euclidean colour difference [169]. Additional attempts have been made to adjust the colour difference equation to further improve the uniformity, and in 2001 a new colour difference formula, CIEDE2000 (ΔE* 00 ), was proposed [111] and adopted by the CIE [29]. In addition to the weighting functions for lightness, chroma, and hue, as already used in ΔE* 94, the ΔE* 00 colour difference equation includes a number of further parameters to compensate for the nonuniformity of the CIELAB colour space. Among the new parameters, a term for improving the performance in the blue colours and a rescaling factor of the CIELAB a* axis improving the performance for colours close to the L* axis, can be mentioned. For further details and the equations the reader is referred to Luo et al. [111]. It is important to notice that even though small noticeable improvements compared to ΔE* 94 could be made in some parts of the CIELAB colour space the complexity of the ΔE* 00 formula may reduce the performance of the colour difference prediction in other parts of the colour space. Additional concerns with regard to the CIEDE2000 colour difference formula are given by Kuehni and Luo [106, 112]. In a later study by Granger, the author even claimed systematic errors in the CIEDE2000 formula [57]. Nevertheless, in a recent study by Pant and Farup, the authors conclude that, though the CIEDE2000 significantly improves the visual colour differences, orientation problems have been observed in the ellipses in the blue and red region [142]. Therefore, further research is required to improve the rotation term in ΔE* 00 or to solve the concerns with the colour difference equations in general. In general, regarding the concerns mentioned with colour difference equations, there are several fundamental questions associated with the CIELAB colour space itself [107]. 24

36 Background Colour Fundamentals Although the CIE system of colorimetry has been used successfully for almost 80 years it is important to note that the current model is limited to the comparison of samples which are identical in spatial respect and viewed under equal viewing conditions including chromatic adaptation, light adaptation, luminance level, background colour, and surround colour [42]. However, with these restrictions in mind, the model serves sufficiently for the upcoming discussions and definitions in the present thesis. 25

37 Colour Fundamentals Background 26

38 Background Colour Measurement 2.2 Colour Measurement As we have seen previously, there are three factors needed for colour perception (unless we have coloured light, then the three factors are reduced to two, the light and the observer, respectively). Firstly, as illustrated in Figure 1, a light source is illuminating the object and consequently an observer detects the reflected light from the object and converts the signals into a response, which the human brain interprets as a colour. In sensory substitution of the human sensor by an instrumental colour measurement device, the goal of the instrumental colour measurement is to estimate what an observer sees, the colour appearance of a stimulus [16]. In the graphic arts and printing industry colour measurement instruments are used primarily for determining the colour characteristics of objects, such as prints and proofs, and for measuring the behaviour of colour reproduction input and output devices. In the context of colour reproduction device calibration and characterization colour measurement is a very important task determining the appropriate parameters and verifying the behaviour of the devices. In colour printing process control the goal of obtaining colour measurements is to verify the specified target or reference values to ensure the repeatability of the process and the ink uniformity across the printing substrate. Colour measurement instruments include both filter-based instruments that directly measure colorimetry and devices, which are designed to measure spectral reflectance or transmittance such as spectrophotometers and instruments intended to measure spectral radiance or irradiance such as spectroradiometers. In the next section we briefly address the most important measurement instruments used in the graphic arts and printing industry and that are also applied in this work. For more comprehensive details on colour measurement devices the reader is referred to literature by Berns [16], Battle [14], Hunt [70], Sharma [169], Schanda [159], Völz [187], Wyble [195] and CIE publication 130 [28] Measurement Instruments and their Application In this sub-section different types of (colour) measurement instruments are presented with respect to their viewing and illumination properties defined by CIE. For completeness, other types of measurement instruments used in the graphic arts and printing industry are also presented and discussed. 27

39 Colour Measurement Background Tristimulus colorimeter The simplest device to measure colours is a tristimulus colorimeter whose spectral responsivity mimics the colour matching functions of the Standard Observer [160]. It measures, as suggested by its name, colour tristimuli and reports these as colour values in CIEXYZ, and CIELAB. The key elements include a light source (for measuring materials) usually illuminant C or D65, a set of colour filters and commonly a silicon photodiode as a photo-detector. The spectral separation is obtained by either using colour filters placed in front of the photodiode or by using spectrally different light sources to illuminate the sample, often using 45 :0 geometry, i.e. circumferential 45 illumination and normal viewing [25]. To be able to report CIE colorimetry the sensitivities of the colour filters are intended to have a close match to a set of colour matching function. As seen in Figure 10, for filter based colorimeters, four independent filters are usually used, two approximate the y (λ) and z (λ) colour matching functions and two approximate the two humps of the x (λ) colour matching functions [169]. Photodiodes Colour Filter Display CIEXYZ λ λ λ Standard Observer response Light source 0 Light source 45 Sample Figure 10: Basic features of a tristimulus colorimeter. In general, the tristimulus colorimeter is easy to use, it is fast and inexpensive, but does not provide any detailed spectral information. Nevertheless, colorimeters are often used in the graphic arts industry to calibrate and characterize self-luminous sources, such as monitor displays, although there might be some performance restrictions in terms of accuracy due to the aging of the colour filters and poor reproducibility to agree with the 28

40 Background Colour Measurement CIE colour matching functions [8, 14]. A study by Gardner compares different calibration methods for tristimulus colorimeters to reduce the measurement uncertainties [51]. It is also important to note that for surface colours, the colorimetric values measured are valid for one illuminant only, the instrument light source, but it is often desirable to know the tristimulus values or colour difference for different illuminants. Thus a colorimeter cannot give any indication of metamerism [70]. Spectrophotometer Nowadays, the most common colour measurement instrument in the graphic arts and printing industry is the spectrophotometer which measures the ratio of reflected to incident light (the reflectance) from a sample at many points across the visible spectrum, that is, as seen in Section 2.1.1, between about 380 nm and 780 nm. Reflectance = Reflected light / Incident light The main components of all spectrophotometers for colour measurement include a light source (for measuring materials), a wavelength selection device, and a photo detector as shown in Figure 11. Reflectance Factor R Detector Array Display Light source Lens 45 0 Sample Aperture Diffraction Grating Figure 11: Main measurement components of a 45 x:0 geometry spectrophotometer. Because we are measuring the ratio of incident to reflected light with a spectrophotometer, the type of light source illuminating the sample should not matter. However, as certain paper materials contain a fluorescent whitening agent (FWA), they absorb energy at one part of the spectrum and re-emit at another. Thus the spectral power distribution (SPD) of the light source can have a dramatic effect on the measured colour. A paper published by 29

41 Colour Measurement Background Andersson and Norberg [3] points out how important the UV-content of the illuminant can be when carrying out colour measurements on prints containing FWA; even small changes in UV-content can affect the measured reflectance spectra. Pulsed xenon, quartz halogen and tungsten gas filled, type A lamps are common light sources for spectrophotometers [16]. The reflected or transmitted light is passed on to the wavelength selection device or spectral analyser, where the light is split into its spectral components to be measured. In a spectrophotometer, prisms, gratings, and interference filters are the technologies used to separate light into narrow bands. Prisms and gratings both separate the wavelength spatially. Although both methods of dispersing light can still be found, today most spectrophotometers use diffraction gratings. Light passing through the grating will be diffracted at a fixed angle, which is dependent on its wavelength and the dispersed light is focused directly onto the photo detector array. Most spectrophotometers work in the range 380 nm nm with sampling intervals reporting reflectance at 5-, 10-, or 20-nm intervals. CIE geometries for illumination and viewing for reflection The illumination and viewing geometry of a colour measurement instrument is very important and can affect the way a colour is measured. The CIE has defined some standard terms and geometrical conditions denoted by symbols, such as d:8 or 45 :0. The symbol d indicates the use of an integrating sphere, and the numbers before and after the colon indicate the angles of incidence and viewing, respectively. For instruments using an integrating sphere a glossy sample can be measured in two conditions either with the specular component included or excluded. Hence, the geometry is distinguished by adding the letter e for excluded or i for included. The current CIE [25] recommendations for reflection measurements with diffuse geometries: Diffuse Geometries: Diffuse: eight-degree geometry, specular component included (di:8 ) Diffuse: eight-degree geometry, specular component excluded (de:8 ) Eight degree: diffuse geometry, specular component included (8 :di) Eight degree: diffuse geometry, specular component excluded (8 :de) Alternative diffuse geometries (d:0 ) and (0 :d) 30

42 Background Colour Measurement Diffuse/diffuse geometry (d:d) Directional Geometries For 45 geometries, the CIE distinguishes between an annular and a directional arrangement by adding the letters a or x, respectively. Forty-five degree directional/normal geometry (45 x:0 ) Forty-five degree annular/normal geometry (45 a:0 ) Normal/Forty-five degree directional geometry (0 :45 x) Normal/Forty-five degree annular geometry (0 :45 a) With these CIE recommendations the 45 x:0 geometry replaces typically the 45 /0 geometry. The 45 geometry spectrophotometer is widely used within the graphic arts and printing industry, where the sample is illuminated at an angle of 45 and the measurement takes place at 0 (perpendicular to the sample), see Figure 11. CIE has also recommended six different standard geometries of illumination and viewing for transmitting material. For further details on measuring geometries we refer to Schanda [159] and Otha and Robertson [139]. It is important for an operator to understand the critical parameters of an instrument. Therefore CIE has worked out a terminology (first published in 2004) that makes the definition of the measurement geometry easier. CIE has recommended that the manufactures follow these new guidelines. However, most of the largest manufactures of colour measurement instruments use still (after six years) the old terminology (e.g. reflection 45 /0 ). Spectrocolorimeter ASTM E1347 [7] defines a spectrocolorimeter as a spectrometer which has a dispersive element (such as a prism, grating, or interference filter) instead of RGB filters that is normally capable of producing as output only colorimetric data (such as tristimulus values and derived colour coordinates) but not the underlying spectral data from which colorimetric data are derived. According to Rich [151] a better characterization accuracy of the colour and colour difference of metameric pairs can be obtained. This is achieved by sampling and digitizing the standard illuminant and the measured object at discrete points, along the visible spectrum, and eventually to calculate the tristimulus values. 31

43 Colour Measurement Background Spectroradiometer Radiometry is the science of measurement of radiant energy, including light with respect to absolute power. Spectroradiometers are designed to measure radiometric quantities (irradiance 2 and radiance 3 ) in a narrow spectral bandpass as a function of wavelength. They have the same principal components as the spectrophotometer, except for the light source. They are designed to measure the spectral properties of light sources, such as viewing booth, monitor displays, and projectors. The tele-spectroradiometer (TSR) is the most frequently used instrument in this category. The key components of the TSR are a telescope, a monochromator, and a detector. As for most of the spectrophotometer the measurement range interval of a TSR is between 380 nm and 780 nm. But the sampling interval of the TSR often has a spectral resolution ranging from 1 nm to 10 nm, which is sufficient for colour work using e.g. fluorescent lamps as sources (some light sources have monochromatic emission lines in their spectra) [169]. The advantage of the TSR is the measurement of the object at a distance corresponding to its actual observing position including common viewing distance and viewing conditions. In other words it can measure all forms of colours (surface and selfluminous), which are particularly important for cross-media reproduction, e.g. to match an image displayed on a monitor to the output from a printer in the viewing booth (e.g. verifying typically a soft proofing set up). However, due to the high costs and complexity the TCR is not a common measurement device used in the graphic arts and printing industry. Abridged Spectrometers According to Berns [16] abridged spectrometers are instruments that sample the visible spectrum at several discrete bands by e.g. replacing the broadband coloured filters with narrowband interference filters with approximately 50 nm bandwidth, resulting in seven signals, and via matricing L*,a*,b* coordinates can be estimated. An alternative approach is to use a monochromatic sensor and a series of different coloured light-emitting diodes (LED s) arranged in a mosaic forming 45 circumferential illumination. Each type of LED can be turned on one-at-a-time to generate a sequence of signals which are detected at 0 and recorded. A matrix converts the multi-signal into colorimetric coordinates. 2 Radiant flux incident per unit area (ASTM 284 E) 3 Radiant flux emitted per unit projected area of a source (ASTM 284 E) 32

44 Background Colour Measurement Imaging system Another interesting group of capturing devices to record colours are colour scanners and digital colour cameras. Although most of these three-signal imaging systems do not have spectral sensitivities to obtain directly colorimetric values, models are available to estimate device independent values from device dependent RGB camera response taken from the coloured sample. For instance Brydges et al. [22] and Hong et al. [69] used a CCD colour camera for estimating colorimetric and densitometric measurements based on polynomial modelling. Even though the model performs well in terms of the accuracy, it is dependent upon both the media and colorant. In other words, applying the model to other media/colorant combinations can lead to serious eye-camera metamerism problems [140]. Instead of predicting CIELAB values from camera RGB the spectral distribution can be estimated. This approach offers different types of post-processing, e.g. simulating the colour coordinates using different illuminations. A study by Solli et al. [173] predicts the reflectance spectra from camera RGB measurements. More sophisticated models to estimate spectral reflectance from trichromatic camera systems are proposed by Imai et al. [74]. In a recent study by Seymour [164] the author provides a review of technical papers on the subject of the accuracy of the camera RGB to CIELAB transformation. According to his findings, the accuracy reported is still considerably worse than one would like from a colour measurement device. However, he is concerned about the accuracy of reporting changes in colour within a printing process. He is proposing the use of CIELAB-like measurements with the camera taken from a print that is considered as the reference. Eventually the deviations from these initial values could then be used for process control. Photometer Light sources radiate energy in the form of electromagnetic waves. Photometry is the science of the measurement of light, with respect to its perceived brightness to the human eye. The photometric quantities are related to the corresponding radiometric quantities by the luminous efficiency function V(λ) of the CIE Standard observer that models human brightness sensitivity. Four basic photometric quantities are defined, namely the luminous flux, luminous intensity, illuminance, and luminance. The appropriate definitions of the terms and units are provided by ASTM E 284 [4] and ISO 3664 [79]. Luminous flux: Flux as the light energy and luminous flux as the measure of the flow of light energy emitted by a source, or received by a surface. The quantity is 33

45 Colour Measurement Background derived from the radiant flux, W (in Watts), by evaluating the radiation in accordance with the relative luminous efficiency function V(λ) of the CIE Standard observer. The unit is lumen (lm). lm = 683 x W x V(λ) Luminous intensity: Expresses the power of a light source. It is defined as the quantity of luminous flux emitted in a given direction per solid angle (in steradian). The unit is candela (cd). 1 cd = 1 lumen per steradian. Illuminance: Illuminance (E) is a measure of the concentration of luminous flux falling upon a surface and is expressed in lumens per unit area. The unit is lux (lx). 1 lx = 1 lumen per square meter (lm/m 2 ). Luminance: Luminance (L) is a measure of the flux emitted from, or reflected by, a relatively flat and uniform surface. The unit is candelas per square meter (cd/m 2 ) and is also known as photometric brightness. For in-depth information on radiometric and photometric terms and units we recommend the reader to review Hunt s book Measuring Colour, Appendix 1 [70]. Essentially, a photometer instrument has the same geometric specification as a spectroradiometer and measures either the illuminance E or luminance L of light sources. It is built similar to a colorimeter except that it only has a single channel. The tristimulus value Y of a colorimeter reading is matching the y (λ) function of the 1931 standard observer and by definition reads E and L. 34

46 Background Colour Measurement In the field of printing process control further measurement instruments have to be briefly mentioned and addressed, measuring densities, dot area and gloss. Optical densitometer The optical density is the degree to which reflective surfaces absorb light, or transparent surfaces allow the light to pass. The density is given logarithmically by, D ink = log 10 I i I m (12) where I m is the reflected light intensity and I i is the intensity of the incident light [15]. High-density values correspond to high light absorptions. Density (absorbance) measurements can be performed by the use of densitometers and it is often applied to control colours in the printing process. A reflection densitometer consists of a light source to illuminate the sample, optics to focus the light, filters to define the spectral response of the sample and a detector to monitor the reflected light from the surface of the sample. The sample is viewed at 45 from the surface and the reflected light is converted to density with a logarithmic amplifier and displayed digitally. Murray [127] expressed the relationship between the reflection density of halftone prints and the dot area R, known as the Murray-Davies equation: R = 1 10 D R 1 10 D H x100% (13) where D R is the density of the sample and D H is the solid ink density. Even though the revision of the ISO standard for process control for the production of half-tone colour separations for offset printing processes [85] is focusing mostly on colorimetric parameters, the density control still has a minor part of the standard as quoted from ISO (section , Note 4): Density values can be very valuable for process control during a print run, where the instrument, the ink and the print substrate remain the same; see ISO [88]. However, in a general situation, density values do not define a colour to the required degree. Therefore, for the purpose of this part of ISO 12647, reflection density values are only recommended for the determination of tone values. For newspaper printing the ISO , Table A.1 of the informative Annex A [86] gives reflection densities values provided for the solid process colours on newsprint. It is still very common in the printing industry that the press operator first is calibrating the solid colours according to the target values, consequently the densities obtained with the instrument from the OK print are recorded and the densities are then 35

47 Colour Measurement Background used as target values for process control during the production run. Although colorimetry makes its way into the pressroom and the revised standards specify the range of acceptable colour of the four process solids in terms of ΔE* ab, rather than in the traditional density tolerances optical densitometers are still very often used in daily production. A study by Seymour [165] considered the relation between ΔE* ab and Delta density, and concluded that densitometry and colorimetry are equivalent in terms of maintaining consistent colour on press. Electronic planimeter In planimetric measurements, the dot area coverage is measured by using instruments, including the basic components of a microscope and a CCD imaging sensor. Such devices, designed for measuring printing plates, are often called dot meters. According to Romano [158] the major variables in such instruments are the image capturing system, aperture selection and thresholding process. The dot meter analyzes the digital image according to the mage histogram and decides what area coverage is a part of the dot based on a certain defined threshold. The dot meter is actually measuring the dot area by taking a camera snap-shot and provides an absolute value of dot coverage depending on the counts of the number of black and white pixels in the image [34]. Research work has been done by Wroldsen et al. [193, 194] (Paper E) to investigate whether it was possible to estimate halftone values measured by a densitometer, from the halftone values measured by different dot meters. Due to the poor repeatability performance for dot meters applied to newspaper print the applied model did perform rather poorly. Hence the authors confirmed the use of dot meters for measuring printing plates only. Glossmeter Gloss is an optical property of a surface and is characterized by its ability to reflect light. The reflected light intensity from a sample can be measured by a glossmeter. The measured intensity is dependent on the material and the angle of the illumination. If a light beam strikes a non-metallic surface (coatings or plastics), the amount of reflected light increases with the increase of the illumination angle. Part of the illuminated light will penetrate into the layer of paint and the remaining light will be reflected. The resulting amount of reflected light from a sample is given in relation to that reflected from a black glass calibration standard with a defined refractive index. The gloss value of the reference standard is defined to be 100 gloss units (GU). 36

48 Background Colour Measurement Depending on the application different illumination angles are used. The gloss characteristics of different paper type substrates are classified in ISO and are in accordance with ISO , TAPPI method 75 [80] Sources of Error and Measurement Uncertainty When selecting a colour measurement instrument a number of factors might be considered. Most of the applications require spectral data, which provide information about the properties of raw materials, content of fluorescent whitening agents (FWA), potential problems with metamerism, and the ability to compute colorimetric data independently of light source. Furthermore the selection of the measurement conditions (radiance, irradiance) and the appropriate geometry is significantly determining the measurements [16, 151]. The suitable software to calculate and read out the measurements, service, and support agreements are further important factors to consider. An additional aspect is the question about the reliability of the measurement data obtained from a certain sample and the procedure to minimize the measurement uncertainty. Uncertainty of measurement is an estimation of the dispersion of the values within which the true value lies [138]. It is known that two models of spectrophotometer will produce different results from measurements on the same sample due to different instrument design in the area of the instrument s measurement system including the measurement geometry, the aperture size, measure position and filter used, light source, dispersion function, the scanning interval and the bandwidth. Depending on the properties of the measured samples (texture, gloss or FWA) these differences in the measurement system of the instrument can affect the measurements [14]. The instrument calibration and verification procedure is contributing to reducing the source of error as demonstrated by Briggs et al. [21] where the authors discuss reliability issues for colour measurement in quality control applications. They conclude that the measurement accuracy is an important aspect of any system for colour measurement and that the calibration procedures can significantly reduce colour differences. A study by Zwinkels [200] demonstrates the steps in calibrating and verifying the performance of colour measurement instruments. Furthermore, the author emphasizes the importance of specifying the procedures of spectrophotometric measurement on surface colours with 1) the quantity measured (reflectance or transmittance), 2) the geometry of illumination and viewing, 3) backing material used, 4) aperture size, and 5) the white reflectance standard for instrument calibration and its traceability. Berns [16] 37

49 Colour Measurement Background defines the calibration as the process of adjusting an instrument in such a way that its readings reproduce a national or international scale. Furthermore, the verification is the process of assessing an instrument s ability to precisely and accurately reproduce, national or international measurement scales. According to Berns [16] measurement uncertainty can be divided into precision and accuracy. Precision describes the dispersion of the measurements taken and accuracy describes the conformance of a series of measurements to the accepted or true value. Precision can be further divided into repeatability and reproducibility. ASTM E2214 [9] specifies repeatability as a measure of how well an instrument repeats its reading of the same sample over a certain period of time (e.g. short-term, medium-term, and long-term repeatability). The instrument s repeatability is quantified by calculating the colour difference between the mean of the measurements and each individual measurement taken (mean colour difference from the mean (MCDM)). MCDM = ΔE i /N (14) where ΔE i is the colour difference between each measurement and the mean of the measurements and N is the number of measurements. Reproducibility is a form of repeatability in which one or more of the measurement parameters have been systematically changed such as the sample is different, the time frame of measurements are very long, the procedures or instrument are different or the operator has changed. A report by Scorer et al. [163] defines procedures to be used to monitor the performance of an instrument in terms of its repeatability and reproducibility. Inter-instrument agreement describes the reproducibility of two or more instruments of identical design. On the other hand inter-model agreement expresses the reproducibility of two or more instruments of different design. In other words, reproducibility determines the variations between instrument s readings. In instruments, which have the same design, the random amount of bias is reduced compared to instruments with different design. A comprehensive set of works was carried out by Wyble and Rich addressing the evaluation of methods for verifying the performance of colour measurement instrument. Part I is addressing the repeatability issue [196] and part II is covering inter-instrument reproducibility [197]. Accuracy is affected by systematic errors, which are characterized by the fact that they remain fixed when the measurement is repeated under the same conditions. In a report published by the National Physical Laboratory (NPL) [33] spectrophotometric errors 38

50 Background Colour Measurement affecting the measurement results are discussed. According to the author, possible uncertainty sources are defined such as: Scale uncertainty (usually via instrument reference standard, e.g. white tile standard) Photometric linearity Dark level uncertainty Wavelength uncertainty Bandwidth uncertainty Gloss trap or specular beam uncertainty, for integrating sphere geometries Repeatability uncertainty It is beyond the scope of this introduction to discuss in-depth the details of these systematic errors. Therefore we refer to the valuable references [33, 186, 200]. A work by Robertsen [155] is addressing the issues of systematic errors and inter-instrumental agreement and proposes a method to assess the severity of these errors. Berns and Petersen [17] have further redefined and extended the proposed method. In another study by Rich et al. [152] maximum colour differences of up to 4.0 CIELAB units were observed between pairs of instruments. Similar pairwise colour differences were reported by Nussbaum et al. [136] (Paper F) on BCRA tiles. But on paper substrate the inter-instrument agreement was even higher in terms of colour differences (ΔE* ab > 6). Another source of error can be thermochromism which is the change of colour with changing temperature. In fact, the instrument s light source should not cause the sample to change temperature [33]. A study by Fairchild and Grum [43] demonstrates the effect on highly coloured samples, such as red, orange or yellow having spectral distributions with steep slopes, and propose a correction method. Various other studies and research addressing measurement uncertainties and comparative studies of colour measurement instrumentations have been presented (see e.g. Billmeyer [18, 19] and Rodgers et al. [157]). Several attempts have been made to correct the measurement data from different instrument readings applying mathematical models to improve the inter-instrumental agreement. De Garcia et al. [35] evaluate the colorimetric behaviour of different spectrophotometers. A study by Steder et al. [174] is assessing the selection of training data in terms of which and how many samples are needed for training the model while maintaining a good performance. Correction models applied to the spectral data have been proposed by Berns and Petersen [17] and Chung et al. [24] to improve the interinstrumental agreement on textile. Nussbaum et al. [133] (Paper G) demonstrate a 39

51 Colour Measurement Background correction model applied to CIELAB measurement data set to improve the colorimetric performance and hence the inter-instrument and inter-model agreement in printing applications. For measurement and calculation procedures for self-luminous video display devices, such as CRTs and flat-panel displays, other standards or recommendations are provided including CIE 122 [27], the IEC [73], and the ASTM standards E1336 [6] and E1455 [8]. These specifications present measurement procedures as well as measurement instrument characteristics in particular for spectroradiometers and tristimulus colorimeters. 40

52 Background Image Reproduction and Colour Management 2.3 Image Reproduction and Colour Management Having reviewed the colour fundamentals, and discussed methods and instruments for measuring colour and light, this section is aiming to give an overview on image reproduction and colour management. A wide range of comprehensive literature and textbooks providing details on image reproduction and colour management exist and the reader is referred to literature of e.g. Giorgianni and Madden [56], Green [58, 60], Morovic [121], G. Sharma [169], A. Sharma [167], and Adams et al. [2] Additive or Subtractive Colour Mixing Additive colour systems produce colour on a dark background by the combination (adding) of different wavelengths, known as primary colours. The term additive is used to signify the fact that the final spectrum is the weighted sum of the spectra of the individual lights. Typically, the additive primaries are red, green, and blue (RGB) and the combination of red and green forms yellow, red and blue forms magenta, and blue and green forms cyan as illustrated in Figure 12 a). The combination of all three primaries at full intensities produces white and the intermediate colours are obtained by varying the individual primary intensities. In contrast, subtractive colour reproduction is used for transparent and reflective media, and is produced by removing (subtracting) unwanted spectral components from white light. By deposing different colorants onto these media different colours are generated by selectively absorbing light of certain wavelengths while transmitting other wavelengths. The most common subtractive colorants are based on cyan, magenta, and yellow (CMY) that absorb light in the corresponding spectral regions of red, green, and blue, respectively. Roughly speaking, each colorant absorbs its complementary colour and transmits the rest of the visible range of the spectrum. In other words, the individual CMY colorants eliminate RGB spectral regions, respectively, as illustrated in Figure 12 b). The combination (overlay) of cyan and magenta eliminates both red and green, producing blue; the mixture of cyan and yellow eliminates red and blue, producing green; and the combination of magenta and yellow eliminates green and blue, producing red. The combination of the maximum amounts of all three produces black, and by varying the colorant amounts the intermediate colours are produced. 41

53 Image Reproduction and Colour Management Background K W R C Y W M B K G G C B M R Y a) b) Figure 12: a) The additive mixing of RGB primaries and b) subtractive colour mixing of CMY. It is common to add a fourth black (K) colorant in subtractive systems to improve the reproduction of achromatic (grey) colours and to extend the gamut by producing darker colours. Halftone colour printing can be considered as a hybrid system because the colorants combine subtractively, but the perceived colour is the average of the differently colored regions over a small area [170] Image Reproduction In the context of image reproduction the original image can originate from multiple sources. Images of real objects in real scenes exist as spatial variations of spectral distribution of radiance and reflectance. To record the real colour information, the scene has to be captured with a colour reproduction device (medium) such as a digital camera, by sampling the colour information both spatially and spectrally. Consequently, the recorded image can be processed and reproduced by using additive or subtractive colour mixing with a particular set of primary colours. Before commencing this introductory discussion on the principles of image reproduction and colour management it is worth clarifying some issues of terminology. In the CIE 156 publication [30] Guidelines for the evaluation of gamut mapping algorithms, the key components of image reproduction and their terms for image, colour reproduction medium, colour gamut, and colour gamut mapping are defined as follow: Image: Two-dimensional stimulus containing pictoral or graphical information whereby the original image is the image to which its reproduction is compared in terms of some characteristics (e.g. accuracy). Colour reproduction medium: A medium for displaying or capturing colour information, e.g. a monitor display, a digital camera or a scanner. Note, that in the 42

54 Background Image Reproduction and Colour Management case of printing, the colour reproduction medium is not the printer but the combination printer, colorants and substrate. Colour gamut: A range of colours achievable on a given colour reproduction medium (or present in an image on that medium) under a given set of viewing conditions it is a volume in colour space. The colour gamut of an image or colour reproduction medium can be represented in a three-dimensional color space such as CIEXYZ or CIELAB. Colour gamut mapping: A method for assigning colour from the reproduction medium to colours from the original medium or image (i.e. a mapping in colour space). Figure 13 illustrates the key components of the image reproduction process including two colour reproduction media or devices with their corresponding colour gamut. The input colour reproduction media are those capturing digital colour information from a visual stimulus, either a real scene or an object. Scanners and digital cameras are typically considered as input media reporting RGB values for each of the pixels it records. The values represent the amount of light detected at a given stimulus location through the device s RGB filters. Since the RGB values recorded by an input device have a colorimetric meaning that is a function of the device s photodetector properties, the filter RGB s of each individual camera end up having somewhat different colour meanings. On the other hand, output media take a two-dimensional image array of digital values and generate a stimulus as a result. In case of a printed medium as an output the printer deposits device dependent CMYK colorants on a substrate based on digital inputs and produce a subtractive colour reproduction. Technologies for depositing colorants on substrates include dye sublimation, thermal wax, inkjet, laser, colour photographic prints, and halftone colour printers. All these are representatives of using the subtractive colour mixing process. The key property of device dependent colour spaces is that they do not in themselves have colorimetric meaning. For example, a device CMYK of [0%, 10%, 20%, 100%] does not give any information about colorimetry or colour appearance, it is just an instruction to the printer engine how much colorant it has to deposit onto the substrate. Device colour spaces also map onto the colour range of the device that they are used with, which makes them particularly suitable for controlling device outputs. The appearance of a final print is a function of the printing device s characteristics (e.g. tone reproduction 43

55 Image Reproduction and Colour Management Background curve), the primary colour used, the properties of the substrate and the viewing conditions under which the print is viewed. There are other types of output colour reproduction media to consider. These are displays using different technologies such as e.g. cathod-ray tubes (CRT), liquid-crystal display (LCD), organic light emitting diode displays (OLED), plasma displays and digital micro-mirror devices (DMD) respectively. The display media use mainly additive colour reproduction. However, the colour appearance is strongly dependent on the viewing conditions [121]. Device specific colour models means that equal sets of RGB or CMYK numbers will produce different colours (different colorimetric meaning) on different devices (or different substrates on a printer or a printing press). To produce the same colour on different devices the RGB and CMYK numbers need to be changed and adjusted to each device. Hence, RGB and CMYK are just instructions for a device of how much colorant to use (e.g. to produce equal colours on different devices). L* L* -a* -a* -b* +b* -b* +b* Original Image +a* Colour Gamut Mapping +a* Image Reproduction Input Colour Reproduction Medium Output Colour Reproduction Medium Figure 13: Schematic overview of the key components of image reproduction. 44

56 Background Image Reproduction and Colour Management Colour Gamut Mapping A key part in image reproduction is colour gamut mapping, which is to transform colour information between colour spaces. In other words, gamut mapping provides the connection between the original s appearance (source) and the appearances possible in the reproduction (destination) and has to map each colour in the source gamut to a colour in the destination gamut. In order to be able to reproduce the original image properly, modifications in some colours often have to be done which either affects the accuracy or the pleasantness of the reproduction. The substitution of original colours in the image source colour space by colours which are reproducible in the destination colour space can often lead to loss of information caused by shape deformation and size reduction of the colour gamut between an original image and its reproduction. To overcome these challenges an impressive number of different gamut mapping algorithms (GMAs) have been proposed in the literature. Morovic and Luo have made a comprehensive survey in [121, 123, 124]. The authors classified the point-wise GMAs into two categories, gamut compression and gamut clipping, respectively. For further details considering the two reproduction intents we refer to Morovic and Luo [125]. For a long time, most of the research activities on GMAs have been related to simple mappings in colour space. A different concept of performing gamut mapping is to take the spatial dimension into account. Farup et al. [46] present a review of different types of spatial GMAs and their characteristics. Further work in perceptual evaluation assessing the performance and gamut mapping quality of different type of GMAs has been done by Bakke et al. [10], Bando et al. [12], Bonnier et al. [20], Dugay et al. [37], and Morovic [121]. The construction of a gamut boundary descriptor (GBD) is the first step in performing a colour gamut mapping transformation. The performance of a GMA depends on the used GBD estimating the appropriate gamut volume. Bakke et al. [11] introduced a method for GBD evaluation; among a number of GBD s tested the authors conclude that the modified convex hull algorithm performs well on a range of different data sets. 45

57 Image Reproduction and Colour Management Background After reviewing the fundamentals of colour image reproduction let s focus on the principles of ICC colour management for print production Principles of ICC Colour Management The basic aim of colour management (in the graphic arts industry) is to ensure colour accuracy throughout the entire workflow from initial draft through the finished printed product. To achieve this goal it is necessary to map the colour reproduction characteristics of each input and output colour reproduction medium within the device independent colour space (CIEXYZ or CIELAB) and to store this information in ICC colour profiles. Thus, the final colour of the print can be simulated at any stage in the production workflow [68]. Colour management as such is no different from the process of colour reproduction used in the graphic arts industry for many years except moving from a closed-loop colour management system to an open architecture using a device independent colour reproduction system where each source and destination is being handled independently. To be able to adopt and implement this new architecture managing colour in the printing industry the International Color Consortium (ICC) 4 was established in 1993 by eight industry vendors for the purpose of creating, promoting and encouraging the standardization and evolution of an open, vendor-neutral, cross-platform colour management architecture and components [178]. The development of ICC profile specification with the current version called version 4 (v4) profiles [72] was the result of this co-operation. The ICC profile specification is also standardized and published in ISO [90] which specifies a colour profile format and describes the architecture within which it can operate. The key components in the ICC architecture are profile connection space (PCS), ICC profile, rendering intents (RI), and colour matching module (CMM). Essentially the solution proposed by the ICC is based on a process where the colour reproduction is divided into two transformations. The forward colour transformation takes the device dependent colour information (e.g. from a camera or a scanner) and transforms it into a device independent colour space (colorimetric), called profile connection space (PCS). Then, the inverse colour transformation takes the colorimetric information and transforms it into a device colour space such as RGB or CMYK (see Figure 14). The ICC device profile for a given colour reproduction medium describes the relationship between device control signals (RGB or CMYK) and the actual colour those signals predict or produce. It defines the device independent values (CIEXYZ or CIELAB) that correspond 4 International Color Consortium (ICC), 46

58 Background Image Reproduction and Colour Management to a given set of device dependent numbers and vice versa. ICC [72] provides detailed specifications of the structure of a profile including their internal operations and the specification of the architecture. Note that each colour reproduction medium is required to have its particular device profile rather than just one for each imaging device. For example a printing machine will need different profiles for different type of substrate (e.g. glossy paper and recycling paper). All colour information is transformed through the PCS combining ICC source profile and ICC destination profile. (Here the term source-destination is used instead of original-reproduction. However, the meaning of these pairs is the same.) The PCS is based on the colorimetric colour space CIEXYZ or CIELAB determined for a specific observer (CIE Standard 1931 Colorimetric Observer), relative to a specific illuminant (D50), and measured with specified measurement geometry for reflecting media (0 /45 or 45 /0 ). Measurement procedures are also defined for transmitting media. In the current specification, two variations of PCS are defined. One is an original-referred variation for colorimetric intent profiles, and the other is a standard output-referred variation for perceptual intent profiles where the measured data was not made relative to D50. The profile builder application is expected to correct the data to achieve this (derived from ICC [72]). As discussed above, each colour reproduction medium (device) has a fixed range of colours that it can reproduce under certain viewing conditions, called, its colour gamut. Colours that are represented in the source colour space but are not reproducible in the destination colour space are called out-of-gamut colours. Consequently those colours have to be replaced by other colours in the destination colour space. The ICC framework proposes four different types of colour transformation methods called rendering intents (RI) to allow the user to select the one appropriate for the desired purpose. The rendering intents constitute a permanent component of the ICC profile. Media-Relative Colorimetric Intent: Generally, the colorimetric rendering intents permit within gamut colours to be reproduced accurately. It rescales the in-gamut colours in such a way that the medium's white point is mapped to the PCS white point (for either input or output). It is useful for reproduction in particular for professional photography where the print is similar to the original viewed on a display. 47

59 Image Reproduction and Colour Management Background ICC-Absolute Colorimetric Intent: The tristimulus values of the in-gamut colours remain unchanged by using the ICC-Absolute Colorimetric Intent. It is typically useful for artwork reproduction where the transformation is intended to reproduce the original appearance in terms of the illumination and the substrate. Furthermore, a very important application in the graphic arts industry is the simulation of one medium on another (proofing). Perceptual Intent: This is intended for preferred or pleasing reproduction of images, particularly pictorial or photographic-type images. It is especially useful when the source and destination media are substantially different and a pleasing colour output is desired. For example if the destination medium has a smaller colour gamut than the source medium, perceptual rendering may alter in-gamut colours to allow gamut compression. If the destination medium allows for greater chroma than the source medium, then chroma may be increased to produce a more pleasing result. Since perceptual renderings are vendor specific the results may differ between profiles produced by different vendors. Therefore users have to be aware of this by transmitting output profiles along with their images if a consistent workflow is required as in distributed printing. Saturation Intent: The saturation intent is also a vendor specific gamut mapping transformation. In order to preserve the vividness of pure colours it involves compromises, such as trading off preservation of hue. It is useful for rendering images, such as business graphics (pie and bar charts), where we simply want vivid colours and aren't particularly concerned with reproducing the colours exactly. In the graphic arts industry the saturation rendering intent is the least used in day-today practice. The choice of the appropriate rendering intent can have a significant effect on the colour reproduction. Therefore the selection is an important decision for the user to determine the aim of the reproduction. A colour management module (CMM) performs the actual conversion of the colour data from one colour space to another with the aid of colour profiles. A CMM performs the actual conversion of the colour data from one colour space to another with the aid of ICC profiles. An ICC profile contains a) device colour to PCS translation parameters and b) PCS to device colour translation parameters. 48

60 Background Image Reproduction and Colour Management Typically the communication between the PCS and input device profiles is one-way, because the PCS needs to know what colours the input RGB numbers represent (e.g. from a digital camera). On the other hand, monitor display ICC profiles must be two-way, because the display acts as both an input and an output device (Figure 14). If you create or edit a colour image based on its appearance on your monitor display, you are using your monitor as an input device. If you display an image on your screen, the monitor is the output device. Printer profiles are always two-way profiles, too. We use them not only to convert them from the PCS to the output colour space for printing, but also to display files already converted to output space on the monitor display, as discussed in the study of Sole et al. [172] (Paper C) considering soft proofing, or to convert a press CMYK image to some other output device s space for proofing [50, 167]. Display ICC profile and RI Digital Printer ICC profile and RI Scanner ICC profile and RI RGB Camera ICC profile and RI RGB RGB PCS CMYK CMYK Offset press ICC profile and RI Figure 14: Overview of ICC workflow including device profiles via the PCS and rendering intent (RI) choices. For the procedure and creation of ICC device profiles and issues addressing assigning and converting source profiles to destination profiles including suggestions of different types of colour workflows (e.g. RGB or CMYK) we refer to textbooks focusing on practical colour management application from Homann [68], Fraser et al. [50], Sharma [167], and Johansson et al. [98] Calibration and Characterization As we have seen above, device profiles describe the relationship between device dependent colour signals (RGB or CMYK) and the corresponding device independent colour output signals (CIEXYZ or CIELAB) under certain given calibration. 49

61 Image Reproduction and Colour Management Background Johnson [99] defines calibration as an operation of establishing fixed, repeatable conditions for a certain device or process to always produce the same colour according to a given set of numbers. The measured values agree with the values specified by a standard or a characterization process and it is an essential first step before any colour imaging process takes place. On the other hand, characterization simply defines the process of relating device-dependent colour (e.g. CMYK values) to device-independent colour values (colours measured on the printed sheet) according to a given printing definition. For example using a colour printing system the characterization process defines the relationship between a digital CMYK colour data set and the resultant CIE colour measurement data obtained. ISO [82] defined an expanded input data set for characterization of 4-colour process printing represented either with the 928 patch IT8.7/3 target or the 1617 patch IT8.7/4 target and the corresponding description of its associated printing definition. Recently, another standard defining Methods of adjustment of the colour reproduction of a printing system to match a set of characterization data has been approved in ISO [97]. In order to retain the validity of a characterization, it is necessary to re-calibrate the image reproduction devices (such as a monitor display, digital printer, or conventional offset press) at regular intervals. From any type of characterization different profiles with different properties can be created Standardization in Offset Printing and Prepress The characterization of traditional printing processes is much more complex and time consuming compared to the generation of ICC profiles for scanners, monitors, and digital colour printer systems. The colour appearance of a printed product is affected by a number of analogue and digital sub-processes, and their respective process parameters applied. The production processes themselves including the raster image processor and digital plate production, the materials used, such as paper substrate type and inks, the printing process used (including all its specific technical characteristics), and operator dependent variables, such as the quantity of ink applied, the ink/water balance, all affect the appearance of the colour in some measure. Obviously the standardization of all these variables is indispensable for the successful use of a colour management system in a professional production workflow. All these factors and parameters are implicitly included in an ICC colour printer profile established for a particular printing process. At the very least, inhouse procedures must ensure a stable production sequence, otherwise the validity of the ICC profiles generated for the printing process will be compromised. 50

62 Background Image Reproduction and Colour Management Essentially, there are two different approaches considering printing press conformance, namely optimized or standardized press behaviour. A fully optimized press aims at maximizing its capability in terms of lowest possible dot gain, highest ink densities and best contrast that the individual printing press can achieve, without considerations of any external specifications or standards. Such individual parameters can create unique press conditions that require custom ICC profiles for creating the appropriate separations. Initially, after the new ICC colour management architecture was introduced in the printing industry and commercialized profiling tools were available it was very common to generate custom printer profiles. Very often, such custom printer profiles were related to a very small number of test prints and colour measurements representing the print conditions. Together with inconsistency of the process calibration in terms of absent repeatability and uniformity specifications, the performance of these profiles and the corresponding print quality has been considered as not satisfactory. Moreover the variations between printing presses and the corresponding conditions were rather large, caused by inconsistency of using common target values of the solid primary colours and common dot gain convention even though the same type of printing substrates had been used [ ] (Paper B and Paper D). Another approach is to make the presses conform to a certain reference or standard such as e.g. ISO , and applying predefined parameters. ISO specifies a number of process parameters and tolerances including the corresponding values to be applied when preparing colour separations for four-colour offset printing [85]. Similar process parameters are specified in ISO for coldset offset lithography on newsprint [86]. By using the second approach and by standardising the behaviour of the press, industry standard ICC profiles can be used. The European Color Initiative (ECI) 5 is providing ICC output profiles for offset and other printing technologies based on reference characterization data according to the ISO series of standards [84-87]. Any single printer output profile contains a specific set of parameters for different separation elements; a single level of grey component replacement (GCR), one certain colour separation, one total-dot-area setting, one method of gamut compression, one tone reproduction curve, and so on. Hence, different output profiles contain different combinations of these separation parameters and thus provide multiple options to adapt data to a particular set of characterization data [58]

63 Image Reproduction and Colour Management Background In terms of aiming for a common print appearance across printing plants (e.g. preserving print appearance of an ad campaign) the second approach including consistency will be the most suitable one to ensure a predictable and equivalent print result. However, a constant maintenance of the calibration including using the appropriate parameters for the solid primary colours, dot gain and grey balance is a very important requirement for predictable print quality. Methods and procedures for print quality assessment and process control according to objective evaluation and perceptive judgement including appropriate viewing conditions according to ISO 3664 [79] are proposed by Nussbaum and Hardeberg [131] (Paper D) and presented in Part II in this thesis. The demand for offset printing standardization in terms of using international print parameters and the exchange of colour-managed data is increasing with corresponding effects on how modern print enterprises work. International standards and Process-Standard Offset (PSO) 6 offer predictable results and consistent, comparable quality location and output device independent. To reduce expensive and time consuming iterations in standardized printing workflow soft proofing has become an important method in predicting the final print product on a monitor display. ISO [83] defines soft proofing as the ability to match colour images displayed on colour monitors to the images produced when the same digital file is rendered by proofing and printing systems. In the past a number of studies and research work have addressed the issue of soft proofing. For more details we refer to the work of Gatt et al. [53, 54], Hardeberg et al. [65], and Roch et al. [156]. However, although the concept of soft proofing is not new, in practical applications the colour appearance between two different media (e.g. softcopy simulation of a hardcopy) can differ a lot due to unsuitable type of devices, viewing conditions, incorrect use of standards and parameters, inaccurate device calibration and characterization, or inappropriate measurement methods. Although ISO defines parameters for monitor and viewing booth condition setup for soft proofing environment, the practical methods to implement these standards and parameters as per the job requirements have not been clearly defined. Consequently, a mismatch between the viewed stimulus on the monitor display and the corresponding stimulus on printed substrate in the viewing booth can occur. The paper by Sole et al. [172] (Paper C) is aiming to describe in detail how to set up an appropriate soft 52 6 PSO assures predictable quality print production from data creation to the final print production. Industrially orientated and standardized method for the creation of print products. and

64 Background Image Reproduction and Colour Management proofing work station (comprising of a monitor and a viewing booth with the appropriate ambient lighting conditions) according to the parameters given by ISO for soft copy and hard copy proof comparison in the graphic arts industry, in order to evaluate the performance of the entire soft proof setup according to the values and tolerances defined in ISO This is also in accordance with soft proofing in a standardized printing workflow according to PSO. In print standardization and Process-Standard Offset (PSO) certification requirements, viewing conditions in the prepress and pressroom have to be measured and verified according to the parameters defined in the ISO standards [79, 83]. Consequently, appropriate measurement instruments are required, and often spectrophotometers and colorimeters are used to determine the photometric quantities (illuminance and luminance values) in the production workflow. Most of the commercial measurement instrument manufactures supply the instruments with an ambient light measurement adaptor to be placed in front of the measure aperture Process-Standard Offset (PSO) PSO is an industrially orientated method or standardized procedure to control the various steps in the creation of print products. The idea of PSO was developed by the The German Printing and Media Industries Federation 7, in collaboration with Fogra 8 and UGRA 9 and is in conformance with the international standardization series ISO The implementation of PSO assures predictable quality print production from data creation to the final print production following the functions: Organization and documentation ISO 9000 [81] Data reception and data creation ISO x [91-94] Display/Soft proofing ISO Proofing ISO Standard illumination ISO 3664 Plate making ISO 9000 Print production ISO In PSO, testing devices and control methods are described with which the production process can be controlled. It describes the input to output workflow, which includes a number of ISO standards describing this workflow Fogra Graphic Technology Research Association, 9 Association for the Promotion of Research in the Graphic Arts Industry, 53

65 Image Reproduction and Colour Management Background 54

66 Background Print Quality and Print Assessment 2.4 Print Quality and Print Assessment The previous section discussed the fundamentals of image reproduction and colour management including printing standards and methods. This section is focusing on print quality, discussing various factors affecting the print quality and the use of different assessment methods and procedures to determine the print quality. Beside the introduction given in the thesis a considerable amount of literature is devoted to image quality and assessment methods. In particular, the reader is referred to the textbooks by Gescheider [55] and Fairchild [42] for the fundamentals of psychophysics, Keelan [103] for characterization and production of image quality and Engeldrum [38] who proposes an image quality toolkit for imaging systems. Hunt [71] explains the visual appreciation and the basis of judgement regarding colour reproduction. Kipphan [105] discusses factors for print quality assessment and specifications for process control. Different aspects of colour printing quality are proposed by Field [48]. The textbook by Wang and Bovik [189] provide details about objective image quality assessment including a broad treatment of the current state-of-the-art and future directions. Since the human observer is the ultimate receiver in most image reproduction applications, the most reliable way of assessing the quality of an image, a print or a document is by subjective evaluation. On the other hand, depending on the type of application, subjective assessment methods are not preferred or appropriate due to significant individual differences in the observer judgements. Thus, objective methods should be applied, attempting to simulate the functional components in the human visual system. How can we classify image quality, print quality, and document quality? Before we discuss the subject of this section let s briefly define the three types of measures in the context of the present work. Even though there is no distinct borderline between image, print, and document quality assessment we can address them as different measures. Often, image quality is related to digital images which are subject to a range of distortions, caused by acquisition, processing, compression, and reproduction, resulting in a degradation of visual quality [190]. According to Silverstein and Farrell [171] there is a natural tendency to confuse image quality with image fidelity and the two terms are often used interchangeably. The authors point out a clear distinction between the two measures and define image fidelity as the ability to discriminate between two images and image quality as the measure of preference for one image over another. For example an original image 55

67 Print Quality and Print Assessment Background reproduced and printed in different location: if we cannot detect the differences between the prints, we might conclude a high print fidelity. On the other hand, if an original image is enhanced by a distortion, people may detect the difference between the original and the distorted version and prefer the distorted version over the original. We may conclude that it is either the image and its content that is determined by the quality aspect or it is the behaviour of the reproduction medium, which is in focus. For example by assessing different types of GMAs the image quality obtained by the different transformations is in focus, independently of the behaviour of the reproduction medium used to view or print the image reproduction. In contrast, assessing the performance of the reproduction medium itself, the image quality is not necessarily of interest but the behaviour and the quality of the device to capture, view or print a stimulus is of importance. In other words, even though the performance of a device conforms to specification, the image quality can be distorted due to some error in the input data. A further term, document quality, has been used in other studies by Falkenstern et al. [44, 45] to indicate that images are only one graphical element among others. Text, vector illustrations, charts, and vignettes are also visual element integrated in a document. If the behaviour and performance of an input or output reproduction medium has to be evaluated then the measure of the device quality has to be considered. As the scope of this thesis is related to printed media, the quality and control of the printing process is relevant What is (Print) Quality? The term print quality can have many aspects depending on the context or the way the issue is considered. The quotes 10 made by print customers regarding their printed copies and the judged print quality illustrate the complexity in terms of quality and expectations: I just received my first book and thought everything was really good except the print quality. The print shop advertised they can print bookstore quality photo books, but I can see the printer roller marks on the page and all my images have lost their detail and look blotchy same problems, no grey, some of the pictures have a colour cast others are too dark. I'm happy with the print quality - the colours turned out vivid and sharp - the pictures look close to what I see on my monitor 10 Quotes collected from Blurb Forums: 56

68 Background Print Quality and Print Assessment I calibrate my monitor once a week and, well, I know what I'm doing. I just received my colour prints and I notice no detail in blacks, lots of print grain in the mid-tones, blotchy skin tones, banding on every page... The term "quality" is often left unclear because definitions are vague or incomplete. Even though the common definition, "quality is meeting or exceeding customers' expectations," is quite true, it is completely useless if those expectations are undefined. Even if print specifications are identified in advance, and the definition "quality is conformance to specifications" is proposed, this statement can be extremely narrow and does not necessarily reflect what customers really mean by the term "quality." Field [48] adapted a generic framework of product quality, proposed by Garvin [52] and applied it for the specific needs of the traditional colour printing industry. He is considering nine aspects of colour printing quality, divided into four appearance factors and five valued by customers. Appearance factors: Conformance to specifications: This factor refers to how closely the final product conforms to pre-defined tolerances in terms of density or colour targets and registration goals. Technical excellence: Concerns the physical and psychophysical properties of the colour reproduction. Tone and colour reproduction, image definition, interference patterns and surface characteristics are the main components of this aspect of quality. The printing conditions, the characteristic of the original, and the specific instructions or demands influence the optimal values for these attributes. Aesthetics: It refers to the creative decisions made by the photographer or graphic designer to express their message in the advertising or fashion illustration business. The printers as such do not have significant influence on the aesthetic aspect of quality. Permanence: It is related to the ability of the substrate and inks to resist environmental influence of light, chemicals, and moisture. Overprint varnishes or coatings can also influence the permanence. The non-appearance aspects of colour printing quality are related to customer service, production, logistic, and economics. Recently, a new quality label of the graphic arts 57

69 Print Quality and Print Assessment Background industry has been introduced. The swisspso 11 label stands for a comprehensive quality covering the entire workflow, from initial customer contact through the production until to the disposal of the waste Factors Contributing to Print Quality At the highest level, printing can be seen as having digital data as its input and a visual stimulus as its output. The relationship between the digital data and the corresponding visual stimulus, i.e., a print viewed in a certain way, can be highly dynamic. As various factors are involved in the printing process as well as in the viewing of the resulting print, there is great potential for variations determining the final print quality. Considering a generic printing workflow the starting point is a digital image, which is sent to a printer from some software application using a certain printer driver. For nonimpact printing technology (NIP) the instructions from the printer driver are sent directly to the imaging engine of the printing system, e.g., LED, inkjet, dye-sublimation, which deposits colorants on a substrate. In terms of conventional printing the instructions from the printer driver are sent via a raster imaging processor (RIP) to the computer to plate (CTP) system where the printing plate or image carrier is the material by which ink is transferred to the printing substrate [105]. The final print, when viewed in a certain environment under a certain light source, results in a certain appearance for a viewer. Any change to the properties of any of the elements involved in the workflow has the potential to alter the final appearance corresponding to the digital input. Thus, the workflow can be typically divided into categories and each category can have certain factors affecting the appearance of the print as demonstrated in Figure 15. The different factors can be categorized into two classes. Those that determine the resulting print for a given digital input, and those that determine the colour appearance of a given print. Multiple factors and their alternative states, (e.g., the light source can be A, D50, or office lighting ) affecting the appearance and consequently the quality of the print have been discussed in studies by Nussbaum and Morovic [135] and Morovic and Nussbaum [126]. Although, this reproduction process is rather complex in its nature high fidelity and consistency of the reproduced images are expected in terms of the perceived print quality

70 Background Print Quality and Print Assessment Digital Input to Print Print to Appearance Digital Input Data format Colour space Resolution Compression Printer driver Calibration Colour management RIP/CTP Printing System Printing technology Characterization Primary ink colours Dot gain Dot shape Grey balance Register Substrates Print Uniformity Repeatability Temporal stability Colour measurement Surface Postpress Figure 15: From digital input to print appearance. Viewing Stimulus and viewing conditions (e.g. size, distance, geometry, flare background, surround) Illumination Light source White point Intensity Essentially, print quality can be seen as a combination of a number of individual attributes or factors contributing to the overall perceived print quality. A work conducted by Rasmussen [150] identifies the most significant attributes affecting the printed image quality produced by a colour printer. He proposes different approaches to measure print quality dependent on a specific goal (e.g. manufacturing, development), and to evaluate which measurement techniques meet those fundamental requirements. Marcu [115, 116] then deals with various factors that determine print quality such as printing technology, colorant/media interaction, geometric resolution, halftoning, separation, black generation, under colour removal (UCR), grey component replacement (GCR) and tone reproduction. Kipphan [105] refers to factors of influence and specifications determining the quality of print and is dividing the production process into the components prepress, print, postpress, and material. Each component has multiple factors influencing the print quality, similar to what we have seen in Figure 15. On the other hand the specification to determine the print quality must be definable and measurable. Kipphan defines four categories named colour, resolution, register, and surface and attributes attached to them. For example colour coordinates, optical density, dot gain, dot shape, evenness of ink distribution (ink layer) are specification for the category colour. Sharpness, addressability, gradation (tone value range) are attributes dedicated to the specification resolution. Dot/colour separation position, and printed image position are attributes attached to the specification register, while gloss, mottling and evenness belong to the specification surface. Specific test elements (targets and patches to measure) are printed 59

71 Print Quality and Print Assessment Background together with the print image and appropriate measurements instruments (see Section 2.2) are available for continous quality control. In addition, print quality can be checked visually. However, some minimum requirements in terms of appropriate illumination and viewing conditions according to ISO 3664 [79] must be fulfilled to perform a visual quality control. Depending on the image quality factor definition there are a number of other attributes which obviously can be considered as dimensions of image quality, such as tone reproduction, dynamic range, contrast, noise, graininess, artifacts etc. Nussbaum et al. [134] (Paper A) proposed a number of quality factors customized for the evaluation of the appropriate digital print system. Engeldrum [38], in his famous Image Quality Circle, is using the Nesses, which reflects the ending of many words describing some the important perceptual attributes. He is also using the expression psychometric scaling as the process in which observers are asked to assign numbers to the perceptual attributes according to their visual judgement. Typically, lightness, colourfulness, pleasantness, and sharpness are examples of human perceptions reflecting these attributes. For further details on perceptual colour attributes and their definition we refer to Hunt [70]. Very early two types of image/print quality assessment methods were recognised. The first method is based on observation, using psychophysical experiments to gather the judgement of human observers. The second method is by measurement, using instruments to determine values for the various quality factors. These are mathematical models based on perception Print Quality Evaluation by Visual Observation The evaluation and judgement of print quality performed by human observers involves psychophysical experiments. The discipline concerned with the measurement of perception or sensation is psychophysics, which according to Fairchild [42], is the scientific study of the relationships between the physical measurements of stimuli and the sensations and perceptions that those stimuli evoke. And in the context of perceived image or print quality methods have been developed to measure the human perception or sensation. Historically, psychophysics has its origins in the early 19th century, when Ernst Heinrich Weber was asking observers to lift a certain weight and then added increments to the weight until the observers just realised the new weight different from the original. Weber found that the ratio of ΔI/I is constant, whereby I represent the magnitude of the stimulus (weight in this case) and ΔI is the change required in the stimulus (increment of 60

72 Background Print Quality and Print Assessment weights) to be just noticeable (the least difference that the observer still can perceive as a difference). Another important work (built on Weber s law) is the transformation of the physical stimulus intensity scale into a perceptual magnitude proposed by Gustav Fechner. Fechner s law, which relates the physical magnitude to the perceptual magnitude by a logarithmic function is built on two assumptions, firstly, that Weber s law is valid and that secondly a just-noticeable difference (JND) is a unit of the perception scale. This was published in Elements of Psychophysics in 1860 [47]. Stevens studied the relationship between physical stimulus intensities and the corresponding perceptual response for a large number of different types of perception. He conducted experiments using magnitude estimation and magnitude production ratio scaling to study brightness and loudness [55]. One of his particular improvements compared to Fechner is the direct estimation of the perceived intensity using psychophysical ratio scaling. It results in Steven s power law S = οi m, where S indicates the subjective, perceived intensity, I the physical stimulus intensity, ο is a constant (which depends on the measurement units used) and m is the exponent which changes depending on the nature of the physical stimuli judged (typically for brightness <1, more precisely 0.33) Scaling types In order to evaluate an image reproduction in terms of the subjective quality, a number of methods can be applied to gather the observer s judgements and to quantify the perceived image quality considering the reproduction properties. Before the most common methods are discussed four scaling types proposed by Stevens [175] have to be addressed. He called them: nominal scale, ordinal scale, interval scale, and ratio scale. Nominal scale is the simplest scaling type (also denoted as categorical), which uses names or labels for certain characteristics to distinguish among them. In an ordinal scale the assessed samples are assigned with numbers to represent the rank order either in an ascending or descending order dependent on the judged attribute. However, the distance between samples can be large or small and can change up or down the scale. For instance a set of images with different intensity of sharpening applied could by ranked from the most sharpen image to the least. Although, we can obtain a ranking order from most sharp to least sharp or vice versa the scale does not provide information about the magnitude of sharpness differences between the scale values, only information about the greater than (>) or less than (<) property. 61

73 Print Quality and Print Assessment Background An interval scale adds the magnitude of distance between the images to the ordinal scale. The differences between the scale values on the interval scales represent equal distance anywhere along the scale. For example, if a pair of images (e.g. with different sharpness) are judged and separated by two units and a second pair of images at some other point on the scale represent the same two units of magnitude in difference, the perceptual difference of the two pairs will be equal [39]. The ratio scale facilitates all the properties of the first three including a meaningful zero-point. However, according to Engeldrum, defining a meaningful zero-point in visual work determining image quality can be rather difficult [39]. For example, classifying a grey image equal to zero colourfulness sounds straightforward. But, what about a hue scale with a hue having a starting point of zero? Besides the mathematical operations addition and subtraction, which can be performed with interval scaling, multiplication and division are further operations, which can be used on a ratio scale. Most of the applied psychophysical experiments are intended to perform on an interval scale or a ratio scale, either directly or, by observer estimation or by data post processing analysis according to an appropriate probabilistic model [103]. In general, it can be concluded that the greater the power of the scaling type the more difficult the task can be for the observer performing the appropriate judgement. In other words, performing a nominal scaling by labelling or naming items or generating an ordinal scale ranking of a set of images according to a most sharp least sharp sequence can be considered as rather easy tasks for observers. A much more demanding task for observers (at least without training) is to respond with a certain number for the ratio to judge a given attribute Psychophysical methods Essentially, four types of psychophysical methods, presenting the stimuli to the observers and the technique of collecting the observer s judgement have been proposed: Threshold, matching, measuring differences and direct ratio scaling, see Figure 16. The first two methods (threshold and matching) are dealing with discrimination each in its own way, which means that threshold discrimination is aiming to find a stimulus that can be distinguished as different from a certain standard stimulus. On the other hand, matching is almost the opposite process: finding a stimulus that can t be discriminated from a standard stimulus. However, both methods can use the same discrimination technique. 62

74 Background Print Quality and Print Assessment For generating measuring difference for stimuli with supra-threshold perception, four methods are frequently used: rank order, pair comparison, categorical judgement, and rating. Their final results are given in an interval scale. For obtaining a ratio scale, a direct ratio scaling method is used including magnitude estimation, magnitude production, ratio estimation and ratio production. Psychophysical methods Measuring Threshold Matching differences Direct ratio scaling Method of limits Method of single stimulus Method of adjustment Memory matching Asymmetric matching Rank order Pair comparison Categorical judgement Graphical rating Figure 16: Overview psychophysical methods. Magnitude estimation Magnitude production Threshold Threshold experiments are designed to determine the JND. It is used to measure an observer s sensitivity to a given change of a stimulus. Engeldrum [38] describes thresholds in two ways. Firstly, he gives a classical definition where the threshold is the amount of a physical stimulus needed to evoke a JND. Moreover, thresholds are expressed as a physical specification of the stimulus (e.g. a brightness threshold can be measured in luminance units of cd/m 2 ). His second view is based on his Image Quality Circle and the customer s perception on different nesses, which are typically lightness, colourfulness and sharpness. In general, threshold techniques are very useful for determining visual tolerances such as those for perceived colour differences. The basic idea of JND or threshold is related to the concepts of discriminal dispersion or the probability function. Observers are typically asked Do you see a difference? or Is stimulus A different from stimulus B? Although the observers responses might be simple Yes or No, large variations can occur by cumulating the observers answers and the result is described by a probability distribution, like the discriminal dispersion of Thurstone [181]. Three types of experiments measuring thresholds have been evolved to determine absolute threshold and JND, which 63

75 Print Quality and Print Assessment Background are methods of limits, methods of single stimulus, and methods of adjustments [55]. Principally, they vary in the method presenting the stimuli, in recording the observer s response, and in the way data analysis is performed. For further details we refer to Gescheider [55]. Matching As mentioned previously, the matching method is rather similar to the threshold technique, except that the aim is to determine when two stimuli are not perceptibly different. Matching experiments provided the basis to derive the colour matching function of the CIE standard. A colour match across some certain changes in viewing conditions are called asymmetric matching. For example, a test stimulus viewed in daylight illumination has to be matched to another stimulus viewed under incandescent illumination. The resulting pair of corresponding colours can be used to determine and test colour appearance models generated to compensate for different viewing conditions. For some experiments, observes are asked to view two stimuli using each eye separately. Each stimulus is viewed under different viewing conditions simultaneously and the observer s task is to produce a colour match. This type of matching technique is called haploscopic matching methods [42]. Another type of matching experiment is where observers are asked to produce a match or to give a response to a previously memorized colour. For more details on memory match under different viewing conditions see the work of Panak et al. [141]. Measuring differences In the field of image reproduction the measuring differences method is very often used to assess image quality and is intended to obtain a relationship between perceptual magnitudes of a reference stimulus and a test stimulus. In such experiments observers are asked to make their judgement on a single perceptible attribute or ness in terms of greater-than or smaller-than properties of the viewed samples. Note, that for some applications this kind of properties (ordinal scaling) might be sufficient. However, as mentioned previously, to determine the magnitude between the samples (images) an interval scale is a minimum requirement. Given an original and a reproduction (including a set of different reproductions according to a certain ness ), there are a number of different methods of presenting the samples to the observers to obtain their judgements about the various reproduction properties. The two most important psychophysical methods used in the field of image reproduction are the rank order technique and pair comparison 64

76 Background Print Quality and Print Assessment technique. A third important method, which has to be considered for the evaluation of image quality, is the category judgement and finally, the graphical rating method. Rank order method Asking observers to arrange or rank image samples in increasing or decreasing order, (e.g. from best to worst) along a particular perceptual attribute, such as pleasantness, is a typical task for this particular method as applied in the study by Nussbaum and Hardeberg [131] (Paper D). A common data collection procedure is to ask the observer to assign the order from 1 (best) to n (worst) reproduction, where n represents the number of reproductions to be assessed. The obtained data set (averaged from J observers) is providing information on an ordinal scale showing which reproduction is performing best, second best, etc. However, the data does not express the magnitude of how much the reproductions differ. In other words, although the reproduction ranked first and second could be very similar or perhaps very different, it only possesses the greater-than property of an interval scale. On the other hand, the rank order method is considered as rather simple for the observer to perform. Furthermore, due to the simplicity of the method, the observer s repeatability performance is rather high, meaning that repeating the experiment at different times results in similar answers. Pair comparison method In pair comparison, observers are asked to choose the stimulus (or reproduction) of a pair that exhibits more of the desired reproduction property. For example, in the work conducted by Nussbaum and Hardeberg [132] (Paper B) the observers were asked the preference question in terms of Choose the image of the pair you prefer in terms of naturalness. Suppose that a reproduction is judged to be the same (due to identical appearance to the observer) in terms of the chosen property, the observer is either to be forced to make a choice of the one or the other anyway (forced choice) or the observer is allowed to point out a match. The proportion of times a reproduction is judged greater in some particular attribute than another reproduction is calculated and recorded. Although such pair-wise judgement results can immediately express a relative magnitude of differences (due to the ordinal scale), they can t give information about their absolute variation. Hence, an interval scale from the pair comparison data has to be generated. A further constraint of the pair comparison method (even though it provides the most accurate results) is the large number of judgements required for all pair combinations. For n stimuli the number of pair wise combination is n(n-1)/2. 65

77 Print Quality and Print Assessment Background Thurstone s Law of Comparative Judgement As already discussed previously, when observers judge samples either with the rank order method or the pair comparison technique they do not produce an interval scale to quantify the difference between the stimuli. With the help of Thurstone s law of comparative judgement [181] the collected data can be transformed into interval scale data where z- scores represent the distance of a certain stimulus to the mean, and compared to the other stimuli being assessed. Thurstone formalized a model for the judgement process, including various cases of the model based on different assumptions made. His model is based on the notion that a discriminal process (as he called the observer s process of judging samples) performed by observers produce variations in the response of judging a stimulus (due to fluctuations of the organism) which was called the discriminal dispersion (nowadays it is named standard deviation of responses). He postulates that the observer s responses have a variable effect on a hypothetical perception scale (psychological continuum), which follows a Gaussian or a normal distribution. Given two stimuli A and B being judged, the difference between them is determined by the distance between the mean of their response distribution, which in general can be calculated using the following formula: Ψ B Ψ A = Z AB 2 σ ΨA 2 +σ ΨB 2r ΨAΨB σ ΨA σ ΨB (15) The symbol Ψ is used to indicate the discriminal process generated by a particular stimulus, where Ψ A and Ψ B represent the mean response distribution of the stimuli A and B respectively, r ΨAΨB is the correlation coefficient between the two distributions, indicating the extent to which a sample stimulus from the distribution Ψ B and a sample stimulus from the distribution Ψ A tend to be interdependent. σ represents the standard deviation for the stimulus A and the stimulus B and z is the normal deviate or z-score corresponding to the proportion of times stimulus A is judged greater than stimulus B according to some perceptual attributes. Assuming having a stimulus A, which has been chosen over stimulus B to a certain percentage, then the corresponding z-score is the distance from the zero mean (on a scale where the unit represents the standard deviation of the distribution). The distance is corresponding to the area under the normal distribution curve too, which again is equal to the percentage of times that stimulus B was chosen over stimulus A. In other words, z-scores have the effect of transforming the original normal distribution into a 66

78 Background Print Quality and Print Assessment metric which indicates how many standard deviations an observation is above or below the mean. Figure 17 illustrates a normal distribution with zero mean and unit standard deviation (i.e. the Gaussian or standard normal distribution). It can be seen that for an area of 0.84 (i.e. 84 %) under the normal distribution, which represents the percentage of times that stimulus B was chosen over stimulus A, the corresponding z-score is Consequently, this is also the distance between the two stimuli on the interval scale obtained using the law of comparative judgement. Normal distribution Percentage=0.84 { Probability Standard deviation ( z-score=1.00 Figure 17:Normal distribution with given area percentage and corresponding z-score (after Morovic [121]). In practice, the values of the standard deviation σ and the correlation coefficient r are very often unknown and can t be measured experimentally. Therefore, Thurstone outlined six cases (I-VI) applying different assumptions. Case V is most frequently used making the assumption that σ ΨA = σ ΨB (i.e. discriminal dispersions or standard deviations are equal), and that r ΨAΨB = 0 (i.e. there is no correlation between the responses of different stimuli). Given these assumption, case V can be simplified and computed as: Ψ B Ψ A = Z AB 2 (16) For more details on cases I-IV and case VI see the work of Engeldrum [38], and Gescheider [55]. The precision of the experimental results can be described in terms of the 95% confidence interval (CI), which indicates the critical distances for significant differences. It can be computed as: 67

79 Print Quality and Print Assessment Background R ±1.96 σ N (17) where R is the mean, σ reflects an estimate of the standard deviation and N represents the number of observations. Because R has a scale of σ 2, σ =1/ 2, thus the CI can be calculated by: ±1.96* ( 1/ 2)/ N (18) Category judgement method If the number of stimuli is large and the observer is asked to separate the stimuli into various categories, the category judgement method is appropriate. This method involves the definition of equally spaced categories in which the observer is asked to assign the viewed sample to one of the defined categories, e.g. 1=No colour difference, 2=JND, 3=Noticeable colour difference,, 7=Largest colour difference. This method requires more judgement on the part of observers than pair comparison and is more suited to the evaluation of a large number of stimuli which can be impracticable by the pair comparison method. The data analysis of the observer s judgements is dependent on the assumption whether the observations were made on a equally spaced scale (where the perceived colour difference between e.g. categories 2 and 3 is equal to that between categories 6 and 7, etc.), then the mean-category-value method can be applied or that the scale used by the observers is not evenly spaced [13]. If that is the case a transformation of the observers response has to be applied to obtain an even spacing by using Torgerson s Law of categorical judgement [183]. As discussed above, Thurstone formulated his Law of Comparative judgement according to the scale difference between the samples to be judged. However, Torgerson s model, which is an extension to Thorstone s model, calculates the difference between the sample scale value and the category boundary. B k Ψ A = Z Ak σ 2 A +σ 2 k 2r ΨABk σ ΨA σ k (19) where B k is the mean location of the K th boundary, Ψ A is the mean response to stimulus A, σ A is the discriminal dispersion (standard deviation) of stimulus A, σ k is the discriminal dispersion of the K th boundary, r ΨABk is the coefficient of the correlation between momentary positions of positions of stimulus A and category boundary k on the scale. 68

80 Background Print Quality and Print Assessment Finally, Z Ak is the normal deviate corresponding to the proportion of times A is placed below category boundary k. By comparing Equation 15 and Equation 19 we observe the same form. However, the difference between the two models is simply that Torgerson s law of categorical judgement relates to the relative positions of stimuli with respect to category boundaries rather than with respect to another stimuli. Graphical rating method This method provides a direct interval scaling estimation. Here, the observers are asked to indicate the magnitude of their perceptions on a one-dimensional scale that has defined end points. For example, the observers are asked to judge the amount of sharpness in a set of reproduction samples. The observer is given a card with a line of about 20 cm in length. On both ends of the card are adjectives describing the property of the particular attribute such as Extreme sharp on the left side and Extreme un-sharp on the right side. Now, the observer is asked to indicate (by marking on the line) the physical distance on the line as the distance between the viewed reproduction and the corresponding perceived attribute scale. Consequently, the observer s judgement provides an interval scale, which indicates the mean location on the graphical scale for each stimulus. (Similar method can be applied on the computer monitor recording the movement of the cursor distance.) Direct Ratio Scaling Method Previously, we have addressed methods of ordinal and interval scale generation using different methods of data collections and analysis techniques. The following methods will provide results in ratio scale applying the Steven s law [176] which states the relationship between a physical stimulus magnitude and its perceived intensity. Basically, there are two types of ratio scale, magnitude estimation, and magnitude production, respectively. Magnitude estimation In the magnitude estimation the observer is asked to give a numerical response in proportion to the perceived strengths of the particular attribute to be judged. Suppose an observer is assigning the number 10 to a darkness reference sample. Consequently, the observer is asked to give a number that is twice the reference for a stimulus that is twice as dark as the reference sample. The key point in the magnitude estimation is that the numerical response of the observer should represent the ratio of the strength of the attribute of the sample compared to the strength of the reference. This method has been applied in 69

81 Print Quality and Print Assessment Background the study by Nussbaum et al. [134] (Paper A) using an expert panel to conduct the psychophysical experiment. Magnitude production The inverse of the magnitude estimation is the magnitude production. Essentially, the observer is given a number and is asked to generate a stimulus that matches the perceived magnitude compared to the given number. Due to difficulties of the appropriate adjustments of the selected attributes this technique doesn t have many applications in imaging. However, it is assumed that with better understanding of image quality and its component this technique will play a bigger role in the future [38] Experimental set up As we have seen previously, the aim of the scaling process is to have observers assign numbers to certain chosen perceptual attributes. Independently of which scaling method is used, the appropriate experimental setup is determining the human judgement. Therefore, information regarding the viewing conditions, image selection, image size, type of observer, and observer instructions has to be provided. The number of samples and observers needed is task dependent. For example CIE 156 [30] Guidelines for the Evaluation of Gamut Mapping Algorithms recommends to have at least 15 observers to perform a psychophysical experiment using the methods of pair comparison, category judgement, or ranking Quantitative Print Quality Evaluation Applying print quality evaluation in the context of process control, the observer based assessment method is a subjective process and is considered as time-consuming, inconsistent, resource demanding and even expensive. A quantitative measure is usually developed to take into account the human visual system and thus being correlated with the subjective assessment. As seen previously in Section print quality can be affected by many factors from the digital input to the print or from the print to the appearance. In case of a mismatch between the original and the reproduction some of these factors may have contributed to the distortion and the question can be asked which factors have caused the mismatch and how can the distortion be assessed? Suppose the printing process is conducted according to a certain conformance but the printed image is distorted in colour, then the digital input is not appropriate transformed according to the given output parameters. On the other hand, 70

82 Background Print Quality and Print Assessment the digital input can be prepared for a certain output device but the printing conditions do not conform to specification and do not match the input data. In the printing industry, quality inspections are used to control process conditions, and to ensure that the end products meet certain quality criteria. A common approach to measure print quality is to make instrumental measurements of certain printed properties, such as colour, resolution, register, and surface, and use a quantitative metric to determine the print quality. In this work the property colour with its specifications colour accuracy, optical density, dot gain, evenness of ink distribution (ink layer) in different printing workflows has been determined by using quantitative colour measurement [ , 134, 172, 193, 194] (Papers A-E). To determine the print quality in terms of process control the obtained measurement data were compared to specifications (target values and tolerances) provided by ISO [79, 80, 83-87]. Operating with quantitative evaluation using measurement instruments the measurement uncertainty has to be considered, see Section As discussed in the previous section, colour management and ICC device profiles are widely used in the graphic arts and printing industry to transfer colour information between colour reproduction media. The print quality and the accuracy of a colour from a given input, to the display, and further to the printed image in a colour management system depends among other factors on the quality of the profiles involved. The assessment of ICC profiles is a rather complex issue and involves a number of aspects to consider depending on the users requirement and the colour image reproduction s objectives. Sharma [166, 168] proposed methods to evaluate the aspect of colorimetric accuracy of device profiles considering in-gamut CIELAB values. As mentioned earlier, printer profiles are typical two-way profiles; the invertibility assessment of an ICC printer profile is considering the performance of the transformation from the PCS-to-device lookup table (B2A LUT) and the device-to-pcs lookup table (A2b LUT). The "Round-Trip" test is a useful method proposed by Sharma. The B2A LUT is relevant for printing an image while A2B LUT is used when we preview or soft proof images on a monitor display or performing a print simulation on a hard proof system. Depending on the chosen rendering intent these transforms should be either accurate or pleasing. Green [58] provides a recommendation to assist in the evaluation of the colorimetric and perceptual rendering intent transform in ICC v4 profiles [72]. Another aspect in device profile evaluation is to consider the performance in terms of gamut mapping. Different GMA s are proposed (as seen in the previous section) to transform colour values between gamut s of different sizes. The behaviour and performance can be evaluated by comparing GMA s to each other. 71

83 Print Quality and Print Assessment Background Consequently gamut mapping affects the aspect smoothness, which is perceived when transitioning between colour shades. In visual evaluations of colour reproductions the aspect of smoothness is a very desirable property of colour transforms and often given a high ranking. If the viewer perceives a lack of smoothness in a colour transformation it is considered as an unexpected jump in the difference between adjacent regions [59]. A work by Falkenstern et al. [44, 45] proposes a framework containing a number of aspects (e.g. colorimetric accuracy, colourfulness, smoothness, gamut volume, etc.) and the corresponding metric to quantify the performance of a ICC printer profile. It is known that manufacturers of commercial profiling tools use different techniques to fit measurement data into a model of the device to be characterized, thus each tool has its own strengths and weaknesses. A study by Büring et al. [23] investigates profiling tools and the generation of ICC profiles based on one single set of measurement data. The work focuses on image quality asking observers to judge the reproductions of natural as well as artificial images with various image content with respect to a given original. It is worth mentioning other research areas in the field of quantitative print quality evaluation. Considering print quality control Kuenzli et al. [108] proposed a mini-target for on-line measurements in newspaper printing production using an RGB camera to detect the target values and an image analysis software which is able to assess the registration, solid-tone density, dot gain, and colorimetric values. A print quality toolkit which calculates the print quality metrics specified in the ISO/IEC guidelines on office equipment - measurement of image quality attributes for hardcopy output [96] is presented by Grice and Allebach [61]. The printed test target is scanned at a desired resolution, and the resulting scanned test target is analysed according to four distinct categories of printed areas: line and character metrics, solid-fill (colorant blocks) metrics, tint solid (halftoned blocks) metrics, and background field (non-printed areas) metrics. A similar approach, using a scanner or camera to perform objective print quality measurement is presented by Streckel et al. [180] and Kipman [104]. The authors present a series of specific metrics that can be used for quantifying printed image quality for digital printer and paper manufactures. Recently studies have been presented proposing a framework for the evaluation of colour prints using image quality metrics [144]. Multiple work by Pedersen et al. [ ] proposes quality attributes (such as colour, lightness, contrast, sharpness, and artifacts) for the evaluation of colour prints. In combination with an appropriate image 72

84 Background Print Quality and Print Assessment quality metric these attributes are estimated and validated to be able to predict print quality. 73

85 Print Quality and Print Assessment Background 74

86 Summary of included Papers 3 Summary of included Papers In order to demonstrate the relation between the papers and topics addressed, the presented research area can be divided into categories and keywords or subjects. Three categories can be pointed out. They are called human, workflow and physics, indicating the type of core characterization. Moreover, each category includes different keywords specifying the subjects. The category workflow is focusing on the subjects print quality evaluation, colour measurement, standardization, process control, calibration/character-ization, standards/parameters and colour management. The category human considers the subjects print quality factors and psychophysical experiment. The category physics is focusing on the subjects measurement uncertainty and measurement instrument characterization, respectively. The bold letters A, B,, and G, as shown in Figure 18, represent the seven individual research papers and indicate the main area and subjects of each contribution and how they relate to each other. It is important to point out that the structure of categories and subjects is very dynamic in the sense that most of the subjects have large intersections. The overlaps among the subjects and between the categories indicate their relation. Print quality evaluation APrint quality Af Psychophysical Bment Standardization ProcessD control C Colour ma ECalibration / ment MeasF CStandards / characterization Cparameters E uncef Gent Gzation A B D C E F G Colour measurement Human Workflow Physics Figure 18: Diagram demonstrating the categories and subjects and the contribution of the papers related to them. 75

87 Summary of included Papers 76

88 Summary of included Papers Paper A 3.1 Paper A Print Quality Evaluation for Governmental Purchase Decisions Abstract Potential customers within the digital printer market have various demands considering their requirements with regards to desired print quality. This paper aims to investigate the print quality according to predefined quality factors to determine the appropriate printing equipment. In particular the study describes the methods and results of a research project conducted at the Norwegian Colour Research Laboratory for the Norwegian Government Administration Service. The objective of the project was to develop methods and procedures to perform test prints from digital printers including the evaluation and assessment of the print quality according to predefined quality factors. The results of the study indicate the performance of various digital printers in terms of their obtained print quality. This will be useful for the governmental purchase decisions Motivation After several years of market hesitation, digital presses have now become common, and in today s printing market, where flexibility, variable content, shorter lead times, and on demand publishing, are being requested, digital presses represent an attractive supplement to conventional offset presses [66, 105, 110]. The Norwegian Ministry for Transport and Communications who is responsible for the design handbook defining the colours used in the ministries logo (National Coat-of-Arms) has decided to embrace this new technology and replace the conventional offset process for printing letterheads by digital printing systems. However, such potential customers within the digital printer market have various demands considering their requirements with regards to desired print quality. In order to maintain the colour reproduction specifications, as outlined in the design handbook, the Norwegian Colour Research Laboratory at Gjøvik University College in collaboration with Norwegian Government Administration Service (NGAS) has carried out a project to develop methods to evaluate an appropriate digital printer. Although a number of studies and research have been done in the field of print quality it has repeatedly been concluded to be a very complex issue. The subject has been discussed at various conferences e.g. by Stokes [177], Hardeberg and Skarsbø [66], Yendrikhovskij [199] or Norberg et al. [129]. 77

89 Paper A Summary of included Papers In the context of print quality evaluation this contribution [134] (Paper A) presents a solution of developing methods and quality factors to evaluate digital printers in terms of their potential impact on print quality Methods Considering the methods applied to assess the digital print quality in this study we have proposed a number of quality factors including their weights. Furthermore, the quality factors have been divided into three assessment categories, namely, visual logo assessments, general print quality assessments and finally the copy quality assessments. Principally there are two different methods when assessing print quality. The first method is typically done by quantitative analysis, using measurement instruments to determine values for the various quality factors. Considering the evaluation of relative colorimetric colour reproduction of printing devices Microsoft has proposed a quality assurance system [118]. In this study the quality factors including text quality, colour gamut, repeatability and register are determined by quantitative analysis. The second method is based on observation, using psychophysical experiments to gather the judgement of human observers. In this work the magnitude estimation method [55] was employed to judge the quality factors determining the colour logo reproduction in terms of colour match, visual resolution, surface texture, logo alignment and artefacts. The results are provided in ratio scale applying Steven s law [176] which states the relationship between a physical stimulus magnitude and its perceived intensity. Four expert observers gave individual ratings on a scale from excellent (6) to very poor (0), and the average results were listed by ranking. To complete the final ranking the results of the three categories have been weighted once again according to the priorities of the quality requirements defined by NGAS. The category visual logo assessments has been weighted highest by 50% due to the importance of the colour match of the colour logo reproduction, the category general print quality assessments by 30%, and finally the third category copy quality assessments by 20%. NGAS, as the customer of this project, nationally announced a printer quality contest and printer manufactures were invited to participate in the test. Consequently methods had to be designed, including the development of four different digital test documents and test procedures. Finally, six competitors A, B,, F have applied to conduct the test. 78

90 Summary of included Papers Paper A Result The data analysis identifies competitor A as the candidate that scored best in all the quality criteria in the category visual logo assessments. Especially for the quality factor colour match which NGAS prioritised highest candidate A showed an excellent performance followed by candidate B and C. Although the candidates D, E, and F have shown reasonable results for the quality factors «surface texture», «register» and «artefacts» the performance regarding «colour match» was rather poor. In fact, the expert panel has decided that candidate D, E, and F will be excluded in the further evaluation process because they couldn t match the requirements of this category at all. In the category general print quality assessments candidate B performs best followed by the candidates C and A. Although candidate A has shown the largest range of colours achievable compared to the other two competitors the repeatability performance has shown some rather poor results. Finally, in the third category copy quality assessments all three competitors achieved reasonable results without significant differences among them. Lastly, although the final ranking doesn t identify a significant difference between competitors A and B in terms of weighted mean, competitor B has the best score Conclusions The evaluation method including the selection of the quality factors used in this particular study has been unique in terms of print quality. Although various factors have been chosen and applied in the evaluation process the main target to assess the competitors has been the accuracy of the colour reproduction of the three colours of the national logo. In this work we demonstrated and proposed categories and quality factors, which can be applied and adjusted for other applications according to the customer s requirements to evaluate the appropriate print quality. 79

91 Paper A Summary of included Papers 80

92 Summary of included Papers Paper B 3.2 Paper B Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset Abstract This paper aims to investigate print quality in heat-set web offset by applying colour management. In particular it looks at the colorimetric properties of five heat-set web offset presses in order to evaluate the appropriate colour separation approach, either by applying individual separation profiles or by using an industry standard profile such as ISOwebcoated.icc. The key method relies on obtaining colour measurements to determine the repeatability of each participant in terms of colour differences. Furthermore the variation between the five heat-set web offset printing processes and the variation according to the colorimetric values of the ICC profile ISOwebcoated are important parts of the quantitative evaluation. According to the colour measurements two custom ICC profiles were generated and applied to four test images, which were printed by five heat-set web offset presses. Furthermore the industry standard ICC profile ISOwebcoated was applied too. A psychophysical experiment was carried out to determine naturalness of the reproductions made according to the three profiles applied. Finally the results of the study indicate the performance of the appropriate profile applying to the five heat-set web offset presses to obtain significant best print quality Motivation Although process control for the production of half-tone colour separations, proof and production prints are clearly defined in ISO [85] printing processes often show major variations, which affect the appearance of print. The print variations can be detected within the sheet, within the press or between the presses. Due to the print-on-demand concept, different parts of a total publication edition can be printed in different print locations. However, the appearance of the total print edition must be identical. Similarly the requirements for advertising campaigns, which are published in different magazines but printed on equal substrates, have to have identical appearance. Eventually, to strengthen their position in the heat-set market, five of the largest Norwegian heat-set web offset printing plants started a collaboration to evaluate their common print quality and print control. Hence, the aim of this work [132] (Paper B) is to evaluate five heat-set web offset 81

93 Paper B Summary of included Papers printing presses in terms of their conformance to specified values, in accordance with the requirements of ISO Furthermore, the assessment of each individual printing press and the variation within the five participants are important parts of this study, in order to evaluate the appropriate colour separation approach, either by applying individual separation profiles or by using industry standard profile such as ISOwebcoated.icc [41] Methods To assess print quality a number of test prints produced under certain print conditions and parameters are required. Often quantitative evaluation is used in combination with a psychophysical experiment determining a certain quality attribute such as sharpness or colour. In this work a colour measurement instrument has been used to investigate the print run repeatability of each printing press, the variations between the presses and finally the colorimetric variations in accordance with the target characterization table FOGRA28L.txt [49]. The aim of the psychophysical experiment was to determine naturalness of the image reproductions made according to different web coated prints. Four test images have been prepared and converted according to one industry standard profile ISOwebcoated.icc and two custom profiles. The comparisons between the printed images were carried within each individual web offset press. The law of comparative judgement was applied and the method used was pair comparison judgement [38]. The observer s task was to decide which of the two prints in the viewing cabinet was most preferred in terms of naturalness Results According to the colour measurement evaluation all printing plants show a short-term repeatability (within press) performance within an acceptable tolerance. On the other hand except for one printing plant the long-term repeatability shows variations, which can be considered as rather large. In the course of our work, we have found that some of the printing plants have changes in the calibration set up between the two test prints and therefore the variations have been unveiled. An important factor to consider in terms of print quality and print control is the variation performance between the presses. As a result from the quantitative evaluation of the first test print run the variations between the five printing plants demonstrate some undesired bias, which needs further adjustments. The comparison between the obtained colour measurements and the target values of the FOGRA28L.txt show colour differences, which again can be considered as rather large. 82

94 Summary of included Papers Paper B Especially, for the primary colours solid magenta and solid yellow almost all printing plants show colour differences which exceeds the defined ISO tolerances in both test prints. The results from the psychophysical experiment demonstrate that the industry standard profile ISOwebcoated performs significant best even though the calibration set up aiming for ISO parameters is not fulfilled as seen in the quantitative evaluation. Assuming that the print variations within the printing press between the first and second test print would have been rather small it could be expected that the custom profile (which is based on the first test print conditions) would perform best. However, as seen from the long-term repeatability evaluation the print conditions in the second test print have changed. Generally, industry standard profiles are related to a characterization data set with a much larger number of measurements compared to the custom profile. Hence, the industry standard profiles behave smoother in terms of preserving large print tolerances than the custom profile. Moreover, the GCR (grey component replacement) degree in the ISOwebcoated industry standard profile is much stronger compared to the custom profiles and therefore less sensitive to large areas with neutral colours. In other words, the use of low degree of GCR can have a stronger impact to print variations and finally affect the appearance of the print Conclusions In this paper we have demonstrated a method to evaluate the print quality and print control in heat-set web offset. Although the industry standard profile did perform best according to the results of the psychophysical experiment some deficiencies have been detected. The outcomes of the quantitative evaluation demonstrate clearly that there is still potential to improve the target values of the ISO to obtain a better coherence between the five heat-set web offset printing plants. To be able to obtain the colorimetric target values defined in ISO [85] it is very important to ensure the use of the appropriate ink set according to ISO [77]. This specifies the colour characteristics that have to be met by each ink in a process colour ink set intended for heat-set web offset. Eventually the outcome in this study demonstrates an obvious need for standardising the behaviour of the heat-set web offset presses. 83

95 Paper B Summary of included Papers 84

96 Summary of included Papers Paper C 3.3 Paper C Implementing ISO Standards for soft Proofing in a Standardized Printing Workflow according to PSO Abstract This paper defines one of many ways to set up a soft proofing workstation comprising a monitor display and viewing booth in a printing workflow as per the Function 4 requirements of PSO certification. Soft proofing requirements defined by ISO are explained and are implemented in this paper using a Nec SpectraView LCD2180WG LED display along with Just colourcommunicator 2 viewing booth and X-rite EyeOne Pro spectrophotometer. The display monitor colour gamut is checked for its ability to simulate the ISO standard printer profile (ISOcoated_v2_300_eci.icc) as per the ISO requirements. Methods and procedures to perform ambient light measurements and viewing booth measurements using EyeOne Pro spectrophotometer are explained. Adobe Photoshop CS4 software is used to simulate the printer profile on to the monitor display, while, Nec SpectraView Profiler software is used to calibrate and characterize the display and also to perform ambient light and viewing booth measurements and adjustments Motivation Related to print quality as seen in the previous two studies [132, 134] (Paper A and Paper B) soft proofing has become a very important quality control check when simulating the final print product. Although the concept of soft proofing is not new and the task may sound rather simple, in practical applications the colour appearance between two different media (e.g. soft copy simulation of a hard copy) can differ a lot due to unsuitable type of devices, incorrect use of parameters and inaccurate device calibration and characterization, or inappropriate measurement methods. In the past a number of studies and research have addressed the issue of soft proofing such as the work from Gatt et al. [53, 54], Leckner and Nordqvist [109], and Roch et al. [156]. Katoh [102] discussed the appearance difference between the CRT soft copy image and the hard copy image. Although ISO 3664 [78] defines the parameters and tolerances for the appropriate viewing conditions and ISO [83] describes the parameters and tolerances for the displays set up for colour proofing, the practical methods to implement these standards as per the job requirements have not been clearly defined. 85

97 Paper C Summary of included Papers Therefore the main contribution of this work [172] (Paper C) is aiming to describe in details one of the ways to set up an appropriate soft proofing station (comprising a high end monitor and a viewing booth with the appropriate ambient lighting conditions). The proposed method is applied and verified according to ISO standard for soft copy and hard copy proof comparison in the graphic arts industry. This is also in accordance with soft proofing in a standardized printing workflow according to PSO [184] Methods The underlying experimental method is related to light and colour measurement using commercial measurement instruments. Parameters for the ambient light setting, the surround conditions, the viewing booth, the display calibration, and display characterization are provided in the standards mentioned above. To measure and to verify the set up of the soft proofing station including the ambient light conditions, commercial measurement devices, used in the graphic arts industry were used. The obtained colorimetric data were compared with the parameters provided in the standard. Furthermore, the appropriate position and angle of the measurement instrument, determining the measurements of the defined conditions, are demonstrated. The geometry of the measurement device is very critical in terms of the measured values. Small changes of the measurement instrument s angle or the distance to the light source can affect the measurement results dramatically Results The measurement results of the monitor display analysis include the performance of the white point and black point, grey balance, colour gamut, and the uniformity of the luminance. The comparisons between the measured values and the target values in terms of white point, black point and luminance show acceptable results. Although, the max ΔC* ab is 2.35 units, the grey balance is still within the given tolerance. Flare light could be considered influencing the measurement on dark colours. An important requirement of the soft proof set up is the colour gamut of the display which should be as such, that it totally encloses the colour gamut produced by the inks specified in the appropriate part of ISO for which the display is required to provide a proof [83]. According to the measurement results the colour gamut of the monitor is big enough to simulate the ISO coated printer profile. A further important quality criterion is the uniformity of the luminance across the display. Different colour measurements (white, medium grey and dark grey colour patches) are taken at nine positions on the monitor display. The obtained 86

98 Summary of included Papers Paper C results demonstrate that the luminance is uniform across the display and corresponds to the parameters defined in the ISO standard for soft proofing. To analyse and evaluate the viewing booth and ambient lighting conditions, measurements were performed using the X-rite EyeOne Pro spectrophotometer with the ambient light adaptor. Subsequently, the measured values were compared against the P2 viewing conditions defined in ISO 3664 standards. The results show that the obtained values are clearly within the given tolerances. Finally, implementation of the ISO standard alone does not guarantee that a displayed image (soft copy) will match the colour of the same image produced on the hard copy in the viewing booth. Hence, to obtain an appropriate colour match, colour transformation is required to convert the colour data from the printer colour space to the display colour space by using a colour management system and the appropriate ICC profiles [56, 58, 167] Conclusions In conclusion, in this paper we demonstrated a method to set up a soft proofing station according to ISO standards for soft copy and hard copy proof comparison in the graphic arts industry. The performance of the entire soft proof set up according to ISO has been evaluated. It was observed that, in spite of being within the standard s tolerance level the two images (soft copy image on the display and the corresponding hard copy image in the viewing booth) might not show an exact visual match either. Therefore, to investigate the disparity a psychophysical experiment could evaluate the magnitude of visual differences between the soft proofed image and the hard copy in terms of perceptibility and acceptability threshold. Factors and their magnitude affecting the appearance on the monitor display and the viewing booth need to be considered. New reflective monitor display technology, used in combination with high level ambient light, needs to be investigated in terms of image quality for soft proofing application. As a consequence, the colour measurement method including technology and geometry needs to be verified. 87

99 Paper C Summary of included Papers 88

100 Summary of included Papers Paper D 3.4 Paper D Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print Abstract This paper aims to investigate print quality in newspaper print by considering the appropriate calibration standard and applying colour management. In particular, this paper examines the colorimetric properties of eight Norwegian newspaper printing presses, in order to evaluate the relevant colour separation approach, either by applying custom separation profiles or by using an industry standard profile. The key method relies on obtaining colour measurements to determine the repeatability of each participant. Furthermore, the variation between the eight newspaper printing presses and the variation according to the colorimetric values of the ISO standard are important parts of the quantitative evaluation. Based on the colour measurements two custom ICC profiles were generated and an industry standard profile ISOnewspaper26v4.icc was also used. The first custom profile was generated using averaged colour measurement data set from a test print run, and the second using a data set averaged between measured data and the characterization data set IFRA26.txt provided by IFRA. These three profiles were applied to four test images, which were then printed by the eight newspaper printing presses. A psychophysical experiment was carried out to determine the pleasantness of the reproductions, which were produced using the three profiles. The results show the performance of the appropriate profile, which is applied to the eight newspaper printing presses to obtain significant best print quality. It reveals the importance of adopting international standards and methods instead of using insufficiently defined house standards to preserve equal results among different newspaper printing presses Motivation Although process control parameters for the production of half-tone colour separations, proofs and production prints, are clearly defined in ISO [86], often, newspaper printing processes show significant variations which affect the appearance of print. The Norwegian Newspaper Publishers Association (NAL) has provided three custom newspapers ICC profiles in the period of 2000 to 2004 and recommends application of 89

101 Paper D Summary of included Papers these profiles in the national newspapers printing process. However, these custom profiles have two common characteristics in terms of their parameters. Firstly, the number of colour measurements for generating the profiles is very small, and secondly, the degree of GCR (grey component replacement) is rather low considering newspaper print. The press calibration including the performance of the profiles and the corresponding print quality has been considered as not satisfactory and demonstrates the need for further revision. Therefore, the main contribution of this study [131] (Paper D) is the evaluation of eight newspaper printing plants in terms of their conformance to specified values in accordance with the requirements of ISO Furthermore, the assessment of each individual printing press and the variation within the 8 participants are important parts of this study to evaluate the appropriate colour separation approach, either by applying custom separation profiles or by using industry standard profiles such as ISOnewspaper26v4.icc Methods Considering the print quality evaluation in this study, quantitative analyses based on colour measurements and a psychophysical experiment has been applied. The purpose of the quantitative evaluation is to determine the calibration process in terms of the repeatability and uniformity performance [126] of each participant of the project. Furthermore the variation between the eight newspaper printing processes and the variation according to the colorimetric values of the ISO are important parts of the quantitative analyses. This task requires a certain test document with IT8.7-3 CMYK target including solid bars across the paper width to perform density control. The colour measurement data measured on the IT8.7-3 CMYK Target from the first Test Print were the starting point for generating a number of ICC profiles with different separation settings applied to a number of test images. An expert panel determined the two appropriate colour separations, which were then used to carry out the psychophysical experiment using the test prints of the second test print run. A total of two custom profiles and one industry standard ICC profile were applied to each test image (by using the relative colorimetric rendering intent) and printed in all the eight newspaper printing plants [58, 167]. The two custom profiles characterize the IFRA, ISO Profiles. Available:

102 Summary of included Papers Paper D specific Norwegian printing conditions, whereas the industry standard ICC profile considers the conditions in accordance with the international standard ISO The aim of the psychophysical experiment was to determine how pleasant the reproduction of a newspaper print was considered to be, when compared to the remaining newspaper reproduction prints. The observer s task was to rank the three prints in the viewing cabinet in order, from best to worst, in terms of preferred pleasantness [55, 63, 122] Results A calibration set up of a printing press including appropriate repeatability and uniformity performance and the use of certain target values according to the given ISO standard are determined to obtain homogeneous print results among printing presses. However, as seen from the quantitative evaluation results not only the variations in terms repeatability and uniformity demonstrate some unwanted bias but also the colour difference between the individual measurements of each printing press and the target values show disparities, which partly exceed the defined tolerances. In particular the results of the Tone Value Increase have shown large variations between the primary colours cyan, magenta, yellow and black, between the printing presses and between the two test prints. One of the factors, which have affected the differences, is the missing tone value increase (TVI) specification in the Norwegian newspaper production method. Hence, this can explain the inconsistency in terms of dot gain between the primary colours and between the printing presses. Another possible factor affecting the print results is the inappropriate measurement technology used in the newspaper production process recommended by NADA which proposes to use image-based dot meters not only for measuring the screening dots on the printing plates but also for measuring the dot gain on newspaper substrates [128]. A study by Wroldsen et al. [194] (Paper E) investigated the measurement performance of dot meters on newspaper substrate and concluded a very low repeatability confidence using dot meters in newspaper print, compared to using a colorimetric measurement approach. The results from the psychophysical experiment demonstrate that one of the custom profiles performs significant better than the second custom profile and the industry standard ICC profile. However, it is observed that the measurement parameters of the standard characterization data set do not match the calibration parameters of most of the participants in Test Print 2. Hence, it is not expected that the standard ICC profile will 91

103 Paper D Summary of included Papers perform better then the custom profile. Furthermore, the colorimetric difference between the standard characterization data set and the custom characterization data set can be classified as rather small. On the other hand the print variations within the 8 newspaper printing plants in Test Print 2 are partly larger. Hence, the print inconsistency and not the profile selection may have determined the visual print quality Conclusions Although the quantitative evaluation has demonstrated some obvious shortcomings there is a large potential for improving the target values of the ISO to obtain a better coherence between the newspaper printing presses. Nevertheless, to preserve the daily printing conditions and to match the colorimetric requirements of the adopted standard profile it is highly recommended to perform press control according to a well defined standard e.g. ISO The main contribution of this paper is the demonstration of the importance of adopting international standards and methods instead of using insufficiently defined house standards to preserve equal results among different newspaper printing presses. Furthermore the use of appropriate measurement instruments, combined with print consistency in terms of repeatability is the fundamental requirements to obtain predictable print quality. 92

104 Summary of included Papers Paper E 3.5 Paper E A Comparison of Densitometric and Planimetric Measurement Techniques for Newspaper Printing Abstract Two types of measurement technologies are used for process control in newspaper printing, namely densitometric and planimetric technologies. Densitometric measurements are done with densitometers or spectrophotometers, while planimetric measurements are typically done with CCD image sensor-based instruments called dot meters. Although these two technologies are fundamentally different, they are often used interchangeably in print calibration and process control. In this paper we investigate the statistical relationship between densitometric and planimetric measurements on newspaper print. The aim of our project was to investigate whether it was possible to estimate halftone values measured by a densitometer, from the halftone values measured by different dot meters. The applied model is based on regression analysis using second order polynomials. The results are given as estimates of the polynomial parameters, i.e. the polynomials give the relation between halftone measurements with one of the dot meters and halftone measurements with the densitometer. Our statistical analysis showed that due to the large uncertainty of the estimated parameters, the model does not accurately describe the relationship between the two measurement technologies. This can be explained in part by the poor repeatability performance for dot meters applied to newspaper print. Moreover the measurement results also have shown significant variations within the three dot meters used in this experiment Motivation For process control in newspaper printing, essentially two types of measurement technologies are used, namely densitometric and planimetric measurements. In densitometric measurements, the optical density is measured, and if needed converted to halftone values, typically using the Murray-Davies equation [127, 182]. In planimetric measurements, it is attempted to directly measure the halftone values, that is, the dot area coverage, typically using dot meters containing a CCD imaging sensor [34]. Although these two technologies are fundamentally different, they are often used interchangeably in print calibration and process control, in particular in the Norwegian newspaper industry [1, 93

105 Paper E Summary of included Paper 128]. This motivated us to investigate whether there is a statistically significant relationship between halftone measurements on newspaper print done with densitometers (converted into halftone value with the Murray-Davies-equation) and halftone measurements done with dot meters. Eventually the objective of this study [194] (Paper E) is to find out whether it is possible to convert planimetric halftone measurements into densitometric halftone measurements and vice versa. Since these technologies are used interchangeably, it is important to know how to convert from planimetric measurements into densitometric measurements, to keep the printing process under control and to achieve predictable print quality Methods To obtain measurement data with the measurement devices (one densitometer and three dot meters), a specific test target consisting of 16 patches in different halftone values for each process colour (CMYK), was designed and printed in coldset web-offset lithography using AM-screening, in three different newspaper printing plants. Given the halftone values measured by one of the dot meters, the aim was to predict halftone values of the densitometer based on regression analysis using second order polynomials (Third order polynomials tend to result in over-fitting). The results are given as estimates of the polynomial parameters, which means that the polynomials give the relation between halftone measurements with one of the dot meters and halftone measurements with the densitometer. Furthermore, the measurement data were divided into two sets; training set to establish the model (one model for each process colour) and a test set to evaluate its performance. The residuals between the predicted and measured halftone values with the densitometer (with test set = training-set and test set training set, respectively) were used to judge the performance of the model. Because of significant measurement differences between the dot meters and between the process colours for each densitometer dot meter combination, it was necessary to study both the instrument combinations and the process colours individually. To justify that the densitometer could be used as a reliable representative for all densitometers, we did a preliminary test with two different densitometers to verify whether different instruments give the same result. The variation was less than 0.01 density for all the process colours which is within the tolerance according to DIN [36]. A repeatability analysis was also conducted for the three dot meters measuring on newspaper 94

106 Summary of included Papers Paper E print the 50%-patch for each process colour (CMYK) 10 times. Consequently, the repeatability results show rather large variations for all involved dot meters Results Due to the fact that the correlation coefficients were close to 1, it can be assumed that there must be a correlation between the halftone values measured by the various instruments. Based on three measurement series of one test target, second order polynomials were established estimating the relationship between halftone measurements with the dot meters, and the corresponding halftone measurement with the densitometer. If the test set is part of the training set, it is expected that the differences between the predicted and measured halftone values are rather small, as the results have shown. Furthermore, according to the results the variations are colour and halftone value independent and do not follow a certain trend. However, there is no significant trend for the obtained variations. Although the model performs well it is important to test the model with another test-set. When the test-set is not part of training-set the model does not perform that well and the differences between the predicted and measured halftone values are larger. Although the residuals have increased, the variation still does not follow a certain trend. Similar results are obtained for the two other combinations of instruments. The low repeatability of the dot meters is an unfavourable factor that makes our model unsatisfactory. Various factors affecting the repeatability and determining the performance of the model are discussed in the paper. Print irregularities cause noticeable differences in measured halftone values and reduce the repeatability when the aperture is small (in combination with large halftone dots used in newspapers. Due to high optical dot gain (especially for the middle tones and in newspaper print) it is ambiguous to decide what is substratum and what is part of a screen dot. Large residuals between predicted and measured halftone values for the middle tones could partly be explained by the high optical dot gain and problems due to determination of threshold (what is substratum and what is not) in the image analysis Conclusions As expected the residuals between predicted and measured half tone values with the densitometer increased when the test set was different from the training set. Moreover, the measurement results have shown significant variations within the three dot meters. 95

107 Paper E Summary of included Paper Overall, the general conclusion that can be drawn is that the model does not accurately describe the relation between the two measurement technologies due to large uncertainty of the estimated parameters. This can be explained by the poor repeatability performance for dot meters applied in newspaper print. Eventually the main contribution of this paper is to demonstrate that dot meters are not recommended for halftone measurements on paper substrates in newspaper printing. Dot meters are originally developed for measuring printing plates only. 96

108 Summary of included Papers Paper F 3.6 Paper F Analysis of Colour Measurement Uncertainty in a Colour Managed Printing Workflow Abstract Since the recent revision of ISO and ISO , specifying the requirements for systems that are used to produce hard-copy digital proof prints, the use of colour measurement instruments is even more than before required in the printing industry. Therefore, in a modern colour managed printing workflow, most of the printing houses use more than one colour measurement instrument, typically one instrument in each department (pre-press, press, and post-press). In this paper, a total of nine commercial spectrophotometers are compared in terms of measurement uncertainty, precision and accuracy, repeatability and reproducibility. The BCRA series 2 ceramic gloss tiles are used to confirm the accuracy and repeatability of these measuring instruments according to the manufacturer s standards. We focus especially on inter-instrument and inter-model reproducibility and discuss the effect of instrument calibration and certification. For our experimental setup, four different materials are used, one proof print, one commercial print, and one reference print, along with the BCRA series 2 ceramic gloss tiles. In a colour managed printing workflow the use of more than one instrument can impair and complicate the colour process control due to the colour differences between different measurement devices. The effect of the colorimetric measurement errors due to large inter-instrument and inter-model variability between instruments used in different parts of the workflow is discussed and demonstrated Motivation According to the findings in Paper B, D and E measurement instrument uncertainties can highly affect the print quality in a colour managed printing workflow. Especially, the use of more than one instrument in a printing workflow (pre-press, press) can generate unwanted bias in terms of measurement disparities. One measurement instrument may result in measurements which will qualify the print result to pass, and the other one can result as a false-positive. Various studies and research considering measurement uncertainties and comparative studies of different types of colour measurement instrumentations have been 97

109 Paper F Summary of included Papers presented in the past [18, 19, 137, 149, 157, 196, 197]. In a work by Ohno [138] uncertainty is defined as an estimate of the range of values within which the true value lies. However, our main contribution in this article [136] (Paper F) is the analysis of colour measurement variability of using a number of different bidirectional colour measurement instruments in a colour managed printing workflow. Therefore, the aim is to evaluate the performance of nine commercial spectrophotometers (one bench-top and 8 hand-held) typically used in the graphic arts industry in terms of precision (repeatability and reproducibility) and accuracy. Furthermore, the effect of colour measurement variability in a colour managed printing workflow will be demonstrated. In particular the results of inter-model agreement measuring colours on paper substrate will be reviewed Methods ASTM E2214 [9] defines the specification, methods, and procedures evaluating the performance of colour measuring instruments in terms of repeatability and accuracy. Furthermore the reproducibility [197], which determines the variations between instrument s readings can have a huge impact on the evaluation of print quality in the process control. Hence, it is an important aspect of this study. 14 BCRA series 2 ceramic gloss tiles 13 and their true values are used to confirm the accuracy of the instruments. The repeatability performance of the measurement instruments is compared according to the instrument manufacturer s standards. We focus especially on inter-instrument and inter-model reproducibility and discuss the effect of instrument calibration and certification. In order to analyse the measurement uncertainties of the colour measurement instruments on commercial printed substrates, measurements were conducted on three different paper substrates measuring the UGRA/FOGRA Media Wedge CMYK [162] including 46 colour patches. All instruments used in this paper reported spectral reflectance factor values from 380nm to 730nm with 10nm interval. Subsequently, the data were converted to CIEXYZ tristimulus values according to the CIE observer and the CIE Standard illuminant D50 using the method proposed by ASTM 308, Table 1 [5]. Furthermore, CIELAB (D50 as the reference white) values were calculated according to CIE 15 specifications. 13 BCRA, tiles are produced by CERAM Research, Queens road, Penkhull, Stoke-on-Trend, ST4 7LQ, England 98

110 Summary of included Papers Paper F Colorimetric difference ΔE* ab and ΔE* 94 (as some manufacturers quoted the colour difference in ΔE* 94 ) values were computed between the BCRA reference data and the measurements data obtained using each instrument [25] Results As stated in ASTM E2214 the most important specification in colour-measuring instrument is repeatability. According to our results, except of two instruments, all others did pass the manufacturer s agreement in the short-term repeatability test and all in the test available instruments did conform to the long-term agreement. The obtained results from the inter-instrument test have demonstrated a reasonable performance among instruments within the same family, as agreed by the manufactures. Nonetheless, the inter-model measurement results have shown much larger colour differences, especially using instruments from different manufactures. The results of the print measurements did confirm the inter-instrument and inter-model agreement as observed by measuring the BCRA tiles. Considering the accuracy it can be observed that the chromatic BCRA tiles (e.g. Red, Orange and Yellow) produce larger colour differences than the achromatic BCRA tiles, perhaps due to possible thermochromism [43]. Furthermore, all the instruments demonstrate smaller colour differences for the Black tile than for the White tile. The dark to light grey tiles show a very similar behaviour. The larger differences on the White tile may indicate that the instruments do not agree well on the definition of white, which could be traced to the instruments calibration or certification status. However, there is no obvious consistency observed between certified and non-certified instruments in terms of their performance on the White tile Conclusions Measurement instruments with valid certification and instruments with expired certification have been used in this study to be conforming to the common situation in the printing workflow. I might be expected that instruments with valid certification perform better than instrument with expired certification. Although missing instrument recertification did not show a significant effect on the measurements in terms of reduced accuracy or reproducibility performance, it is highly recommended to maintain the instruments according to the manufactures requirement including appropriate recertification procedures. In conclusion, when applying only calibrated and certified instruments, a further obvious consequence will be the use of only one certain instrument 99

111 Paper F Summary of included Papers family (same model, same design of instrument from the same manufacturer with the same parameters) in a colour managed printing workflow. This will minimize the colour differences. To avoid large measurement errors among different instrument within the same workflow it is required to use common filter specifications. In order to improve the colorimetric performance and inter-instrument and inter-model agreement a method of characterizing measurement instruments using colorimetric regression technique can be considered. 100

112 Summary of included Papers Paper G 3.7 Paper G Regression based Characterization of Colour Measurement Instruments in Printing Applications Abstract In the context of print quality and process control colorimetric parameters and tolerance values are clearly defined. Calibration procedures are well defined for colour measurement instruments in printing workflows. Still, using more than one colour measurement instrument measuring the same colour wedge can produce clearly different results due to random and systematic errors of the instruments. The aim of this paper was to determine an appropriate model to characterize colour measurement instruments for printing applications in order to improve the colorimetric performance and hence the inter-instrument agreement. The method proposed is derived from colour image acquisition device characterization methods, which have been applied by performing polynomial regression with a least square technique. Six commercial colour measurement instruments were used for measuring colour patches of a control colour wedge on three different types of paper substrates. The characterization functions were derived using least square polynomial regression, based on the training set of 14 BCRA tiles colorimetric reference values and the corresponding colorimetric measurements obtained by the measurement instruments. The derived functions were then used to correct the colorimetric values of test sets consisting of 46 measurements of the colour control wedge patches. The corrected measurement results obtained from the applied regression model was then used as the starting point with which the corrected measurements from other instruments were compared. The goal is to find the most appropriate polynomial, which results in the least colour difference. The obtained results demonstrate that the proposed regression method works remarkably well with a range of different colour measurement instruments used on three types of substrates Motivation As seen previously, using measurement instruments in a print quality control context reveals questions about the reliability of the obtained measurements. Especially when a certain target colour is measured with different measurement instruments in different locations and the results are qualifying the print differently. According to the results 101

113 Paper G Summary of included Papers presented in Paper F [136] an obvious consequence will be the use of only one certain instrument family in a colour managed printing workflow to preserve reasonable colour differences. However, firstly this is practically almost impossible to implement and secondly even within the same instrument family the measurements will vary. Eventually, in order to improve the colorimetric performance and inter-instrument and inter-model agreement, a method of characterizing measurement instruments has to be considered. Therefore, the main contribution of this study [133] (Paper G) is to propose a method to reduce the colour difference of colour measurement performed with more than one colour measurement instrument measuring the same colour target. In particular, the main contribution of this study is in correcting the measurements obtained by instruments, using colorimetric regression technique. Finally, the appropriate correction model applied to the measurement data sets will reduce the colour differences between the measurements acquired by a master measurement instrument and the measurements performed by a second measurement instrument used. Eventually, the model will improve the colorimetric performance and inter-instrument and inter-model agreement Methods The method we propose in this work is based on colour image acquisition device characterization, which has been applied by implementing polynomial regression with least square technique. Polynomial device characterization technique with least squares fitting for different application has been adequately applied in a number of other studies by Kang [101], Sharma [169], Hong et al. [69] and Johnson [100]. Compare to other studies where the authors, e.g. Berns and Petersen [17] propose the correction of various systematic errors by applying the spectral measurements data using multiple linear regression based on modelling the results to improve the colorimetric performance, in this work, the aim is to correct the instrument s systematic errors by applying regression technique direct to the measured CIELAB data. Generally, to create a characterization model, a training set, existing of a reference data set and the corresponding measurements performed by a measurement instrument is required. In this work the model (training set) is derived by the 14 BCRA reference data set and the corresponding measurements obtained by the measurement instrument. The idea of using the BCRA tiles for the training set is to make the appropriate adjustments in terms of the accuracy. Note, not only the precision between the instruments has to be improved but also the accuracy in terms of the appropriate colorimetric values. Therefore, 102

114 Summary of included Papers Paper G the BCRA tile reference values, which are traceable, have been used to establish the model. Eventually, the model has been tested with independent data using 46 patches of the UGRA/FOGRA Media Wedge CMYK [162] on three different types of printed substrates namely a hard-copy digital proof print, paper substrate type 1 and paper substrate type 5. Six commercial industrial-oriented spectrophotometer measurement instruments typical utilized for daily production control in prepress and press applications are used. One particular measurement instrument has been defined as a reference ( master instrument ). The reference is not meant to represent the best or ideal instrument but instead to be a state, to which to compare the measurements obtained by other instruments. The corrected measurement results obtained from the regression model was then used as the starting point with which the corrected measurements from the other instruments were compared and eventually to find the most appropriate polynomial, which results in the least colour difference Results Even though the training set in the first attempt was limited to 14 samples only due to the available BCRA tiles, the model performed excellent. The colour differences between the corrected measurements of two measurement instruments could be reduced significantly for all three types of substrates. In particular the instrument combination from different instrument families demonstrate a noteworthy performance. Although there is no particular function, which performs best for instrument combination within the same product family, most of the polynomials reduce the colour difference with almost more than half compare to the uncorrected data. Regardless the very small colour differences of the uncorrected measurement data between inter-instruments (the same model from the same manufacturer using equal specifications), applying the regression model have further minimized the colour differences. By extending the training set from 14 samples to 38 samples including the 24 colours from the ColorChecker the results does not significant improve in terms of reducing the mean colour difference for all the measurement. However, the maximum colour difference between all instrument s combinations could be reduced substantially on all three substrates. Hence, this indicates clearly that the model with 38 sample points is more robust. 103

115 Paper G Summary of included Papers Conclusions The paper concludes that first order polynomials (more precise 3x5 polynomial) produces, in most of the cases, the best results in terms of reducing the colour differences between the instruments on different substrates. Although, there is no significant difference in the performance of the model on the three different types of substrates, the proofing substrate results in the least colour differences. Moreover, with instruments used from different product families the inter-model agreement can be significantly improved, by reducing the colour variations between the measurement instruments with more than half by applying the characterization model. Increasing the size of the training set from 14 to 38 samples is slightly reducing the maximum colour differences, but much more important, the model s behaviour is more robust in terms of different applied polynomials. To justify whether thermochromism [43] affected the model, the BCRA tiles red, orange and yellow could be left out in the training sample. The main contribution of this paper is the demonstration of a method to correct the instrument s systematic errors by applying regression technique direct onto the measurement output values in the CIELAB colour space. As a result, the colorimetric performance and hence the inter-instrument and inter-model agreement has improved. In conclusion the proposed method has shown significant improvement for taken measurements with different instruments on reflective light. Therefore, it can be expected that the method perform well too, on measurements obtained from light emitting devices (e.g. display), including the use of different models of spectrophotometers, spectrocolorimeters and colorimeters. However, this assumption has to be proofed. According to ASTM E2214 [9] the most important specification is the repeatability, including the reproducibility. Taking that into account an alternative approach can be considered establishing the model using samples with similar or equal physical properties for both the training set and the test set. In the current work we have trained on 14 BCRA tiles to make the appropriate adjustments in terms of the accuracy and tested on different paper substrates; which means the physical properties between the training and the test set is different. However, a study by Steder et al. [174] illustrates that the inter-instrument agreement increases by using samples of equal physical properties for both the training and test set. In our work the model could be trained on printed sample colours and tested on another set of printed colours on the identical paper substrate. 104

116 Discussion and Conclusion Discussion of Papers in Context 4 Discussion and Conclusion After each individual paper has been summarized in the previous chapter the discussion and conclusion of the thesis will be presented. In the first section the contribution of the papers will be discussed in the context of demonstrating the relation between them. Consequently a summary of the thesis contribution will be presented, before a conclusion can be drawn und some ideas for future work are proposed. 4.1 Discussion of Papers in Context The goal of this thesis has been to develop methodologies and procedures for print quality assessment and process control according to objective evaluation and perceptive print quality judgement. Because the quantitative evaluation has been an important part of the study, the investigation of instrument measurement performance in terms of accuracy and reproducibility for process control and print quality assessment was a focal point of the study. Hence, a method was proposed to reduce the colour differences of colour measurement performed and simultaneously improve the colorimetric performance and inter-instrument and inter-model agreement. Predictable colour reproduction has always been an important issue and a demand within the graphic arts and printing industry. Although common standard parameters to calibrate and characterize devices in the printing workflow have been defined and provided in recent years or even decades, the implementation and the maintenance of the parameters was not always prioritized by many companies. About the motive can be speculated. However, it can be assumed that either missing competence or low market competition are possible reasons for this absent priorities. In the context of print quality assessment the seven papers associated to different categories have contributed to the final present thesis. As illustrated in Figure 18 the contribution of the papers generally cover the three categories human, workflow and physics which encompasses the main subjects colour measurement, print quality evaluation and colour management in this work. Even though Papers A [134], B [132] and D [131] are mainly classified in the human category, because the subjects print quality factors and psychophysical experiment are involved, the main contribution can be related to the category workflow. Paper C [172] in the category workflow is 105

117 Discussion of Papers in Context Discussion and Conclusion mainly dedicated to addressing the subject colour management in the sense of proposing a method of applying ISO standards for soft proofing in a standardized printing workflow according to PSO. In the course of our research work, we have recognized the importance of colour measurement and the consideration of measurement uncertainties in general. Therefore a very important contribution in the category physics are the subjects measurement uncertainty and measurement instrument characterization represented by the Paper contributions E [194], F [136], and G [133]. Recently offset and newspaper printing have adopted uniform standards including parameters and tolerances and process control systems developed by a number of research and standard groups such as ISO (International Standardization Organization), aiming to stabilize and monitor the production processes [143]. It has been argued that the type of standards used in production (official international or in-house standards) is not as important as the fact that some form of standardization regulates the production process to ensure that the same conditions are applied for each sub-process, e.g. from plate output through to actual printing process. However, in our research we experienced great misunderstanding and confusion in applying standards and parameters in the printing industry. Among collaborating printing partners we have noticed the absence of quality and tolerance agreements. A study by Martin et al. [117] confirmed our assumption that a successful implementation of a colour managed printing workflow is rather difficult and that achieving printed results that satisfy the customer is often a complicated business involving a number of print-proof-adjust-reprint cycles. In the beginning of the current research process there was no obvious distinction between colour management and standardization of printing processes. During the research period the difference became more noticeable in the sense that a standardized process is all about calibration, including the aim of obtaining certain target print parameters, and the preservation of the defined tolerances. Consequently, predefined industry standard profiles corresponding to the applied calibration parameters can be used. On the other hand colour management is dealing more with the optimization of a printing process without considerations of any external specifications or standards, and therefore the need for generating custom ICC profiles occurs. Nowadays, it can be argued that a standardization process in printing is all about aiming at certain target values, performing quantitative evaluation and implementing procedures to ensure that the target values are within certain defined specifications and tolerances. Hence, conducting psychophysical experiments might be redundant in this context as discussed in Paper B [132] and Paper D [131]. 106

118 Discussion and Conclusion Discussion of Papers in Context However, the implementation of standardization procedures is more then just applying and maintaining certain process parameters. For a successful implementation high competence and deep understanding of the entire colour managed printing workflow is needed among the people who are responsible for colour management and colour measurements. Therefore, conducting psychophysical experiments involving the production operators by introducing standardization procedures is psychologically and strategically important and is supporting the awareness of the standardization benefits. Together with the use of quantitative instrumental measurement technique this will strengthen the prediction of the expected print quality. Figure 19 demonstrates that multiple aspects and factors contribute to the final appearance of the print and hence affect the print quality. In this work, we have systematized the various aspects interacting with each other. To determine print quality, the definition of print quality factors and the different kind of methods applied to assess them are as important as the concept of process calibration and standardization including colour measurement for process control. For example in Paper A [134] we proposed a number of quality factors to determine the colour reproduction performance of digital printers. As discussed in Section 2.4 the aspect of print quality can be determined either by the image, its content and its transformation or by the behaviour of the reproduction medium itself and the corresponding specification conformance applied. In the context of standardization, standards and parameters, the evaluation of the process and the reproduction media involved are important to determine weather the device performance conforms to specifications. Hence, applying the appropriate calibration parameters and the equivalent characterization is fundamental for the appropriate colour transformation. Quantitative evaluation is based on using measurement instruments to determine process control and print quality. The term process control in this work refers to the quality assurance and prediction in terms of maintaining the print production within predefined tolerance specifications. Since the recent revision of ISO and ISO specifying the requirements for process control for the production of half-tone colour separations system and proofing processes working directly from digital data the use of colour measurement instruments is more than before required in the printing industry. In addition ISO which specifies the minimum requirements for the properties of displays to be used for soft proofing of colour images together with our proposed method implementing the standard presented in Paper C [172] the need for objective measurement 107

119 Discussion of Papers in Context Discussion and Conclusion instruments is obviously. The increasing use of different measurement instruments in the colour reproduction workflow, and as a result of that, the related uncertainty affecting the reliability of the measurement data obtained has to be considered. As discussed in Paper F [136], and Paper G [133] not only calibrated and certified measurement instruments should be used but also the appropriate instrument type has to be applied for certain measurement tasks. The outcome of the research in Paper E [194] clearly demonstrates that dot meters are developed for measuring printing plates only and are not recommended for halftone measurements on paper substrates in newspaper printing. To minimize the colour differences between different types of instruments the use of only one particular instrument family (same model, same design of instrument from the same manufacturer with the same parameters) is recommended. Furthermore, in order to improve the colorimetric performance and inter-instrument and inter-model agreement we proposed a method of characterizing measurement instruments using a colorimetric regression technique. Print quality factors Standardization Standards and parameters Assessment methods Print quality Process control Colour management Colour measurement Concept of calibration, characterization and colour transformation Figure 19: Aspects and factors affecting and contributing to the print quality. The methodology and procedures obtained during our research and the contribution of the individual papers could be used as an application independent framework for those involved in the process of print quality assessment and Process-Standard Offset (PSO). We believe that the media community will benefit from the results and contribution of the present work in terms of producing predictable print quality by applying the procedures and methods proposed. Thus, we hope that we have - at least partially - reached the goal that we aimed for at the outset of this project. 108

120 Discussion and Conclusion Summary of Contribution 4.2 Summary of Contributions This section summarises the scope and contributions of the thesis. The basic objective of this thesis was to propose methods and procedures to evaluate print quality in terms of process control obtaining predictable colour reproduction. An important part of the work is the investigation of measurement devices in the graphic arts and printing industry in terms of their accuracy and reproducibility for process control and print quality assessment. Thus, the measurement uncertainties were analysed and a method proposed to reduce the colour differences of colour measurement performed with different instruments. The proposed model improves the colorimetric performance and inter-instrument and intermodel agreement. The substantial and innovative contributions of the research are as follows: We developed methods and quality factors to evaluate digital printers in terms of their potential impact on print quality. The proposed framework including categories and quality attributes can be applied and adjusted for other applications according to the customer s requirements to evaluate the appropriate print quality. The development of digital test documents and test procedures to obtain the evaluation results is an important part of the concept. With respect to evaluation of print quality and print control in heat-set web offset we have demonstrated a method applying quantitative assessment and in combination with a psychophysical experiment determining the quality attribute of naturalness. The presented method demonstrates the significance of applying common standard target parameters, so that the heat-set web offset printing process is consistent among and between printing presses to provide predictable print quality. Even though ISO standard defines parameters for monitor and viewing booth condition setup for soft proofing environment a clearly defined method to set up an appropriate soft proofing workstation is not stated. We proposed a method and procedure to set up a soft proofing station according to ISO standards for soft copy and hard copy proof comparison in the graphic arts industry and evaluate the performance of the entire soft proof set up according to ISO

121 Summary of Contribution Discussion and Conclusion We demonstrate the importance of adopting international standards and methods instead of using insufficiently defined house standards to preserve equal results among different newspaper printing presses. Furthermore we proposed the use of appropriate measurement instruments, combined with print consistency in terms of repeatability which are the fundamental requirements to obtain predictable newspaper print quality. We have tried to answer the question: Is there a relation between halftone measurements from densitometers (converted into tone value with the Murray- Davies-equation) and halftone measurements from dot meters on newspaper print?" Our statistical analysis showed that due to large uncertainty of the estimated parameters, the model does not accurately describe the relation between the two measurement technologies. This can be explained by the poor repeatability performance for dot meters applied in newspaper print. None of the three dot meters applied in the project fulfilled the requirement of 2% tolerance deviation (note: these are requirements which are not defined in an official standard). Dot meters are originally developed for measuring printing plates only and are not recommended for halftone measurements on paper substrates in newspaper printing. Process control and print quality is often determined according to measurement data obtained by different colour measurement instruments used in a production workflow. We have demonstrated the effect of colour measurement variability (precision and accuracy) in a colour managed printing workflow and in particular the result of inter-model agreement measuring colours on paper substrates has been reviewed. Moreover, in order to minimize and control the colour variations in a colour managed printing workflow, only calibrated and certified instruments within one certain instrument family (same model, same design of instrument from the same manufacturer with the same parameters) should be used. In order to improve the colorimetric performance and inter-instrument and intermodel agreement a method of characterizing measurement instruments using a colorimetric regression technique directly on the measurement output values in the CIELAB colour space has been demonstrated. The proposed method shows significant improvements for measurements taken with different instruments in reflective light. Furthermore, there is a potential to apply the proposed method in 110

122 Discussion and Conclusion Summary of Contribution emission measurements too, using different models of spectrophotometers, spectrocolorimeters and colorimeters. Although we did not demonstrate ground-breaking new technology or algorithms we made several novel contribution useful for the graphic arts and printing industry including the demonstration of methods and procedures to ensure predictable print quality. Moreover, in our contributions we proposed a method for the media community to control the colour measurement uncertainties by reducing the colour differences of colour measurement performed and simultaneously improve the colorimetric performance and inter-instrument and inter-model agreement. 111

123 Summary of Contribution Discussion and Conclusion 112

124 Discussion and Conclusion Conclusion 4.3 Conclusion The major goal of this work was to develop a methodology and procedures that could be used as an application independent framework for those involved in the process of print quality assessment. On first sight, one might assume that assessing print quality in a colour managed printing workflow is a rather trivial task where the measurement results lead to numerical quality estimation. However, it is more complex than it seems. A number of factors are affecting the appearance of print. The specifications and parameters to which the printing process has to be calibrated are often incomplete applied and the appropriate print assessment methods are often absent. In the present work we have demonstrated the implementation of standardization methods and procedures in a colour managed printing workflow to ensure predictable print quality. We found significant measurement variation among measurement instruments used in the graphic arts and printing industry in terms of accuracy and reproducibility performance. Hence, we proposed a solution to correct the systematic errors of the instruments by applying a regression technique directly on the measurement output values in the CIELAB colour space to improve the colorimetric performance and hence the interinstrument and inter-model agreement. The proposed method has shown significant improvement for measurements obtained with different instruments in reflective light. In conclusion, we strongly believe that the methodology and procedures developed as a result of the project can be used as an application independent framework for those involved in the process of print quality assessment. Further, with the results we have obtained, we believe to contribute in finding solutions for the graphic arts and printing industry to strengthen their position against other media and other global competitors in terms of consistent and predictable print quality and cost efficient print production. 113

125 Conclusion Discussion and Conclusion 114

126 Discussion and Conclusion Perspectives 4.4 Perspectives In the course of our research work, we have identified multiple fields for further research in the area of colour managed printing workflow, print quality and print quality assessment. Our thoughts about potential future research areas that should be looked into and ideas for further improvements of the presented methodology and procedures can be divided into three categories. Measurement technologies: Although the proposed measurement instrument characterization model presented in Paper G [133] showed remarkable results in terms of reducing the measurement variations and improving the colorimetric accuracy between different measurement instruments the method needs some further investigations. For example the performance on different paper substrates (e.g. glossy paper, newspaper) and material (e.g. plastic, textile, aluminium, glass) could be analyzed, the method could be extended and tested on emission measurements using different models of measurement devices which are used in the field. With the increasing demand for PSO certifications which requires ambient light measurements in the prepress and press room, the performance and reliability of the measurement instruments used needs to be investigated and perhaps a correction model should be applied to improve the inter-instrument agreement. An interesting aspect is to compare our model with the method applying the correction in the spectral domain as for example proposed by Berns and Petersen [17]. It can be speculated that a combination of the two methods could further improve the colorimetric performance and the inter-instrument and inter-model agreement. Colour difference and print quality metrics: The ultimate goal of implementing standardization procedures and operating with specification conformance is the print quality control according to numbers. Although the current ISO standards define certain tolerance settings for process control for the production of half-tone colour separations, proofs and production prints, the control system is based on colour patches to be measured independently of the document content (colours in high-frequency and low-frequency images) to be printed. In other words, the production tolerances are print content and application independent. However, visual difference between prints from different printing presses (or proof and print) can be perceived for certain colour shades even though the print production on both presses are within the given tolerances. A colour control system could be considered where the printed content automatically is related to the tolerances according to the perceived colour difference. Suppose printing a catalogue with low- 115

127 Perspectives Discussion and Conclusion frequency images (e.g. with large uniform areas) the requirements for the print variations should be reduced compared to printed images that have a high-frequency characteristic. Colour difference metrics for image quality assessment has been widely used for various applications. However, there is seldom a strong correlation between the objective evaluation and the visual assessment. Furthermore, the interpretation of the complete image quality assessment considering the colour difference calculation depends on the application and the acceptance. The acceptability threshold considering print quality assessment is a vaguely defined concept and one that depends strongly on application and industry. Moreover, a further quality metric, which could be considered more suitable for the heat-set web offset and newspaper printing process, is a tolerance threshold. Environmental management: As discussed in the thesis print standardization has a number of goals such as being consistent and predictable in terms of print quality, increasing the production efficiency and reducing costs. A further important aspect to consider is the reduction of print waste and pollution. Pollution, global warming and climate change have become very serious issues and even though a large number of different research projects worldwide have been carried out aiming for environmental management (e.g. ISO [89], Forest Stewardship Council (FSC) 14 ) there is still a huge need for addressing further areas contributing to the reduction of pollution and preserving the environment and climate. The graphic arts industry has taken these issues seriously, in particular concerning the development of new print technologies and processes, and the increased use of recycled paper. Nevertheless, the industry s awareness of the importance of pollution reduction (in terms of saving energy, reducing volatile organic compounds (VOC), and reducing waste could indeed be increased. Furthermore, introducing colour management and standardization of the print production processes, a further positive impact on environment and climate could be obtained. Further research in colour management, standardization and PSO implementation help to reduce the amount of ink and paper substrate used in print reproduction by controlling the total ink coverage in printing on paper substrate. This should in turn help simplify paper recycling due to reduced ink layers on paper. Research in colour management will also help in eliminating hard copy proofing systems in the future by introduction of softcopy proofing using high end energy efficient colour displays. The overall effect of standardization will be to increase efficiency in terms of, increasing 14 Forest Stewardship Council is an independent, non-governmental, not-for-profit organization established to promote the responsible management of the world s forests, 116

128 Discussion and Conclusion Perspectives competence among the graphic art engineers to reduce waste of paper, ink, and other consumables. 117

129 Perspectives Discussion and Conclusion 118

130 References References [1] E. Aasen, Ø. Danielsen, and A. J. Bovolden, "Halftone measurements in newspaper print," BSc, Gjøvik University College, Norway, [2] R. M. Adams II, A. Sharma, and J. J. Suffoletto, Color Management Handbook: A Practical Guide. Pittsburgh, PA, US: PIA/GATF Press [3] M. Andersson and O. Norberg, "Color measurements on prints containing fluorescent whitening agents," in Proc. SPIE: Color Imaging XII: Processing, Hardcopy, and Applications, 2007, Vol [4] ASTM E284-08, Standard Terminology of Appearance, American Society for Testing and Materials West Conshohocken, PA, USA, [5] ASTM E308-08, Standard Practice for Computing the Colors of Objects by Using the CIE-System, American Society for Testing and Materials, West Conshohocken, PA, [6] ASTM E , Standard Test Method for Obtaining Colorimetric Data From a Visual Display Unit by Spectroradiometry, American Society for Testing and Materials, West Conshohocken, PA, [7] ASTM E , Standard Test Method for Color and Color-Difference Measurement by Tristimulus Colorimetry, American Society for the Testing of Materials, West Conshohocken, PA, [8] ASTM E , Standard Practice for Obtaining Colorimetric Data from a Visual Display Unit Using Tristimulus Colorimeters, American Society for the Testing of Materials, West Conshohocken, PA, [9] ASTM E , Standard Practice for Specifying and Verifying the Performance of Color Measuring Instruments, American Society for the Testing of Materials, West Conshohocken, PA, [10] A. Bakke, I. Farup, and J. Hardeberg, "Predicting the Performance of a Spatial Gamut Mapping Algorithm," in Color Imaging XIV: Displaying, Hardcopy, Processing, and Applications, San Jose, CA, USA, 2009, Society of Photo-Optical Instrumentation Engineers, SPIE Proceedings,

131 References [11] A. M. Bakke, I. Farup, and J. Y. Hardeberg, "Evaluation of algorithms for the determination of color gamut boundaries," Journal of Imaging Science and Technology, vol. 54, pp (11), [12] E. Bando, J. Y. Hardeberg, and D. Connah, "Can gamut mapping quality be predicted by colour image difference formulae?," in Human Vision and Electronic Imaging X, ed. B. Rogowitz, T. Pappas, S. Daly, Proc. of SPIE-IST Electronic Imaging, SPIE,, 2005, Volume 5666, pp [13] C. Bartleson and F. Grum, Optical Radiation Measurement. Vol 5 Visual measurements: Academic Press, [14] D. Battle, "The measurement of colour," in Colour physics for industry, R. McDonald. 2nd: Society of Dyers and Colourists, 1997, 2, pp [15] L. Bergman, "Using Multicoloured Halftone Screens for Offset Print Quality Monitoring," Licentiate, Linköping University, Sweden, [16] R. Berns, Billmeyer and Saltzman's principles of color technology: Wiley New York, [17] R. Berns and K. Petersen, "Empirical modeling of systematic spectrophotometric errors," Color Research & Application, vol. 13, 4, pp , [18] F. Billmeyer Jr, "Comparative performance of color-measuring instruments," Applied Optics, vol. 8, 4, pp , [19] F. Billmeyer Jr and P. Alessi, "Assessment of color-measuring instruments," Color Research & Application, vol. 6, 4, pp , [20] N. Bonnier, F. Schmitt, H. Brettel, and S. Berche, "Evaluation of spatial gamut mapping algorithms," in Proc. of 14th Color Imaging Conference, IS&T, Springfield, VA, 2006, IS&T, 14, pp [21] J. Briggs, D. Forrest, and M.-K. Tse, "Reliability Issues for Color Measurement in Quality Control Applications," in Proc. NIP 14: International Conference on Digital Printing Technologies, 1998, IS&T, pp [22] D. Brydges, F. Deppner, H. Kunzil, K. Heuberger, and R. Hersch, "Application of a 3-CCD color Camera for Colorimetric and Densitometric Measurements," in SPIE Proceedings 3300, Color Imaging : Device Independent Color, Color Hardcopy and Graphics Arts III, 1998, Citeseer, pp [23] H. Büring, P. G. Herzog, and E. Jung, "Evaluation of Current Color Management Tools: Image Quality Assessments," in CGIV -- Second European Conference on 120

132 References Color in Graphics, Imaging and Vision, Aachen, Germany, 2004, IS&T, 2, pp [24] Y. Chung, J. Xin, and K. Sin, "Improvement of inter instrumental agreement for reflectance spectrophotometers," Coloration Technology, vol. 120, 6, pp , [25] CIE 15 Technical report, Colorimetry, Commission Internationale de l'éclairage, Vienna, Austria, [26] CIE 116, Industrial colour-difference evaluation, CIE Publication No , Commission Internationale de l'éclairage, Vienna, Austria, [27] CIE 122, The relationship between digital and colorimetric data for computer controlled CRT displays., CIE Publication No , Commission Internationale de l'éclairage, Vienna, Austria, [28] CIE 130, Practical Methods for the Measurement of Reflectance and Transmittance, CIE Publication No , Commission Internationale de l'éclairage, Vienna, Austria, [29] CIE 142, Improvement to Industrial Color-difference Evaluation, CIE Publication No , Commission Internationale de l'éclairage, Vienna, Austria, [30] CIE 156, Guidelines for the Evaluation of Gamut Mapping Algorithms, CIE Publication No , Commission Internationale de l'éclairage, Vienna, Austria, [31] CIE Commission Internationale de l Eclairage. Selected Colorimetric Tables Available: [32] F. Clarke, R. McDonald, and B. Rigg, "Modification to the JPC79 Colour difference formula," Journal of the Society of Dyers and Colourists, vol. 100, 4, pp , [33] P. J. Clarke, "Measurement Good Practice Guide No.96, Surface Colour Measurements," National Physical Laboratory, Teddington, Middlesex, United Kingdom, TW11 0LW, [34] S. Colthorpe and G. Imhoff, "CTP Why densitometers do not work, White paper," Centurfax Ltd and Grip Digital Inc [35] V. de Garcia, J. Inarejos, S. Otero, E. Perales, and F. Martínez-Verdú, "Assessment of the Colorimetric Behaviour of Different Spectrophotometers," in Advances in Printing and Media Technology: Proceedings of the 36th International Research 121

133 References Conference of iarigai vol. 36, N. Enlund and M. Lovrecek. Stockholm, Sweden, [36] DIN , Prüfung von Drucken und Druckfarben der Drucktechnik Farbdichtemessung an Drucken, Deutsches Institut für Normung, [37] F. Dugay, I. Farup, and J. Y. Hardeberg, "Perceptual evaluation of color gamut mapping algorithms," Color Research & Application, vol. 33, 6, pp , [38] P. Engeldrum, Psychometric scaling: a toolkit for imaging systems development: Imcotek Press, Winchester, Massachusetts, USA, [39] P. Engeldrum, "Psychometric Scaling: Avoiding the pitfalls and hazards," in Proc. IS&T s PICS: Image Processing, Image Quality, Image Capture Systems Conference, Montreal, Quebec, Canada, 2001, Citeseer, pp [40] M. Enroth, "Environmental impact of printed and electronic teaching aids, a screening study focusing on fossil carbon dioxide emissions," in Advances in Printing and Media Technology: Proceedings of the 36th International Research Conference of iarigai vol. 36, N. Enlund and M. Lovrecek. Stockholm, Sweden, [41] European Color Initiative (ECI). ICC profiles from ECI Available: [42] M. Fairchild, Color appearance models. Reading, Massachusetts, USA: Addison Wesley Longman, Inc, [43] M. Fairchild and F. Grum, "Thermochromism of ceramic reference tile," Applied Optics, vol. 24, 21, pp , [44] K. Falkenstern, N. Bonnier, H. Brettel, M. Pedersen, and F. Vienot, "Using Image Quality Metrics to Evaluate an ICC Printer Profile.," in Proc. 18th Color Imaging Conference, IS&T, San Antonio, TX, USA, 2010, IS\&T and SID, pp [45] K. Falkenstern, N. Bonnier, M. Pedersen, H. Brettel, and F. Vienot, "Using Metrics to Assess the ICC Perceptual Rendering Intent," in Image Quality and System Performance VIII, San Francisco, CA, 2011, SPIE Proceedings, [46] I. Farup, C. Gatta, and A. Rizzi, "A multiscale framework for spatial gamut mapping," Image Processing, IEEE Transactions on, vol. 16, 10, pp , [47] G. Fechner, Elements of Psychophysics, Vol 1 (1860): (translated by H.E. Adler), Holt, Rinehart and Winston, New York, N.Y,

134 References [48] G. Field, Color and its reproduction: Fundamentals for the Digital Imaging and Printing Industry, 3 ed.: Graphic Arts Technical Foundation, GATF, [49] FOGRA. FOGRA28L.txt, Characterization data for standardized printing conditions Available: [50] B. Fraser, C. Murphy, and F. Bunting, Real world color management, 2 ed.: Peachpit Press, Berkeley, CA, US, [51] J. Gardner, "Comparison of Calibration Methods for Tristimulus Colorimeters," Journal of Research of the National Institute of Standards and Technology, vol. 112, 3, pp , [52] D. A. Garvin, "What Does 'Product Quality' Really Mean?," MIT Sloan Management Review, vol. 26, 1, [53] A. Gatt, S. Westland, and R. Bala, "Testing the softproofing paradigm," in Proc. of 12th Color Imaging Conference, Scottsdale, Arizona, 2004, Citeseer, pp [54] A. Gatt, S. Westland, R. Bala, and Y. Ling, "Testing the Softproofing Paradigm II," AIC Colour 05-10th Congress of the International Colour Association, pp , [55] G. Gescheider, Psychophysics: The Fundamentals, 3 ed. Mahwah, New Jersey: Lawrence Erlbaum Associates, Inc., Publishers, [56] E. Giorgianni and T. Madden, Digital color management: encoding solutions, 2nd ed.: Wiley, [57] E. M. Granger, "Comparison of Color Difference Data and Formulas," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 60th Annual Meeting, San Francisco, CA, USA, 2008, pp [58] P. Green, Color Management: Understanding and Using ICC Profiles. Chichester, West Sussex PO19 8SQ, England: John Wiley & Sons Ltd, [59] P. Green, "A smoothness metric for colour transforms," in Proc. SPIE. Color Imaging XIII: Processing, Hardcopy, and Applications San Jose, CA, USA 2008, 6807, pp I. [60] P. Green, Understanding Digital Color, 2nd ed.: GATF Press, [61] J. Grice and J. Allebach, "The print quality toolkit: An integrated print quality assessment tool," The Journal of imaging science and technology, vol. 43, 2, pp ,

135 References [62] J. Guild, "The colorimetric properties of the spectrum," Philosophical Transactions of the Royal Society of London. Series A, Containing Papers of a Mathematical or Physical Character, vol. 230, pp , [63] J. Handley, "Comparative analysis of Bradley-Terry and Thurstone-Mosteller paired comparison models for image quality assessment," in Proceedings of the Image Processing, Image Quality, Image Capture Systems Conference (PICS-01), Montreal, Quebec, Canada, 2001, Citeseer, pp [64] J. Y. Hardeberg, Acquisition and Reproduction of Color Images, Colorimetric and Multispectral Approaches Dissertation.com, [65] J. Y. Hardeberg, P. Nussbaum, S. Roch, and O. Panak, "Time matters in soft proofing," ACTA GRAPHICA, Journal for Printing Science and Graphic Communication, vol. 19, 171, pp. 1-10, [66] J. Y. Hardeberg and S. Skarsbø, "Comparing color image quality of four digital presses," in 11th International Printing and Graphic Arts Conference, Bordeaux, France., 2002, 11th International Printing and Graphic Arts Conference. [67] A. R. Hill, "How we see colour," in Colour Physics for Industry, R. McDonald. Society of Dyers and Colourists, 1997, 8, pp [68] J. Homann, Digital color management: principles and strategies for the standardized print production: Springer-Verlag New York Inc, [69] G. Hong, M. Luo, and P. Rhodes, "A study of digital camera colorimetric characterization based on polynomial modeling," Color Research & Application, vol. 26, 1, pp , [70] R. Hunt, Measuring colour, 3 ed. Kingston-upon-Thames, England: Fountain Press, [71] R. Hunt, The reproduction of colour, 6 ed. Chichester, West Sussex PO19 8SQ, England: John Wiley & Sons Ltd, [72] ICC Specification ICC.1: (Profile version ) Image technology colour management Architecture, profile format, and data structure. Available: November [73] IEC , Multimedia System and Equipment - Colour Measurement and Management, Part 2-5: Colour management Optional RGB colour space oprgb, International Electrotechnical Commission,

136 References [74] F. Imai, R. Berns, and D. Tzeng, "A comparative analysis of spectral reflectance estimated in various spaces using a trichromatic camera system," Journal of Imaging Science and Technology, vol. 44, 4, pp , [75] J. Imai and M. Omodani, "Reasons why we prefer reading on paper rather than displays: Studies for seeking paper-like readability on electronic paper," Journal of Imaging Science and Technology, vol. 52, 5, pp. 1-5, [76] Intergraf, "The Future of the European Printing Industry-in our own Hands," Facta Consult, Intergraf, Brussel, Belgium, [77] ISO , Graphic technology Colour and transparency of printing ink sets for four colour printing Part 1: Sheet-fed and heat-set web offset lithographic printing, International Organization for Standardization, Geneva, [78] ISO 3664, Graphic technology and photography Viewing conditions, International Organization for Standardization, Geneva, [79] ISO 3664, Graphic technology and photography Viewing conditions, International Organization for Standardization, Geneva, [80] ISO , Paper and board -- Measurement of specular gloss -- Part 1: 75 degree gloss with a converging beam, TAPPI method, International Organization for Standardization, Geneva, [81] ISO 9000, Quality management systems -- Fundamentals and vocabulary, International Organization for Standardization, Geneva, [82] ISO , Graphic technology -- Input data for characterization of 4-colour process printing -- Part 2: Expanded data set, International Organization for Standardization, Geneva, [83] ISO 12646, Graphic technology Displays for colour proofing Characteristics and viewing conditions, International Organization for Standardization, Geneva, [84] ISO , Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 1: Parameters and measurement methods, International Organization for Standardization, Geneva, [85] ISO , Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 2: Offset printing processes, International Organization for Standardization, Geneva,

137 References [86] ISO , Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 3: Coldset offset lithography on newsprint, International Organization for Standardization, Geneva, [87] ISO , Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 7: Proofing processes working directly from digital data, International Organization for Standardization, Geneva, [88] ISO 13656, Graphic technology -- Application of reflection densitometry and colorimetry to process control or evaluation of prints and proofs, International Organization for Standardization, Geneva, [89] ISO 14001, Environmental management systems -- Requirements with guidance for use, International Organization for Standardization, Geneva, [90] ISO , Image technology colour management -- Architecture, profile format and data structure -- Part 1: Based on ICC.1: , International Organization for Standardization, Geneva, [91] ISO , Graphic technology -- Prepress digital data exchange using PDF -- Part 4: Complete exchange of CMYK and spot colour printing data using PDF 1.4 (PDF/X-1a) International Organization for Standardization, Geneva, [92] ISO , Graphic technology -- Prepress digital data exchange using PDF -- Part 6: Complete exchange of printing data suitable for colour-managed workflows using PDF 1.4 (PDF/X-3), International Organization for Standardization, Geneva, [93] ISO , Graphic technology -- Prepress digital data exchange using PDF -- Part 7: Complete exchange of printing data (PDF/X-4) and partial exchange of printing data with external profile reference (PDF/X-4p) using PDF 1.6 International Organization for Standardization, Geneva, [94] ISO , Graphic technology -- Prepress digital data exchange using PDF -- Part 8: Partial exchange of printing data using PDF 1.6 (PDF/X-5) International Organization for Standardization, Geneva, [95] ISO 23603, Standard method of assessing the spectral quality of daylight simulators for visual appraisal and measurement of colour, International Organization for Standardization, Geneva,

138 References [96] ISO/IEC 13660, Information technology -- Office equipment -- Measurement of image quality attributes for hardcopy output -- Binary monochrome text and graphic images, International Organization for Standardization, Geneva, [97] ISO/TS 10128, Graphic technology -- Methods of adjustment of the colour reproduction of a printing system to match a set of characterization data, International Organization for Standardization, Geneva, [98] K. Johansson, P. Lundberg, and R. Ryberg, A guide to graphic print production, 2nd ed. Hoboken, New Jersey: Wiley & Sons, Inc., [99] T. Johnson, Colour management in graphic arts and publishing. Surrey, UK: Pira International, [100] T. Johnson, "Methods for characterizing colour scanners and digital cameras," Displays, vol. 16, 4, pp , [101] H. Kang, Color technology for electronic imaging devices: SPIE-International Society for Optical Engineering, [102] N. Katoh, "Practical method for appearance match between soft copy and hard copy," in Device-Independent Color Imaging, Proceeding of SPIE, 1994, SPIE, 2170, pp [103] B. W. Keelan, Handbook of Image Quality: Characterization and Prediction. New York NY 10016: Marcel Dekker, Inc., [104] Y. Kipman, "Image quality metrics for printers and media," in IS&T s Image Processing, Image Quality, Image Capture, Systems Conference, Portland, Oregon, USA, 1998, The Society for Imaging Science and Technology, 1, pp [105] H. Kipphan, Handbook of print media: technologies and production methods: Springer-Verlag Berlin Heidelberg, [106] R. Kuehni, "CIEDE2000, milestone or final answer?," Color Research & Application, vol. 27, 2, pp , [107] R. Kuehni, "Towards an improved uniform color space," Color Research & Application, vol. 24, 4, pp , [108] H. Kuenzli, F. Deppner, K. Heuberger, and Y. Jiang, "Mini-Targets: A New Dimension In Print Quality Control," in Proc. Color imaging: device-independent color, color hardcopy, and graphic arts III, San Jose, California, 1998, Society of Photo Optical, 3300, pp

139 References [109] S. Leckner and S. Nordqvist, "Soft Proofing using LCDs-Case Newspaper Workflow," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 54th Annual Meeting, 2002, TAGA; 1998, pp [110] S. Lindberg, D. C. Donderi, and N. Pauler, "Image quality of four digital printing methods compared to offset and flexographic printing. In Advances in Color Reproduction," in Proceedings of the 28th International Iarigai Research Conference, chapter 4.1., Sewickley, PA, US, 2001, GATF Press. [111] M. Luo, G. Cui, and B. Rigg, "The development of the CIE 2000 colour difference formula: CIEDE2000," Color Research & Application, vol. 26, 5, pp , [112] M. Luo, G. Cui, and B. Rigg, "Further comments on CIEDE2000," Color Research & Application, vol. 27, 2, pp , [113] D. MacAdam, "Visual sensitivities to color differences in daylight./. opt," Soc. Amer, vol. 32, pp , [114] M. Mahy, L. Van Eycken, and A. Oosterlinck, "Evaluation of uniform color spaces developed after the adoption of CIELAB and CIELUV," Color Research and Application, vol. 19, 2, pp , [115] G. Marcu, "Color Quality in Desktop Printing," in IS&T s PICS: Image Processing, Image Quality, Image Capture Systems Conference, Portland, Oregon, USA, 2000, IS&T, Tutorial Notes. [116] G. Marcu, "Factors Governing Print Quality in Color Prints," in Proc. IS&T s PICS: Image Processing, Image Quality, Image Capture Systems Conference, Springfield, VA, USA, 1998, IS&T, pp [117] D. Martin, J. O'Neill, T. Colombino, F. Roulland, and J. Willamowski, "'Colour, itís Just a Constant Problem': An Examination of Practice, Infrastructure and Workflow in Colour Printing," in From CSW to Web 2.0: European Developments in Collaborative Design, Computer Supported Cooperative Work, D. Randall and P. Salembier. London: Springer-Verlag, 2010, 2, pp [118] Microsoft Corporation. Windows Color Quality Specifications for Printer OEMs. Part of Microsoft Hardware Quality Labs (WHQL) s Windows Color Quality Test Kit Available: [119] Å. Moberg, C. Borggren, G. Finnveden, and S. Tyskeng, "Effects of a total change from paper invoicing to electronic invoicing in Sweden," Report from the KTH Centre for Sustainable Communications TRITA-SUS, vol. 3,

140 References [120] Å. Moberg, M. Johansson, G. Finnveden, and A. Jonsson, Screening environmental life cycle assessment of printed, web based and tablet e-paper newspaper: Centre for Sustainable Communications, Royal institute of technology (KTH), [121] J. Morovic, Color Gamut Mapping. Chichester: John Wiley & Sons, Ltd, [122] J. Morovic, "Colour gamut mapping," in Colour Engineering Achiving Device Independent Colours, P. Green and L. MacDonald. Chichester, West Sussex PO19 8SQ, England: John Wiley & Sons Ltd, 2002, pp [123] J. Morovic, "Gamut mapping," in Digital Color Imaging Handbook., G. Sharma. CRC Press, 2003, Chapter 10, pp [124] J. Morovic and M. Luo, "The fundamentals of gamut mapping: A survey," Journal of Imaging Science and Technology, vol. 45, 3, pp , [125] J. Morovic and M. Luo, "The Pleasantness and Accuracy of Gamut Mapping Algorithms," in ICPS Conference 1998, 2, pp [126] J. Morovic and P. Nussbaum, "Factors affecting the appearance of print on opaque and transparent substrates," Journal of Imaging Science and Technology, vol. 47, 6, pp , [127] A. Murray, "Monochrome reproduction in photoengraving," Journal of the Franklin Institute, vol. 221, 6, pp , [128] NADA Support Lineær avisproduksjon kort og godt. [Guideline]. Available: retrieved 13. April [129] O. Norberg, P. Westin, S. Lindberg, M. Klaman, and L. Eidenvall, "Comparison of print quality between digital and traditional technologies," in In Proceedings of DPP2001: International Conference on Digital Production Printing and Industrial Applications, 2002, pp [130] P. Nussbaum and J. Y. Hardeberg, "Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 60th Annual Meeting, San Francisco, CA, USA, 2008, pp [131] P. Nussbaum and J. Y. Hardeberg, "Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print," Color Research & Application, Article first published online:march 8 th 2011, DOI: /col [132] P. Nussbaum and J. Y. Hardeberg, "Print quality evaluation and applied colour management in heat-set web offset," in Advances in Printing and Media Technology: Proceedings of the 33rd International Research Conference of 129

141 References iarigai. vol. 33, N. Enlund and M. Lovrecek. Acta Graphica Publishers, 2006, pp [133] P. Nussbaum, J. Y. Hardeberg, and F. Albregtsen, "Regression based characterization of Color Measurement instruments in printing applications," in Electronic Imaging, Color Imaging XVI: Displaying, Processing, Hardcopy, and Applications, San Francisco, CA, 2011, SPIE Proceedings, [134] P. Nussbaum, J. Y. Hardeberg, and S. E. Skarsbø, "Print quality evaluation for governmental purchase decisions," in Advances in Printing Science and Technology: Proceedings of the 31st International Research Conference of iarigai vol. 31, M. Lovrecek. Acta Graphica Publishers, 2004, pp [135] P. Nussbaum and J. Morovic, "Factors Affecting The Appearance Of Print On Opaque Substrates," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 55th Annual Meeting, Montreal, Canada, 2003, pp [136] P. Nussbaum, A. Sole, and J. Y. Hardeberg, "Analysis of colour measurement uncertainty in a colour managed printing workflow," Journal of Print and Media Technology Research, iarigai, Accepted for publication. [137] P. Nussbaum, A. Sole, and J. Y. Hardeberg, "Consequences of using a number of different color measurement instruments in a color managed printing workflow," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 61st Annual Meeting, New Orleans, LA, USA, 2009, pp [138] Y. Ohno, "CIE Fundamentals for color measurements," in Proc. NIP 16: International Conference on Digital Printing Technologies, Vancouver, CA, 2000, IS&T, pp [139] N. Ohta and A. Robertson, Colorimetry: Fundamentals and Applications: J. Wiley, [140] Z. Pan, Y. Noyes, J. Hardeberg, L. Lee, and G. Healey, "Color scanner characterization with scan targets of different media types and printing mechanisms," in Color Imaging: Device-Independent Color, Color Hardcopy, and Graphic Arts VI, San Jose, CA, 2001, SPIE Proceedings, 4300, pp [141] O. Panak, P. Nussbaum, and J. Y. Hardeberg, "Colour Memory Match Under Disparate Viewing Conditions," in Fifteenth Color Imaging Conference: Color Science and Engineering Systems, Technologies, and Applications, IS&T, Albuquerque, New Mexico, 2007, pp

142 References [142] D. R. Pant and I. Farup, "Riemannian Formulation of the CIEDE2000 Color Difference Formula," in Proc. 18th Color Imaging Conference, IS&T, San Antonio, TX, USA, 2010, IS&T, pp [143] M. Pedersen, Internationalt Standardiseret Grafisk Produktion ISO 12647: Grafisk Arbejdsgiverforening, [144] M. Pedersen and S. A. Amirshahi, "Framework for the Evaluation of Color Prints Using Image Quality Metrics," in Proc. 5th European Conference on Colour in Graphics, Imaging, and Vision (CGIV), Joensuu, Finland, 2010, IS&T, pp [145] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Attributes of a New Image Quality Model for Color Prints," in Proc. 17th Color Imaging Conference, IS&T, Albuquerque, NM, USA, 2009, IS&T, pp [146] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Attributes of image quality for color prints," Journal of Electronic Imaging, vol. 19, 1, pp , [147] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Estimating Print Quality Attributes by Image Quality Metrics," in Proc. 18th Color Imaging Conference, IS&T, San Antonio, TX, USA, 2010, IS&T, pp [148] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Validation of Quality Attributes for Evaluation of Color Prints," in Proc. 18th Color Imaging Conference, IS&T, San Antonio, TX, USA, 2010, IS&T, pp [149] G. Radencic, E. Neumann, and M. Bohan, "Spectrophotometer inter-instrument agreement on the color measured from reference and printed samples," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 60th Annual Meeting, San Francisco, CA, 2008, pp [150] R. Rasmussen, "Measuring Color Printer Image Quality," in Tutorial Notes, IS&T/SID 9th Color Imaging Conference, Springfield, VA, USA, 2000, course SC4. [151] D. C. Rich, "Instruments and methods for colour measurement," in Colour engineering: achieving device independent colour, P. J. Green and L. W. MacDonald. Wiley, [152] D. C. Rich, Y. Okumura, and V. Lovell, "The Effect of Spectrocolorimeter Reproducibility on a Fully Color-Managed Print Production Workflow," in 4th European Conference on Color in Graphics, Imaging, and Vision (CGIV), Barcelona, Spain 2008, Society for Imaging Science and Technology, pp

143 References [153] T. Rijgersberg. Calculation from the original experimental data of the CIE 1931 RGB standard observer spectral chromaticity co-ordinates Available: [154] D. Roberts, Signals and perception: the fundamentals of human sensation: Palgrave, [155] A. Robertson, "Diagnostic performance evaluation of spectrophotometers," in Advances in standards and methodology in spectrophotometry. vol. 2, B. C. a. M. K.D. Elsevier Science Publisher B.V., Amsterdam, 1987, pp [156] S. Roch, J. Y. Hardeberg, and P. Nussbaum, "Effect of time spacing on the perceived color," in Proc. SPIE: Color Imaging XII: Processing, Hardcopy, and Applications 2007, The International Society for Optical Engineering, Color Science, 6493, pp [157] J. Rodgers, K. Wolf, N. Willis, D. Hamilton, R. Ledbetter, and C. Stewart, "A comparative study of color measurement lntstrumentation," Color Research & Application, vol. 19, 5, pp , [158] D. Romano, "The Image Analyzer A True Dot Area Meter?," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 51st Annual Meeting, Vancouver, Canada, 1999, pp [159] J. Schanda, Colorimetry: Understanding the CIE system: Wiley-Interscience, [160] J. Schanda, G. Eppeldauer, and G. Sauter, "Tristimulus Color Measurement of self- Luminous Sources," in Colorimetry: Understanding the CIE system, J. Schanda. Wiley-Interscience, 2007, pp [161] K. Schläpfer, Farbmetrik in der Reproduktionstechnik und im Mehrfarbendruck, 2 ed.: UGRA, [162] U. Schmitt and F. Dolezalek, Ugra/FOGRA Media Wedge CMYK V 2.0, Munich, [163] M. Scorer, M. Luo, and B. Rigg, "Performance monitoring of colour measuring instruments," Coloration Technology, vol. 118, 4, pp , [164] J. Seymour, "Color measurement with an RGB camera," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 61st Annual Meeting, New Orleans, LA, USA, 2009, pp [165] J. Seymour, "How Many Δ Es Are There in a ΔD?," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 59th Annual Meeting, Pittsburgh, PA, USA, 2007, pp

144 References [166] A. Sharma, "Methodology for Evaluating the Quality of ICC Profiles-Scanner, Monitor, and Printer," Journal of Imaging Science and Technology, vol. 50, p. 469, [167] A. Sharma, Understanding Color Management. Thompson Delmar Learning: New York, [168] A. Sharma and P. Fleming, "Measuring the Quality of ICC Profiles and Color- Management Software," Seybold Report, vol. 2, 19, p. 3, [169] G. Sharma, Digital color imaging handbook: New York, CRC, Press, [170] G. Sharma and H. Trussell, "Digital color imaging," IEEE Transactions on Image Processing, vol. 6, 7, pp , [171] D. Silverstein and J. Farrell, "The relationship between image fidelity and image quality," in Proc. IEEE Int. Conference Image Processing, Lausanne, Switzerland, 1996, IEEE, 1, pp [172] A. Sole, P. Nussbaum, and J. Y. Hardeberg, "Implementing ISO standards for soft proofing in a standardized printing workflow according to PSO," in Advances in Printing and Media Technology: Proceedings of the 37th International Research Conference of iarigai, vol. 37, N. Enlund and M. Lovrecek. International Association of Research Organizations for the Information, Media and Graphic Arts Industries, 2010, pp [173] M. Solli, M. Andersson, R. Lenz, and B. Kruse, "Color measurements with a consumer digital camera using spectral estimation techniques," Image Analysis, pp , [174] T. Steder, M. R. Luo, and Li;Changjun, "Training Data Selection Study for Surface Colour Measurement Data Correlation," in CGIV Fourth European Conference on Colour in Graphics, Imaging, and MCS/08 Vision 10th International Symposium on Multispectral Colour Science, Terrassa Barcelona, España, 2008, Society for Imaging Science and Technology. [175] S. Stevens, "On the theory of scales of measurement," Science, vol. 103, 2684, pp , [176] S. Stevens, Psychophysics: Wiley New York, [177] M. Stokes, "The impact of color management terminology on image quality," in PICS Conference, Portland, Oregon, USA, 1998, Society for Imaging Science & Technology, pp

145 References [178] M. Stokes, "Industry adoption of color management systems," in Proceedings of the 8th Concress of the International Colour Association, AIC Color 97, Kyoto, Japan, 1997, Vol I, pp [179] M. Stokes, M. Fairchild, and R. Berns, "Colorimetrically quantified visual tolerances for pictorial images," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 44th Annual Meeting, Williamsburg, VA, USA, 1992, pp [180] B. Streckel, B. Steuernagel, E. Falkenhagen, and E. Jung, "Objective print quality measurements using a scanner and a digital camera," in DPP2003: IS&Ts International Conference on Digital Production Printing and Industrial Applications, Barcelona, Spain, 2003, 5, pp [181] L. Thurstone, "A law of comparative judgment," Psychological review, vol. 34, 4, pp , [182] Tobias Associates Inc Reflection Densitometry. Available: [183] W. Torgerson, "A law of categorical judgment," Consumer behavior, pp , [184] UGRA. (2009, 6. July 2010). PSO certification. Available: [185] A. Valberg, Light vision color: Wiley, [186] J. F. Verrill, P. J. Clarke, and J. O'Halloran, "Study of Improved Methods for Absolute Colorimetry, Fifth Report, Study of Colorimetric Errors in Industrial Instruments," Centre for Optical and Environmental Metrology, National Physical Laboratory, Teddington, Middlesex, United Kingdom, TW11 0LW, [187] H. Völz, Industrial color testing: fundamentals and techniques, 2nd ed.: Wiley, [188] B. Wandell, Foundations of vision: Sinauer Associates, [189] Z. Wang and A. Bovik, Modern image quality assessment: Morgan & Claypool Publishers, [190] Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, "Image quality assessment: From error visibility to structural similarity," Image Processing, IEEE Transactions on, vol. 13, 4, pp ,

146 References [191] S. Westland and C. Ripamonti, Computational colour science using MATLAB: Wiley, [192] W. D. Wright, "A re-determination of the trichromatic coefficients of the spectral colours," Transactions of the Optical Society, vol. 30, p. 141, [193] M. Wroldsen, P. Nussbaum, and J. Y. Hardeberg, "Densitometric and Planimetric Measurement Techniques for Newspaper Printing," in Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 59th Annual Meeting, Pittsburgh, PA, USA, 2007, pp [194] M. Wroldsen, P. Nussbaum, and J. Y. Hardeberg, "Densitometric and Planimetric Measurement Techniques for Newspaper Printing " Journal TAGA - Technical Association of the Graphic Arts, vol. 4, pp , [195] D. R. Wyble. (2010, 24. August 2010). Color Measurement - Introduction, Historical Perspective, Definitions and Terminology, Components of a Spectrophotometer, Light Source, Detector, Dispersing Element. Available: [196] D. R. Wyble and D. Rich, "Evaluation of methods for verifying the performance of color-measuring instruments. Part I: Repeatability," Color Research & Application, vol. 32, 3, pp , [197] D. R. Wyble and D. Rich, "Evaluation of methods for verifying the performance of color-measuring instruments. Part II: Inter-instrument reproducibility," Color Research & Application, vol. 32, 3, pp , [198] G. Wyszecki and W. Stiles, Color science, Concepts and Methods, Quantitative Data and Formulae, 2 ed.: Wiley New York, [199] S. Yendrikhovskij, "Image quality: Between science and fiction," in In Proc. IS&T s PICS Conference, Savannah, Georgia, US., 1999, IS&T, pp [200] J. Zwinkels, "Colour-measuring instruments and their calibration," Displays, vol. 16, 4, pp ,

147 References 136

148 PART II INCLUDED PAPERS 137

149 138

150 Paper A Paper A Peter Nussbaum, Jon Yngve Hardeberg and Sven Erik Skarsbø Print Quality Evaluation for Governmental Purchase Decisions Published In Advances in Printing Science and Technology: Proceedings of 31 st International iarigai Research Conference, Volume 31, pp , M. Lovreček, Ed., Acta Graphica Publishers,

151 Paper A 140

152 Paper A Print Quality Evaluation for Governmental Purchase Decisions Peter Nussbaum, Jon Y. Hardeberg and Sven Erik Skarsbø The Norwegian Color Research Laboratory Faculty of Computer Science and Media Technology Gjøvik University College P.O.Box 191 N-2802 Gjøvik Norway [email protected] [email protected] [email protected] Abstract Potential costumers within the digital printer market have various demands considering their requirements with regards to desired print quality. This paper aims to investigate the print quality according to predefined quality factors to determine the appropriate printing equipment. In particular the study describes the methods and results of a research project conducted by researchers at the Norwegian Color Research Laboratory for the Norwegian Government Administration Service. The objective of the project was to develop methods and procedures to perform test prints from digital printers including the evaluation and assessment of the print quality according to predefined quality factors. The results of the study indicate the performance of various digital printers in terms of their obtained print quality, which will be useful for the governmental purchase decisions. 1 Introduction In 1997 the design program for the Norwegian Ministries was established, and a design handbook was produced (Kulturdepartementet, 2000). The Ministry for Culture and Church Affairs is responsible for the design program and the ministry has delegated the responsibility for the administration, the daily running and the development of the design program to the Norwegian Government Administration Service (NGAS). NGAS is also technical editor for governmental publications, and supports all of the ministries in the printing of publications. Amongst other things, the design handbook describes and defines the colours used in the ministries logos (National Coat-of-Arms). In order to maintain 141

153 Paper A control over colour reproduction of these logos it was decided to use pre-printed letterheads; letterheads produced by offset printing. Originally, the three colours of the national logo have been specified with regard to PMS (Pantone Matching System). Although this printing technique ensures the colour accuracy of the colour logo, it is an expensive and time consuming process. On the other hand, after several years of market hesitation, digital presses have now become common and in today s printing market, where flexibility, variable content, shorter lead times, and on demand publishing, are being demanded, digital presses represent an attractive supplement to conventional offset presses (Hardeberg and Skarsbø, 2002). Consequently, the Ministry for Transport and Communications has decided to embrace this new technology and replace the conventional offset process by digital printing systems. In order to maintain the colour reproduction demands, as outlined in the design handbook, the Norwegian Color Research Laboratory at Gjøvik University College in collaboration with NGAS has carried out a project to develop methods to evaluate digital printers. Consequently, the aim of the presented work has been to develop methods and quality factors to evaluate digital printers in terms of their potential impact on print quality, in accordance with the quality requirements of the NGAS. Furthermore, the development of digital test documents and test procedures to obtain the evaluation results is an important part of this study. Although other aspects, such as the commercial and the technical point of view, might be considered for the acquisition of the appropriate equipment, the results of the printer evaluation will be used by the NGAS to determine the final purchase decisions. We also believe that not only the results but also the methodology will be of interest to other target groups and who are considering the replacement of conventional printing by digital printing. NGAS, as the costumer of this project, nationally announced a printer quality contest and printer manufactures were invited to participate in the test. However, due to the current purchase process by the NGAS, the names of the participating printer manufactures and their printing systems cannot be published in this work. In the reminder of this study the next section gives a short overview in the field of colour image quality. Then section three introduces the experimental method used for evaluating the print quality in terms of the predefined quality factors. Then in section four the data analysis performed on the experimental data is described, followed by a presentation of the 142

154 Paper A results. Finally in the section five the implications of the results will be discussed and ideas will be suggested for future work. 2 Colour Image Quality Various studies and research have been done in the field of print quality and repeatedly it has been concluded as a very complex issue. The subject has been discussed at various conferences (see e.g. Stokes, 1998; Hardeberg and Skarsbø, 2002). In another study by Marcu (2000) addresses various quality factors that determine the print quality such as printing technology, colorant/media interaction, geometric resolution, halftoning, separation, black generation, UCR, GCR and tone reproduction. Principally there are two different methods to assess image quality. The first method is by measurement, using instruments to determine values for the various quality factors. Considering the evaluation of relative colorimetric colour reproduction of printing devices Microsoft has proposed a quality assurance system (Microsoft Corporation, 2001). The second method is based on observation, using psychophysical experiments to gather the judgement of human observers. For instance, the pair comparison method is a robust approach where observers are asked to compare the perceptual magnitude between two stimuli (or pairs of stimuli). This method is based on Thurstone s «law of comparative judgement» (Handley, 2001; Gescheider, 1985). To produce an accurate assessment of image quality the values obtained by measurement and observation metrics can be calculated. The term metric is generally applied to any physical and psychophysical measure of image quality. Image quality metrics can be defined as numbers, derived from physical measurements that can be related to the perceptions of image quality (Jacobson, 1995). Compared to a number of other studies, whose aim is the comparison of colour image quality of digital presses concentrating in the field of rendering complex images, the main focus in this work is printer evaluation according to reproduction requirements of the Norwegian ministries logo (National Coat-of-Arms). Hence, considering the printer evaluation in this work, NGAS has most emphasis the colour logo reproduction (Figure 1). 143

155 Paper A Figure 1: The design handbook describes and defines the colours used in the Norwegian ministries logo (National Coat-of-Arms). 3 Experimental Method Considering the methods to assess the digital print quality in this study we have proposed a number of quality factors. According to the presented list of appropriate quality factors NGAS has determined the most significant quality factors including their weights. Then the selected quality factors have been divided into three assessment categories, namely, visual logo assessments (Section 3.1), general print quality assessments (Section 3.2) and finally the copy quality assessments (Section 3.3). To complete the final ranking the results of the three categories have been weighted once again. According to the priorities of the quality requirements defined by NGAS the category visual logo assessments has been weighted by 50%, the category general print quality assessments by 30% and finally the third category copy quality assessments by 20%. The quality evaluation and the final ranking were done at the Color Research Laboratory at Gjøvik University College, using psychophysical experiments (expert panel) and quantitative analyses based on measurements. The psychophysical evaluations were performed to determine the colour logo reproduction in terms of colour match, visual resolution, surface texture, logo alignment and artefacts. The quantitative analyses included text quality, colour gamut, repeatability and register. The next three sections are describing the assessment categories including the defined quality factors. 3.1 Visual logo assessment As mentioned in the previous section, the Color Research Laboratory has proposed a number of quality factors to determine the colour reproduction performance of digital printers. In the first category, which is considering the visual assessment of the colour logo reproduction, Table 1 lists the selected quality factors and the corresponding weight 144

156 Paper A defined by the NGAS. It can be seen that the quality factor colour match has absolute highest priority. Table 1: Quality factors within the category visual logo assessments and the corresponding weights. Quality factors Weight % Colour match 50 Visual resolution 20 Surface texture 10 Logo alignment 10 Artefacts 10 The next paragraph explains the quality factors in more detail. Colour match: The aim of this factor is to obtain, as close as possible, a colour match between the logo colours defined by the three Pantone colours and the test prints. The competitors have been asked to adjust the source file with the three logo colours according to the supplied Pantone colour patches to achieve a closest possible colour match between the test print and the Pantone target. Visual resolution: Detail reproduction in the colour logo such as the «tongue in the lions head» (Figure 1), sharp lines and smooth edges have an impact in the quality of the logo reproduction. Surface texture: Observation of graininess, texture of the colours, smoothness of the image and banding. Logo alignment: Positioning of the colour logo according to the requirements. The colour logo placed on the A4 document has to be aligned centre and 10mm from the top. Artefacts: Observation of unwanted effects on the print such as stripes, scratches and marks. A psychophysical experiment was carried out to evaluate the visual quality of the colour logo reproduction that the printers can obtain. The test prints have been visually assessed according to standardised illumination and viewing conditions (ISO 3664 P1). Before performing the visual assessment the expert panel, including three members from the NGAS and one member from the Color Research Laboratory, have conducted the ISHIHARA Test for Colour Deficiency to confirm their appropriate colour vision. 145

157 Paper A The magnitude estimation method was employed to judge each quality factor. The four observers gave individual ratings on a scale from excellent (6) to very poor (0). The individual marks obtained by the four experts have been collected and the average result listed by ranking. 3.2 General print quality assessments Within this category, four quality factors have been proposed to assess the print quality independently of the Norwegian ministries logo. Table 2 presents the selected quality factors and the corresponding weights. Note that NGAS has weighted all factors equal. Table 2: Quality factors within the category general print quality assessments and the corresponding weights. Quality factors Weight % Text quality 25 Colour gamut 25 Repeatability 25 Register 25 Text quality: This quality factor analyses the performance of the text reproduction including negative text in various text sizes (3 pica point up). In addition the rendering of text on coloured background, bitmap and bitmap illustrations are further assessed. The assessment method used is based on observation from the expert panel rating on a scale from excellent (6) to very poor (0). Colour gamut: The colour gamut defines the range of colours achievable on a certain substrate under the predefined viewing conditions. Colour difference in terms of ΔE* ab between the logo colours (Pantone 280M, Pantone 185M and Pantone 116M) and the corresponding logo colours printed by the competitors is part of the assessment. A printer with a larger colour gamut than another is typically able to reproduce more saturated colours. However considering the predefined Pantone colours in the colour logo, it is not expected that a larger colour gamut necessarily include the target colours because the match will be determined due to the location of the Pantone colour and the device colour gamut within the entire colour space. The data used in this task are based on colorimetric measurements, which have been analysed and visualised by the icc3d application (Farup et al., 2002). 146

158 Paper A Repeatability: To determine the variation in terms of colour stability over time the quality factor repeatability is required (Morovic and Nussbaum, 2003). Essential, a Microsoft Word document (102 pages), representing an ordinary file including text, images, graphs, illustrations and two colour test charts (inserted on page 6 and page 99) was the source. To quantify the short-term repeatability of the printer, the Word document was printed five times in sequence. This results in ten measurements of the colour test chart with an interval of 93 pages and 9 pages respectively. The variations between the measurements are given in ΔE* ab units. Register: The register between individual colour image layers influences the print quality of the colour reproduction (Field, 1999). Variations in register can result in loss of resolution and sharpness. Figure 2 illustrates a sample of nonius register marks (Ifra, 2002) to determine the register deviations in the print. The evaluation to determine the register performance was carried out by the use of a microscope according to ISO Figure 2: Variations in register can be determined by the aid of nonius register marks (Ifra, 2002). As mentioned in the previous section NGAS is responsible for the design program (design handbook), which describes and defines the colours used in the ministry logo (Figure 1). The three colours of the national logo, accepted by NGAS as «the optimised logo colours», refer to Pantone 280M, Pantone 185M and Pantone 116M. Table 3 shows the CIELABvalues of the three logo colours obtained by colour measurement at the Norwegian Color Research Laboratory using a spectrophotometer, SpectroLino (2, D50). These obtained CIELAB-values for optimised logo colours are registered in the design program. 147

159 Paper A Table 3: CIELAB-values (2, D50) for optimised logo colours registered in the design program. L* a* b* Pantone 280M Pantone 185M Pantone 116M Although this category combines visual observations and the method using instruments to determine values for some of the quality factors, a scaling from excellent (6) to very poor (0) has been employed. 3.3 Copy quality assessments Finally in the last category, three quality factors have been analysed determining the quality performance of the copy function of the printers. The evaluation method used is based on visual assessment with individual marks given by the expert panel rating on a scale from excellent (6) to very poor (0). The quality factors in this test and the corresponding weights are listed in Table 4. Table 4: Quality factors within the category copy quality assessments and the corresponding weights. Quality factors Weight % Colour match between original print and copy 40 Copy accuracy 40 Artefacts 20 Colour match between original print and copy: The original print for this task is the Word document with 102 pages printed from the supplied digital file. Subsequently a copy has been produced, by using the copy function of the printer. Copy accuracy: The rendering ability in the copy reproduction of text and illustrations has been assessed. Artefacts: Undesirable print artefacts, which appear on the copy print such as dust, stripes, scratches and marks have been evaluated. 148

160 Paper A 3.4 Test targets As mentioned previously, NGAS, as the costumer of this project, nationally announced a printer quality contest and printer manufactue r s were invited to participate in the test. Consequently methods had to be designed, including the development of digital test documents and test procedures, for the competitors that have applied to conduct the test. The requirements for conducting the test were defined in four tasks. Included on a CD- ROM each competitor has been asked to print four different test documents. The first test document, defined in pdf-format, srgb colour space and in size A3, contains test images, colour charts and elements for the register control. The second target, shown in Figure 3, was specified in pdf-format, CMYK for offset (coated paper) and in size A3. It consists of test images, colour charts and other control elements according to the CMYK colour space. The third document is a Microsoft Word file, A4 in size, containing 102 pages with text, illustrations and colour test charts. This test file should simulate a customised text document completed by the NGAS. The fourth document is an eps-file consisting of the three Pantone colours, which represent the national colour logo (National Coat-of-arms). The participants have been asked to adjust the Pantone colours in the file to obtain a closest possible match between the print and the «original Pantone patches». NGAS has provided the three Pantone colour patches Pantone 280M, Pantone 185M and Pantone 116M. Finally, the last task in the test, the competitors have been asked to copy the test prints from the Microsoft Word file on the same equipment as the test prints have been made. Note that the presented test documents contains test elements such as complex images, which are not taken into consideration within the test. However, these elements could be used in further image quality assessments. Figure 3: The CMYK test target designed for the printer test. 149

161 Paper A The competitors were allowed to choose their customized printing parameters such as driver settings and RIP software, with the goal of achieving their best colour image quality. Considering the substrate, NGAS has supplied all the participants with the appropriate paper, which has been required conducting the print test. The scanning function of the printers has not been part of the evaluation. 4 Results In this section the results of the printer evaluation will be presented. To recap the aim of this work is to develop methods including quality factors to the evaluation of digital printers in terms of their potential impact on print quality, in accordance with the quality requirements of the NGAS. The Color Research Laboratory received test prints from a total of six participants to evaluate their print quality. To avoid any prejudices throughout the evaluation process NGAS anonymised the test prints by identifying the competitors with A, B F. Firstly the results of the three categories individually will be presented before the final ranking will be shown. Table 5 presents the results according to the visual evaluation of the six test prints. Considering the visual logo assessments the experimental results identify competitor A as the candidate that scored best in all the quality criteria in this category. Table 5: Results of the category visual logo assessments of six test prints. Part 1: Visual logo assessments Quality factors Weight A B C D E F Colour match 50% Visual resolution 20% Surface texture 10% Logo alignment 10% Artefacts 10% Total Weighted mean Ranking

162 Paper A Especially for the quality factor colour match, which the NGAS has prioritised highest, competitor A showed excellent performance. By presenting the results of the other competitors, only B and C could match the minimal requirements of the expert panel. Although the competitors D, E and F have shown reasonable results for the quality factors «surface texture», «register» and «artefacts» the performance regarding «colour match» was rather poor. In fact, the expert panel has decided that competitor D, E and F will be excluded in the further evaluation process because they couldn t match the requirements of this category at all. Table 6 presents the results of the second category, namely, the evaluation of quality factors, which are not direct related to the colour logo. Nevertheless these quality factors contribute to the entire image quality and may affect the rendering of the colour logo. Table 6: Results of the category general print quality assessments of the competitors A, B and C. Part 2: General print quality assessments Quality factors Weight A B C Text quality 25% Colour gamut 25% Repeatability 25% Register 25% Total Weighted mean Ranking It can be seen, that competitor B performs best in this category followed by competitors C and A. Due to the imperfect repeatability performance obtained by competitor A the total score within this category is rather poor. However, considering the quality factor colour gamut competitor A has shown the largest range of colours achievable compared to the other two competitors. 151

163 Paper A Table 7: Results of the category copy quality assessments of the competitors A, B and C. Part 3: Copy quality assessments Quality factors Weight A B C Colour match between original print and copy 40% Copy accuracy 40% Artefacts 20% Total Weighted mean Ranking Table 7 presents the results of the category copy quality assessment according to visual evaluation. Considering the quality factor colour match between original and copy competitor A performs best, whereas competitor B shows some obvious limitations. Although competitor B performs best regarding the quality factor copy accuracy all three competitors achieve a reasonable result. Table 8: Final evaluation results including the three categories and their corresponding weights from competitor A, B and C. Final ranking: part 1, part 2 and part 3 Weight A B C Weighted mean part 1 50% Weighted mean part 2 30% Weighted mean part 3 20% Total Weighted mean Final ranking To obtain the final ranking the results of all three categories has been weighted according to the priorities of the quality requirements defined by NGAS. Although Table 8 shows an insignificant difference between competitor A and competitor B in terms of weighted mean, competitor B has scored best considering the final ranking. 152

164 Paper A 4 Discussions Although the results of our evaluation presented in the previous section are interesting, and the methods are valid themselves, it is appropriate to proceed to a critical discussion. Even if all three competitors showed an acceptable result in the category general print quality assessment competitor B demonstrated the overall best performance in this category. Considering the quality factor text quality all competitors have shown an acceptable performance in terms of legibility for the font size of 3 pica. However, considering the quality factor colour gamut Figure 5 shows that competitor A represents the largest range of colours achievable on this substrate whereas competitor C has a significantly smaller colour gamut. In terms of matching the Pantone colours of the national colour logo the results from competitor A show a strong correlation between the results of the quality factor «colour match» in the category visual logo assessments and the quality factor colour gamut in the category general print quality assessments. Colour match of the logo colours E*ab Pantone 280M Pantone 185M Pantone 116M Competitor C Competitor B Competitor A Figure 4: Colour difference in terms of mean error between the Pantone colours and the test print. Considering the colorimetric measurements of the test print comparing with the target colours Pantone 280M, Pantone 185M and Pantone 116M, non of the three competitors are able to perform an absolute colour match, as seen in Figure 4. However, for the target colour Pantone 280M competitor A shows a colour difference of ΔE* ab < 5 units whereas competitor B and competitor C show a colour difference of ΔE* ab > 10. Regarding the target colour Pantone 185M it can be seen that all three competitors have their limitations in terms of colour match accuracy. 153

165 Paper A Figure 5: Colour gamut projection of the competitors A, B and C onto the a*b*-plane according to the level of lightness (L*) of the three colours of the national logo. Figure 5 shows the colour gamut projection of the three competitors onto the a*b*-plane according to the level of lightness (L*) of the three optimized logo colours (Table 3). Furthermore it illustrates the distance from the nearest possible point in the colour space achieved from the print medium and the target colour. Except for Pantone 116M, competitor A shows the best performance in terms of colour difference between the Pantone colours and the test print. The Pantone 185M was in general the most difficult colour to reproduce. Table 9: Size of colour gamuts, quantified as the volume of the convex hull of the gamut in CIELAB colour space derived to the test prints of the three competitors A, B and C. Competitor: CIELAB volume Relative volume A unit 100% B unit 77% C unit 57% Another way of looking at the size of colour gamuts is by considering the gamut volume. The size of colour gamuts, quantified as the volume of the convex hull of the gamut in CIELAB colour space was derived from the test prints of the three competitors A, B and C. It can be seen in Table 9 that competitor A provides the largest gamut volume, whereas the volume given by competitor C is only 57% of that given by competitor A. 154

166 Paper A E*ab 6.00 Repeatability Competitor A Competitor B Competitor C Time Figure 6: Repeatability variations over time in terms of ΔE* ab of the three competitors A, B and C. To quantify the effect of the printer s repeatability, a digital test chart including 288 colour patches, which representing the achievable colour gamut on a certain substrate, was printed ten times at short-time interval. Figure 6 shows the repeatability variations according to ten measurements in sequence. Apparently competitor B performs best compare to the other two competitors with very small variations. Competitor C presents a larger variations but the result is still within an acceptable tolerance. However it is worth pointing out that competitor A has shown an unexpectedly poor repeatability performance. Currently the exact cause of this could not be determined, although the most likely cause is human error of the operator. Finally, in Table 7, the results of the category copy quality assessments indicate the best characteristics for competitor A in terms of colour match between original print and copy. Nevertheless all three candidates have performed acceptably in terms of the quality factors copy accuracy and artefacts». 155

167 Paper A E*ab Colour difference between original and copy Mean Standard deviation Max Competitor A Competitor B Competitor C Figure 7: Colour difference between original and copy in terms of ΔE* ab of the three competitors A, B and C. Although the final ranking list shows an insignificant difference between candidate A and candidate B, competitor B has demonstrated the best performance. It is obvious that competitor A would have achieved the overall best result by performing an acceptable repeatability performance. As seen in Table 7, competitor A has performed best regarding the visual evaluation of the copy test. However the results of the colorimetric measurements between the original and the copy do not match the visual evaluation in terms of colour difference. Figure 7 illustrates that competitor A shows the highest value according to the ΔE* ab colour difference (Mean ΔE* ab 11.9 units) whereas competitor C shows the best performance. The fact that there is no strong correlation between the visual evaluation and the colorimetric measurement might be due to the different number of colours, which have been used for the two methods of quality assessment. For the visual evaluation, the three logo colours only have been considered whereas for the colorimetric measurement a large number of colours which represent the entire reproducible colour space have been the target. Suppose the weight of the chosen categories and defined quality factors had been changed, the final result may have been different. Moreover it is worth pointing out that we have proposed further quality factors, which could have been applied to this study to evaluate the print quality (e.g. visual image quality, colorimetric reproduction, uniformity and addressability). However, commercial and political circumstances of NGAS have affected the consideration of the applied factors in this study. 156

168 Paper A 6 Conclusions and perspectives The evaluation method including the selection of the quality factors used in this particular project has been unique in terms of print quality. Although various factors have been chosen and applied in the evaluation process the main target to assess the competitors has been the accuracy of the colour reproduction of the three colours of the national logo. The outcomes in this study demonstrate a strong correlation between the performances of the competitors in terms of the obtained results of the various quality factors. The competitors, that have shown good results considering the size of the colour gamut and the accuracy of the colour match, have obtained reasonably good results in other quality factors too. Except for one competitor, which has shown a very poor performance regarding repeatability. However, the result of the printer evaluation can be used by the NGAS to determine the final purchase decision in terms of image quality. A similar approach might be adopted for other printer evaluation to justify the appropriate equipment to match the desired print quality and requirements. The proposed categories and quality factors are adjustable according to the quality requirements. After the NGAS has decided which type of equipment shall be purchased and implemented further measurement can be conducted on this new machine to confirm the results from the prior tests. Nevertheless, the similarity between the purchased printers themselves in terms of colour accuracy and stability over time would be a possibility for further studies. There might be several potential directions for further work on this topic. The acceptability threshold considering print quality assessment is defined as a vague concept and one that depends strongly on application and industry. Moreover, a further quality metric, which could be considered more suitable for the printer market is a tolerance threshold. Acknowledgments The authors thank NGAS for their kind permission to use this material. Furthermore the authors wish to extend their thanks to the participating members of the NGAS for their grateful collaboration. References Farup I., Hardeberg J. Y., Bakke A. M., Kopperud S. and Rindal A. (2002). Visualization and interactive manipulation of color gamuts. In Proceedings of IS&T and SID s 10th 157

169 Paper A Color Imaging Conference: Color Science and Engineering: Systems, Technologies, Applications, pages , Scottsdale, Arizona. Field G. (1999). Color and Its Reproduction. Second Edition, GATFPress, Pitttsburgh US Gescheider G. A. (1985). Psychophysics, Method, Theory, and Application, Second Edition, Lawrence Erlbaum Associates, ISBN , 152. Handley J. (2001). Comparative Analysis of Bradley-Terry and Thurstone-Mosteller Paired Comparison Models for Image Quality Assessment, PICS 2001, Hardeberg J. Y. and Skarsbø S. E. (2002). Comparing color images quality of four digital presses. Proceedings of the 11th International Printing and Graphic Arts Conference, Bordeaux, France. Ifra (2002). Colour quality audit in reproduction and print, Ifra Consulting Module ISO (1996). Graphic technology - Process control for the manufacture of half-tone colour separations, proof and production prints Part 1: Parameters and measurement methods, First edition ISO (1996). Graphic technology - Process control for the manufacture of half-tone colour separations, proof and production prints Part 2: Offset lithographic processes, First edition ISO 3664 (2000). Viewing conditions Graphic technology and photography Second edition Kulturdepartementet (2000). Grafisk designprogram for departementene, Håndbok, annen utgave, Statens forvaltningstjeneste 158

170 Paper A Jacobson R. E. (1995). An Evaluation of Image Quality Metrics, The Journal of Photographic Science, Vol Marcu G. (2000). Color Quality in Desktop Printing. Tutorial Notes, IS&T s PICS Conference, Portland, Oregon. Microsoft Corporation (2001). Windows Color Quality Specifications for Printer OEMs. Part of Microsoft Hardware Quality Labs (WHQL) s Windows Color Quality Test Kit. Morovic J. and Nussbaum P. (2003). Factors Affecting The Appearance Of Print On Opaque and Transparent Substrates, Journal of Imaging Science and Technology, JIST 47 #6 Stokes, M. (1998). The impact of color management terminology on image quality. In Proc. IS&T s 1998 PICS Conference, pages , Portland, Oregon. 159

171 Paper A 160

172 Paper B Paper B Peter Nussbaum and Jon Yngve Hardeberg Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset Published: In Advances in Printing and Media Technology: Proceedings of the 33rd International Research Conference of iarigai, Volume 33, pp , N. Enlund and M. Lovreček, Ed., Acta Graphica Publishers,

173 Paper B 162

174 Paper B Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset Peter Nussbaum and Jon Y. Hardeberg The Norwegian Color Research Laboratory, Faculty of Computer Science and Media Technology, Gjøvik University College, P.O.Box 191, N-2802 Gjøvik, Norway [email protected], [email protected] Keywords: Process control, colour measurement, print quality assessment, psychophysical methods, colour management Abstract This paper aims to investigate print quality in heat-set web offset by applying colour management. In particular it looks at the colorimetric properties of five heat-set web offset presses in order to evaluate the appropriate colour separation approach, either by applying individual separation profiles or by using industry standard profile such as ISOwebcoated.icc. The key method underlying the work described here relies on obtaining colour measurements to determine the repeatability of each participant in terms of colour differences. Furthermore the variation between the five heat-set web offset printing processes and the variation according to the colorimetric values of the ICC profile ISOwebcoated are important parts of the quantitative evaluation. According to the colour measurements two custom ICC profiles were generated and applied to four test images, which were printed of the five heat-set web offset presses in a second test run. Furthermore the industry standard ICC profile ISOwebcoated was applied too. A psychophysical experiment was carried out to determine naturalness of the reproductions made according to the three profiles applied. Finally the results of the study indicate the performance of the appropriate profile applying to the five heat-set web offset presses to obtain significant best print quality. 163

175 Paper B 1. Introduction Although process control for the production of half-tone colour separations, proof and production prints are clearly defined in ISO :2004 often printing processes show major variations, which affect the appearance of print. Essentially, there are two different approaches considering printing press convention, namely optimized or standardized press behaviour. A fully optimized press is all about maximizing its capability in terms of lowest possible dot gain, highest ink densities and best contrast the individual printing press can achieve without regards to any external specifications or standards. Such individual parameters can create unique press condition, which requires custom ICC profiles to create the appropriate separations. Another approach is to make the press conform to a certain reference or standard such as ISO :2004. By using the second approach and by standardising the behaviour of the press industry standard ICC profiles can be used. Recently, five of the largest Norwegian heat-set web offset printing plants started a collaboration to evaluate their common print quality and print control to strengthen their position in the heat-set market. Due to the print-on-demand concept, parts of a total publication edition can be printed in different print locations. However, the appearance of the total print edition must be identical. Therefore the variations of the five heat-set web offset printing plants had to be verified to determine the appropriate adjustments. The project is named webicc and the five participants are AllerTrykk, Hjemmet Mortensen Trykkeri (HMT), Aktietrykkeriet, Norprint and Ålgård. The aim of the presented work is to evaluate five heat-set web offset printing presses in terms of their conformance to specified values, in accordance with the requirements of ISO :2004. Furthermore, the assessment of each individual printing press and the variation within the five participants are important parts of this study, in order to evaluate the appropriate colour separation approach, either by applying individual separation profiles or by using industry standard profile such as ISOwebcoated.icc. In order to obtain the defined goals, the webicc-group in collaboration with the Norwegian Color Research Laboratory has carried out this print quality project. The next section of this paper will introduce the experimental method used for evaluating the print quality. Then the data analysis performed on the experimentally obtained data will be described, followed by a presentation of the results. Finally the implications of the results will be discussed and ideas will be suggested for future work. 164

176 Paper B 2. Research Method Various studies and research have been done in the field of print quality and repeatedly it has been concluded as a very complex issue. The subject has been discussed at various conferences (see e.g. Hardeberg and Skarsbø, 2002; Nussbaum et al., 2004). Principally there are two different approaches to assess image quality. The first approach is by measurement, using instruments to determine values for the various quality factors. The second method is based on observation, using psychophysical experiments to gather the judgment of human observers. For instance, the pair comparison method is a robust approach where observers are asked to compare the perceptual magnitude between two stimuli (or pairs of stimuli). This method is based on Thurstone s «law of comparative judgment» (Handley, 2001; Gescheider, 1985). Considering the print quality evaluation in this study both methods have been applied, using quantitative analyses based on colour measurements and psychophysical experiments (Figure 1). Figure 1 Overview of method 2.1 Quantitative evaluation The purpose of the following test is to determine the print run repeatability of each participant of the webicc project. Furthermore the variation between the five heat-set web offset printing processes and the variation according to the colorimetric values of the ICC profile ISOwebcoated are important parts of the quantitative evaluation. Hence, the characterization test target ECI2002 was used in this work. As seen in Figure 1 two test print runs with four months interval were carried out in this project. The colour 165

177 Paper B measurement data from the first test print run were the source to generate two custom printer profiles, which were applied in the second test print run. Considering the substrate, a paper manufacture supplied all the participants with the appropriate paper, type 3, 70g/m 2. The ink set colours used in the test was supplied by two manufacturers according to ISO :2004. The colour measuring conditions were according to ISO 13655:2000, geometry 45/0, 2ºobserver, CIELAB system. For documentation all measuring were made on white backing. Considering colour difference calculations ΔE* ab values were computed between individual measurements and the ISOwebcoated colorimetric values. The arithmetic mean, standard deviation and maximum of the resulting colour difference distributions were then computed. The quantitative evaluation has been achieved on the basis of data gathered in an experiment involving colour measurement. The Short-term repeatability performance of the measuring instrument (GretagMacbeth Spectrolino) has a mean ΔE* ab of 0,02 (10 measurements at 10 sec. interval on white). Hence the instrument can be considered to have a high degree of repeatability. The mean repeatability error of the reference measuring set-up was ΔE* ab of 0,17 (equal test print measured two times). 2.2 Psychophysical Experiment The aim of the psychophysical experiment was to determine naturalness of the reproductions made according to different web coated prints. As a result of the colour measurements of the first test run two ICC profiles were generated, namely, webicc and custom. Consequently the psychophysical experiment was carried out according to the test prints of the second test print run. In addition to the two custom profiles the ISOwebcoated ICC profile was applied to each test image and printed in the five heat-set web offset printing plants. The properties of the three applied ICC printer profiles are: ISOwebcoated: Webicc: Custom: Based on the offset characterization table FOGRA28L.txt valid for the following reference printing conditions relating to the international standard ISO/DIS :2003 ICC profile according to the average measurement values of all five heat-set web offset printing plants (Profiling tool: GretagMacbeth, ProfileMaker 5.0.4, predefined separation settings Offset ) ICC profile (5 individual profiles) according to individual measurement values of each heat-set web offset printing plant 166

178 Paper B (Profiling tool: GretagMacbeth, ProfileMaker 5.0.4, predefined separation settings Offset ) The source profile of the test images was AdobeRGB(1998). Considering the conversion options the Adobe (ACE) CMM and the relative colorimetric rendering intent has been used. Four test images were chosen to contain a range of different types of pictorial content and tonal and chromatic variety as shown in Figure 2. The chosen images represent typical the type of images used in weekly magazines. Figure 2 Four test images: camera, couple, portrait and house. The images camera and portrait are reproduced with the permission of Ole Jakob Bøe Skattum. The image couple is reproduced with the permission of Se og Hør and the image house with the permission of Hjemmet Mortesen AS Hytteliv. The images were simultaneously viewed side by side, as shown in Figure 3 and no anchor stimuli were used as a reference, The viewing set up was based on the standard condition of the graphic art industry ISO, 3664, vertical geometry 45, background grey, light source D50 simulator, light intensity 100%. The images were viewed in a darkened room (lights off) and the images, approximately 10cm x 15cm in size, were viewed from a distance of approximately 65 cm. Figure 3 Viewing arrangement for psychophysical experiment 167

179 Paper B A total of 27 observers (16 experts and 11 naïve) with normal colour vision and ages from 20 to 64 took part in the experiment. Each observer was required to make 3 pair-wise comparisons per image (4) and per web offset process (5), a total of 60 comparisons. Note that the comparisons were carried within each individual web offset press. The law of comparative judgement was applied and the method used was pair comparison judgement (Engeldrum, 2000). The observer s task was to decide which of the two prints in the viewing cabinet was most preferred in terms of naturalness. In a study by Morovic (2002) naturalness is defined as an attribute, which is determined by identifying the locations of prototypical memory colours such as skin, grass and sky in colour space (which represent a wider range of naturally occurring colours) and then assessing image colours in relation to these. 3. Results As mentioned previously the aim of this work is to evaluate five heat-set web offset printing presses in terms of their conformance to specified values, in accordance with the requirements of ISO :2004. Furthermore, the assessment of each individual printing press and the variation within the five participants are important tasks to evaluate the appropriate colour separation approach, either by applying individual separation profiles or by using an industry standard profile such as ISOwebcoated.icc. 3.1 Results of the quantitative evaluation The Color Research Laboratory received test prints from all five webicc participants and measured a total of six (3 recto printing and 3 verso printing) test charts (ECI2002) per test print run from each participant to evaluate their colorimetric properties. To determine the variation in terms of colour stability over time the quality factor repeatability is required (Morovic and Nussbaum, 2003). Figure 4 shows the performance of the short-term repeatability of each participant. According to the first test print run the colour differences were calculated between the average of the six measurements and each individual measurement. Apparently, participant Norprint shows the best performance with a mean ΔE* ab 1.2 units. Although the other four participants present larger variations the results are still within an acceptable tolerance (Mean ΔE* ab between 1.5 and 2.0 units). 168

180 Paper B Eab (mean) AllerTrykk HMT Aktiettrykkeriet. Ålgård Norpint Recto 1 Recto 10 Recto 21 Verso 1 Verso 10 Verso 21 Average Average Average Average Average Average print number Figure 4 Short-term repeatability performance of each participant To determine the long-term repeatability performance from each participant colour measurement data from the first test print run and colour measurement data from the second test print run has been compared. As mentioned before the two test print runs were carried with four months interval. As expected the results in Table 1 show higher variations in terms of colour differences. However, as already seen in the short-term repeatability participant Norprint shows again the best performance in the long-term repeatability with a mean ΔE* ab 1.92 units. On the other hand participant HMT presents the largest variation with a mean ΔE* ab 6.48 units. Table 1 Long-term repeatability performance of each participant AllerTrykk HMT Aktietrykkeriet Norprint Ålgård Mean STDEV MAX It is important to point out that the first test print run has been carried out according to individual in-house density standards defined by each participant in the webicc project. These density values are unknown and not available to publish. For the second test print run the webicc group has defined a webicc density standard, which is required to obtain by each participant. The webicc densities of dry solids were measured without polarizing filter. The values are for black 1.4, cyan 1.25, magenta 1.2 and yellow 1.1. Another approach to assess the colorimetric properties is to analyse the variations between the five heat-set web offset printing processes in the first test print run. Figure 5 presents the colour difference between the average measurements of all five heat-set web offset 169

181 Paper B plants and each individual participant. Notice, that participant HMT shows the largest difference (mean ΔE* ab > 4 units) and participant Norprint illustrates the smallest difference Eab Mean STDEV MAX AllerTrykk HMT Aktietrykkeriet Norprint Ålgård Figure 5 Variations between the five participants An important part of the evaluation is to analyse the variation between all five webicc participants and the colorimetric values of the ISOwebcoated ICC profile. According to the results given in Table 2, the participant AllerTrykk shows the best match (mean ΔE* ab < 5 units) and participant HMT presents the largest difference (mean ΔE* ab > 8 units). Note that all the measurements from the first test run were carried out according to individual in-house density standards. Although paper type 3 was used in the project minor variations in terms of colour differences (Table 3) according to ISO :2004 and ISOwebcoated were detected which also may contribute to the overall colour differences. Table 2 Colorimetric colour differences between each heat-set web offset plants and ISOwebcoated values ISOwebcoated ISOwebcoated ISOwebcoated ISOwebcoated ISOwebcoated AllerTrykk HMT Aktietrykkeriet Norprint Ålgård Mean STDEV MAX Table 3 CIELAB coordinates for paper type 3 L* a* b* ISOwebcoated webicc

182 Paper B ISO : Tolerance ±3 ±2 ±2 Another way of looking at the colour differences between colour measurement data of each participant and the data of the ISOwebcoated profile is to consider the colour patches with the highest fluctuations. Figure 6 shows the colorimetric colour differences between each heat-set web offset plant and the ISOwebcoated values including the worst 10% colour differences of the colour patches which are outlined in yellow. Figure 6 The worst 10% colour differences outlined in yellow between each heat-set web offset plant and ISOwebcoated Looking at the CIELAB values for the primary colours cyan, magenta, yellow and black specified in ISO :2004 the differences between the actual values and the nominal values must not exceed the tolerances shown in Table 5. Table 4 presents the CIELAB values for the primary colours from all five webicc participants, ISOwebcoated and the ISO :2004 specification. Moreover the Table shows colour differences calculated between the participants, ISOwebcoated and the ISO :2004 from the first test print run (1.TP) and the second test print run (2.TP). As expected ISOwebcoated is within the tolerance values ΔE ab < 5 units for all four primary colours. For the primary colours cyan and black all five webicc participants show reasonable results in terms of tolerances. Only participant HMT shows colour differences above the CIELAB tolerances. On the other hand the performance of all five participants is rather 171

183 Paper B poor for the primary colours magenta and yellow comparing to the ISOwebcoated CIELAB values. Note that the target value b* for the primary colour yellow in ISO is defined by b*=94. Apparently, none of the participants is able to obtain an appropriate match. Consequently the colour differences in yellow are rather high for all participants. This has also been confirmed, in Figure 6 where the worst 10% colour differences of the colour patches are outlined in yellow, which are typically in the area of yellow, magenta and red. Table 4 CIELAB coordinates for paper type 3 according to measurements on white backing (ISO :2004) for 1. Test Print (1.TP) and 2. Test Print (2.TP) Cyan Magenta Yellow Black 1.TP 2.TP 1.TP 2.TP 1.TP 2.TP 1.TP 2.TP House density webicc density House density webicc density House dens webicc density House density webicc density AllerTrykk L* a* b* ΔE* ab HMT L* a* b* ΔE* ab Aktietrykk eriet L* a* b* ΔE* ab Norprint L* a* b* ΔE* ab Ålgård L* a* b* ΔE* ab ISO webcoated L* a* b* ΔE* ab ISO L* a* b* < CIELAB tolerances > CIELAB tolerances 172

184 Paper B Table 5 CIELAB tolerances (ΔE* ab ) for the solid tones of primaries according to ISO :2004 Cyan Magenta Yellow Black Deviation Figure 7 shows the colour gamut projection of the five heat-set web offset presses and the ICC profile ISOwebcoated onto the a*b*-plane according to the level of lightness L*=50. The data used in this task are based on colorimetric measurements, which have been analysed and visualised by the icc3d application (Farup et al., 2002). The Figure to the left (Figure 7A) illustrate the gamut projection of the five heat-set web offset presses according to the first test print and the Figure to the right (Figure 7B) presents the gamut projection of the second test print with webicc density. It can be seen that the ISOwebcoated profile has a slightly larger colour gamut than the five participants of the webicc project. ISOwebcoated AllerTrykk HMT Aktietrykkeriet Norprint Ålgård Figure 7 2D colour gamuts comparison of the five heat-set web offset presses and the ICC profile ISOwebcoated onto the a*b*-plane according to the level of lightness L*=50 A) First test print run according to in-house densities B) Second test print run according to webicc density Note that the size of the colour gamut of the five participants has changed from the first test print to the second test print. Remember that the density values from the first test print run are not identical to the density of the second test print. Furthermore variations in terms of long-term repeatability affect the colour gamut too. However, as seen in Figure 7B the improvement of the webicc density values results in a closer match to the 173

185 Paper B ISOwebcoated colour gamut. Particularly participant HMT shows in the first test print (Figure 7A) a much larger gamut in the green blue area. Further the largest colour differences between ISOwebcoated and the wecicc participants observed previously in the primary colours magenta and yellow (Table 4) can also be seen in the colour gamut comparison in Figure 7. The three-dimensional CIELAB plot in Figure 8 reveal the colour gamut properties of the ISOwebcoated profile and the webicc profile which contains the average measurement values of all five heat-set web offset printing plants. The plot shows obvious limitations of the webicc profile in high-saturated colours such as yellow and magenta. While the ISOwebcoated profile has some minor limitation in the blue area comparing to the webicc profile. Figure 8 3-D comparison of the ISOwebcoated colour gamut (wireframe red) and the webicc colour gamut (wireframe green) Another way of looking at the size of colour gamuts is by considering the gamut volume. The size of colour gamuts, quantified as the volume of the convex hull (Morovic, 2003) of the gamut in CIELAB colour space was derived from the test prints of both test print runs of the five heat-set web offset presses and the ICC profile ISOwebcoated. Table 6 Comparison of the approximate relative gamut volumes of the five heat-set web offset presses and the ICC profile ISOwebcoated Participants: 1. Test Print House density 2. Test Print Webicc density CIELAB volume Relative volume CIELAB volume Relative volume ISOwebcoated unit 100% unit 100% 174

186 Paper B AllerTrykk unit 99% unit 108% HMT unit 111% unit 101% Aktietrykkeriet unit 97% unit 97% Norprint unit 95% unit 96% Ålgård unit 98% unit 87% The density changes from the first test print run ( house-density ) to the second ( webicc density) affect the gamut volume. It can be seen in Table 6 that AllerTrykk provides the largest gamut volume (even 8% larger than the ISOwebcoated ) in the second test print run, whereas the volume given by participant Ålgård is 87% of that given by ISOwebcoated. Although participant HMT has shown the largest mean colour difference comparing to ISOwebcoated as seen in Table 2, the relative gamut volume is virtually equal to that of the ISOwebcoated. This reinforces the fact that colour differences say something more specific about individual colours whereas relative gamut volumes refer to their ranges. 3.2 Results of the psychophysical experiment The following are the results in terms of z-scores, which have been obtained in the psychophysical experiment. For each image and heat-set web offset printing plant the 3 x 3 matrices of comparisons results for each observer were arranged over the 27 observers and transformed into z-scores. The precisions of the experimental results are described in terms of 95% confidence interval (CI), which is calculated using equation (1) using the mean (R), standard deviation (σ ) and the number of observations (N). Using case V of the method proposed by Thurstone, the standard deviation of the z-scores is assumed to be σ =1/ 2. 95% confidence interval = R ±1.96 σ N (1) 175

187 Paper B z-scores ISOWEBCOATED WEBICC CUSTOM AllerTrykk HMT Aktietrykkeriet Algard Norprint Separations profiles Figure 9 Pair comparison z-scores for each heat-set web press and separation (The error bars represent 95% of population distribution). Figure 9 presents the results in terms of z-scores for each heat-set web press. For the number of images and observations, CI was calculated to be ±0.13. Except for the printing plant AllerTrykk the profile ISOwebcoated shows the best performance for the other four printing plants. The profile custom was ranked significantly best and the profile webicc significantly worst for AllerTrykk. Note that the profiles ISOwebcoated and webicc do not indicate a significant difference in terms of preferred naturalness. It is interesting to note that despite a large colour difference in the quantitative evaluation (Table 2) between the colorimetric values of the participant HMT and the profile ISOwebcoated the performance with the ISOwebcoated profile ranked significantly best. On the other hand considering the long-term variation between the first test print run and the second test print run the participant HMT has shown the largest colour difference. Hence the custom profile for HMT performs worst. 176

188 Paper B z-scores ISOWEBCOATED WEBICC CUSTOM Camera Couple House Portrait Separations profiles Figure 10 Pair comparison z-scores for each type of image and separation (The error bars represent 95% of population distribution). Considering Figure 10 the experimental results identify the profile ISOwebcoated as the separation, which scored significant best for the images Camera, House and Portrait. For the image Couple the profile custom performed best in terms of preferred naturalness. Considering the images Camera, House and portrait the profile custom shows a lower performance comparing to the profile webicc. However there is no significant performance difference between the profile webicc and the profile custom. Table 7 Ranking of separation profiles for each image (1=best, 3=worst) Image ISOWEBCOATED WEBICC CUSTOM Camera Couple House Girl Overall Table 7 gives the ranking of the performance of the three separation profiles for the four images. As it can be seen the profile ISOwebcoated was significant best and the profile webicc was second and the profile custom was third respectively. Figure 11 presents the results in terms of z-scores for all four images and all five webicc-project participants. For the total number of images, heat-set web offset presses and observations, CI was calculated to be ±0.06. Notice, overall, the ISOwebcoated profile performed significant 177

189 Paper B best. Although the results don t show a significant difference between the custom and the webicc profile the custom profile performed second best. 0.5 ISOWEBCOATED WEBICC CUSTOM z-scores Separations profiles Figure 11 Overall z-scores for the three separation profiles 4. Discussion Considering the colorimetric results and the colour differences between the individual printing plants and the colorimetric data of the ISOwebcoated profile (Table 3) it might be expected that a custom profile will perform best. However, it is worth pointing out that the characterization data for generating the ISOwebcoated profile were obtained by measuring a large number of printed samples including averaging, rounding and further adjustments to smoothen the data. Although the results of the repeatability of the individual presses have shown reasonable performances small deviations in the print process can cause unwanted variations in the appearance of the print by using custom profiles. As seen in Figure 9 the webicc profile performs similar comparing to the custom profile. Due to the fact that the webicc profile is related to a larger number of measurements comparing to the custom profile the behaviour is similar to the ISOwebcoated profile considering the smoothness, which tolerates larger variation than the custom profile. Furthermore it must be mentioned that the profile webicc and the profile custom were generated using GretagMacbeth ProfileMaker whereas the profile ISOwebcoated has been generated using Heidelberg Printopen Although the predefined separation settings Offset in ProfileMaker has been chosen, which 178

190 Paper B is, according to GretagMacbeth, compliant with the specifications laid down by FOGRA and ISO for glossy and matt papers the separations using by ISOwebcoated gives a results with a much stronger GCR (grey component replacement) degree. The test images camera and portrait contain large areas with neutral colours, which might be easier exposed to print variations and finally affect the appearance of the print. Hence the reproductions made with the ISOwebcoated profile gives better print results in terms of preserving grey balance. Note, based on the psychophysical experiment the results shown in Figure 10 demonstrate a significant best performance for the test images camera and portrait reproduced by the ISOwebcoated profile. It is worth pointing out that in the beginning of the project a generic Norwegian heat-set web offset profile has been considered for application in the Norwegian market. However, the Norwegian Color Research Laboratory argued very soon to make the presses conform to a certain reference or standard such as ISO As a result of the present study it is possible for all participants of the webicc -group to adopt the industry standard profile ISOwebcoated. Nevertheless to preserve the daily printing conditions and to match the colorimetric requirements of the adopted standard profile it is highly recommended press control according to the standard ISO Conclusions and Perspectives As can be seen from the results of the psychophysical experiment the ICC profile ISOwebcoated performs significant best than the other two profiles, webicc and custom respectively. The poor performance of the custom and webicc profile is not only due to the low degree of GCR. Rather a much larger number of printed samples including averaging, rounding and further adjustments are required to smoothen the data. Although some of the participants of the project still show significant colour variations between their individual press behaviour and the colorimetric values of the ISO :2004 the use of the ISOwebcoated profile perform reasonable results. However, the outcomes of the quantitative evaluation demonstrate clearly that there is still potential to improve the target values of the ISO :2004 to obtain a better coherence between the webicc participants. Furthermore the outcomes in this study demonstrate an obvious need for standardising the behaviour of the heat-set web offset presses. Finally, it is of interest to consider other potential directions for further work in the field of print quality assessment. Colour difference metrics for image quality assessment has been 179

191 Paper B used widely for various applications. However there is not very often a strong correlation between the objective evaluation and the visual assessment. Furthermore the interpretation of the complete image quality assessment considering the colour difference calculation is dependent on the application and the acceptance. The acceptability threshold considering print quality assessment is defined as a vague concept and one that depends strongly on application and industry. Moreover, a further quality metric, which could be considered more suitable for the heat-set web offset printing process, is a tolerance threshold. Acknowledgements The authors thank webicc for their kind permission to use this material. Furthermore the authors wish to extend their thanks to the participating members of the webicc for their grateful collaboration. Literature Engeldrum P.G. (2000) Psychometric Scaling, A Toolkit for Imaging Systems Development, Imcotek Press, Winchester USA, 1 st ed. Farup I., Hardeberg J. Y., Bakke A. M., Kopperud S. and Rindal A. (2002). Visualization and interactive manipulation of color gamuts. In Proceedings of IS&T and SID s 10th Color Imaging Conference: Color Science and Engineering: Systems, Technologies, Applications, pages , Scottsdale, Arizona. Gescheider G. A. (1985) Psychophysics, Method, Theory, and Application, Second Edition, Lawrence Erlbaum Associates, ISBN , 152. Handley J. (2001) Comparative Analysis of Bradley-Terry and Thurstone-Mosteller Paired Comparison Models for Image Quality Assessment, PICS 2001, Hardeberg J. Y. and Skarsbø S. E. (2002) Comparing color images quality of four digital presses. Proceedings of the 11th International Printing and Graphic Arts Conference, Bordeaux, France. ISO (2004) ISO :2004 Graphic Technology Process control for the production of half-tone colour separation, proof and production prints. 180

192 Paper B ISO (2000) ISO 13655:2000 Graphic Technology Spectral measurements and colorimetric computation for graphic arts images. ISO (1999) ISO 3664:1999 Viewing conditions Prints, transparencies and substrates for graphic arts technology and photography. Morovic J. (2003) Gamut mapping. In Digital Color Imaging Handbook. S. Sharma, CRC Press, pp Morovic J. (2002) Colour gamut mapping. In Colour Engineering Achieving Device Independent Colours, P. Green and L. MacDonald, John Wiley & Sons Ltd, pp Morovic J. and Nussbaum P. (2003). Factors Affecting The Appearance Of Print On Opaque and Transparent Substrates, Journal of Imaging Science and Technology, JIST 47 #6. Nussbaum P., Hardeberg J.Y. and Skarsbø S.E. (2004) Print quality evaluation for governmental purchase decisions, In Proc IARIGAI Conference, Copenhagen, Denmark Thurstone L.L. (1927) A law of comparative judgement. Psychol. Rev., 34,

193 Paper B 182

194 Paper C Paper C Aditya Sole, Peter Nussbaum and Jon Yngve Hardeberg Implementing ISO Standards for soft Proofing in a Standardized Printing Workflow according to PSO Published: In Advances in Printing and Media Technology: Proceedings of the 37 th International Research Conference of iarigai, Volume 37, pp , N. Enlund and M. Lovreček, Ed., International Association of Research Organizations for the Information, Media and Graphic Arts Industries,

195 Paper C 184

196 Paper C Implementing ISO12646 Standards for soft Proofing in a Standardized Printing Workflow according to PSO Aditya Sole, Peter Nussbaum, Jon Y. Hardeberg The Norwegian Color Research Laboratory, Faculty of Computer Science and Media Technology, Gjøvik University College, P.O.Box 191, N-2802 Gjøvik, Norway [email protected], [email protected], [email protected] Keywords: Colour measurement, colour management, process control standards, soft proofing, display calibration, display characterisation Abstract This paper defines one of the many ways to setup a soft proofing workstation comprising of a monitor display and viewing booth in a printing workflow as per the Function 4 requirements of PSO certification. Soft proofing requirements defined by ISO are explained and are implemented in this paper. Nec SpectraView LCD2180WG LED display along with Just colorcommunicator 2 viewing booth and X-rite EyeOne Pro spectrophotometer are used in this setup. Display monitor colour gamut is checked for its ability to simulate the ISO standard printer profile (ISOcoated_v2_300_eci.icc) as per the ISO requirements. Methods and procedures to perform ambient light measurements and viewing booth measurements using EyeOne Pro spectrophotometer are explained. Adobe Photoshop CS4 software is used to simulate the printer profile on to the monitor display, while, Nec SpectraView Profiler software is used to calibrate and characterize the display and also to perform ambient light and viewing booth measurements and adjustments. 1. Introduction To reduce expensive and time consuming iterations in standardized printing workflow soft proofing has become an important concept in predicting the final print product. Soft proofing can be defined as the ability to match colour images displayed on colour monitors to the images produced when the same digital file is rendered by proofing and printing systems (ISO ). However, although the concept soft proofing is not 185

197 Paper C new and the task may sound rather simple, in practical applications the colour appearance between two different medias (e.g. softcopy simulation of a hardcopy) can differ a lot due to unsuitable type of devices, incorrect use of parameters and inaccurate device calibration and characterization or inappropriate measurement methods. In the past a number of studies and research work have addressed the issue of soft proofing. For more details see the work of (Gatt et al. 2004) and (Gatt et al. 2005). ISO defines parameters for monitor and viewing booth condition setup for soft proofing environment. The practical methods to implement these standards as per the job requirements are not been clearly defined. This paper therefore is aiming for to describe in details one of the ways to setup an appropriate soft proofing station (comprising of a monitor and a viewing booth with the appropriate ambient lighting conditions) according to ISO standards for soft copy and hard copy proof comparison in the graphic arts industry and evaluate the performance of the entire soft proof set up according to ISO This is also in accordance with soft proofing in a standardized printing workflow according to PSO (UGRA 2009). In this work the current standards and the appropriate parameters for soft proofing are reported before proposing the method of implementing these standards. 2. Methods 2.1 Standards for soft proofing (specifications) ISO specify the soft proofing standards. This standard describes two scenarios. The first scenario consists of comparing a soft copy directly with a hard copy while the second scenario consists of viewing soft copy images on a display independently of any hard copy images. This paper primarily focuses on the first scenario where soft copy is directly compared with the hard copy. The technical requirements defined by ISO for scenario 1 i.e. comparing a soft copy with a hard copy are as follows: Ambient illumination, surroundings and environment The level of ambient illumination should be sufficiently low. The standards recommend for luminance of a perfectly reflecting diffuser placed at position of the faceplate of the monitor, with the monitor switched off, shall not be greater than ¼ of the monitor white point luminance and should not be greater than 1/8 of the monitor white point luminance (ISO 12646, sec 4.7.2) 186

198 Paper C The colour temperature of the ambient light, such as room light, should be within ± 200 k of the colour temperature of the illumination used in the viewing booth. The conditions within the viewing booth shall conform to viewing condition P2 of (ISO )(ISO 12646, sec 4.7.2). P2 viewing conditions are defined as the conditions for practical appraisal of prints, including routine inspection. The luminance of the area surrounding the monitor shall not exceed 1/10 of the luminance of the monitor showing a white screen. Extraneous light, whether from light source or reflected by objects, shall be baffled from view and from illuminating the print or other image being compared (ISO 12646, sec 4.7.2) The surround and backing shall be neutral and matt (ISO 3664, sec 4.3.4). The surround shall have a luminous reflectance between 10% and 60% with the specific value being selected to be consistent with practical viewing. Chromaticity, luminance of the white and black points The black point of the display shall have a luminance that is less than 1% of the maximum luminance of the display (ISO 12646, sec 4.8.1). The conditions within the viewing booth shall conform to viewing condition P2 of ISO 3664 (ISO 12646, sec 4.8.2). P2 viewing conditions specify conditions applicable for the appraisal of tone reproduction of individual images, photographic image inspection or the judgement of prints. The illumination of the plane of viewing shall approximate that of CIE standard illuminant D50. It shall have u 10, v 10 chromaticity co-ordinates within the radius of from u 10 = , v 10 = in the CIE 1976 Uniform Chromaticity Scale (UCS) diagram (calculations using the 10 observer angle). The illuminance at the centre of the viewing surface shall be (500 ± 125 lux). The illumination uniformity should be such that for a viewing area up to 1m 2, the illuminance at any point within the illuminated area shall not be less than 75% of the illumiance measured at the centre of the illuminated viewing surface area. The uniformity should be evaluated by measuring at least 9 points equally distributed on the viewing surface (ISO 3664, sec 4.3.3). The luminance of the white displayed on the monitor shall be at least 80 cd/m 2 but preferably 160 cd/m 2 in order to match an unprinted sheet of white paper located close to monitor having an illuminance of 500 lux, as specified in ISO 3664 for viewing condition P2 (ISO 12646, sec 4.8.2). ISO recommends D50 as white point for the soft proofing display (as the white point in the viewing booth is D50); namely u =0.2092, v =0.4881, as specified in CIE 187

199 Paper C Publication 15 (CIE ). The chromaticity obtained, for the white point chosen shall be within a circle of radius from this point (ISO 12646, sec 4.8.2). Colorimetric accuracy and grey balance of the display Tristimulus values shall be measured for at least 10 neutral colours (R=G=B). For each neutral colour the colour difference, ΔE* c = (Δa 2 +Δb 2 ), where Δa is the difference for the CIELAB (red-green) co-ordinate and Δb is the difference for the CIELAB (yellow-blue) co-ordinate, shall be calculated between these measured CIELAB values and the CIELAB values which are intended to be displayed by the software characterizing the display. The average deviation shall not exceed ΔE* c = 3 and preferably not ΔE c = 2 (ISO 12646, sec 4.10). The average of the ΔE* ab between the measured CIELAB values and the CIELAB values intended to be displayed by the *.icc monitor profile shall not exceed ΔE* ab 5 and preferably not ΔE* ab 2 units respectively. Uniformity of display luminance The display should be visually uniform when displaying flat white, grey and black images. All the measured luminance values should be within 5% of the luminance of the centre and shall be within 10% of it (ISO 12646, sec 4.4). Table 1 below summarizes the specification and tolerances defined by ISO standards for soft proofing. Table 1 Specification and tolerances according to ISO standards for soft proofing Parameters Target Colour Intensity Chromaticity Uniformity Black point Temp Ambient 5000 K ¼ monitor Uniformly -NAlight white point diffuse Surround Neutral and matt with luminous reflectance between 10% and 60% Display 5000 K >80 cd/m 2 u = v = % within the centre measurement Luminance < 1% of max luminance Viewing booth 5000 K 500lux ± 125lux u 10 = v 10 = for 1m 2 area, luminance = >75% luminance at the centre -NA- 188

200 Paper C 2.2 Procedure followed for implementing the conditions/ specifications In this section methods and procedures are proposed for implementing the specifications mentioned previously. To setup the soft proofing station and the ambient light condition following apparatus was used: Hardware NEC SpectraView LCD 2180WG LED display Apple MAC PRO 10.4 Just Normlicht ColorCommunicator 2 viewing booth X-Rite EyeOne Pro spectrophotometer including ambient light measurement adaptor. OSRAM L58W/950, LUMLUX de LUXE, Daylight tubes for ambient light Software SpectraView Profiler Version 4.1 (NEC SpectraView monitor calibration and profiling software) UDACT (UGRA Display Analysis Certification Tool) monitor certifying software by UGRA. Adobe Photoshop CS4 Nec SpectraView LCD 2180EG LED display was used as the display monitor to display the soft copy images. Nec LCD 2180WG LED is a high end display which features individual high power red, green and blue LEDs (Light Emitting diodes) as a backlight light source for the LCD, instead of the typical CCFL (Cold Cathode Fluorescent Lamp) (NEC 2005). LED backlight results in a wide output colour gamut of the display. Hardware calibration can be performed on this display using its own calibration and profiling software. Hardware calibration lets the user adjust the brightness and gradation properties of each RGB primary within the monitor compared to the software calibration were the monitor s primaries are measured and the difference between the measured values and the target values is corrected by adjusting the output of the video card. Software calibration method can create problems like greyscale banding, decline in appearance of colours in greyscale images. Hardware calibration eliminates the need to correct the RGB output for smooth, accurate display of greyscale images. 189

201 Paper C Nec SpectraView Profiler software calibrates and profiles the Nec LCD 2180WG LED display monitor (NEC 2008). The software allows custom target calibrations that can be preset or can be defined by the user. The user can define the white point, light intensity, gamma curves, black point, etc. At the end it also profiles the display and generates a *.icc colour profile. Target and measured results are analysed and displayed, showing information like measured colour, colour difference, gamut etc. (NEC 2005). Just colorcommunicator 2 viewing booth was used for the soft proofing station along with the NEC SpectraView monitor display. Just colorcommunicator 2 can be connected to the system by an USB interface. The light intensity of the viewing booth can therefore be adjusted via software to match the light intensity of the display for soft proofing (JUST 2008). NEC SpectraView Profile software has a facility to communicate with the Just colorcommunicator 2 to adjust the light intensity of the viewing booth as per the ISO soft proofing standards. To perform the measurements of light intensity, white point, colour rendering on the monitor display and on the hard proofing a number of different measuring instruments can be used. To perform the measurement of colour on the monitor display either a colorimeter or a spectrophotometer can be used. While to perform ambient light measurements a light meter, colorimeter or a spectrophotometer with ambient light adaptor can be used. To avoid inter-instrument uncertainty and reproducibility issue only one instrument has been used for the measurements (Nussbaum et al. 2009). For this setup, a X-Rite EyeOne Pro spectrophotometer was used to perform the measurements on monitor display and hard copy and also to perform the ambient light measurements using the ambient light adaptor. Monitor Calibration and viewing booth adjustment To calibrate the monitor and adjust the viewing booth following procedure was followed: The viewing booth and the monitor display were turned on half an hour before any measurements/adjustments were made in order to properly warm and stabilize the performance of the devices. The viewing booth was connected to the system via UBS cable connection. X-rite EyeOne Pro was connected to the system and the emission measurement mode chosen. Consequently the LCD display profiling test chart (with 99 colour patches) was measured using the ProfileMaker Pro MeasureTool software to warm up the measurement instrument. 190

202 Paper C NEC SpectraView Profiler Version 4.1 software was used to calibrate, to characterize and to adjust the viewing booth according to ISO standards parameter for soft proofing. This software communicates with the viewing booth via the USB cable connection (NEC 2008). P2 viewing conditions defined in ISO 3664 were setup in the viewing booth according to ISO specifications (ISO 12646, sec 4.8.2). NEC SpectraView Profiler software asks the user to define the parameters white point, light intensity, gamma, etc to perform the calibration according to the defined parameters. It also has a specific calibration targets for which the monitor can be calibrated. It contains a target for soft proofing according to ISO and ISO 3664 standards where the target for calibration parameters is in accordance with the specifications of the ISO standards. Table 2 shows the parameters selected for the calibration and characterization of the monitor display. Table 2 Preset target settings in NEC SpectraView Profiler software Display Type LCD Calibration method Hardware calibration (monitor LUTs) Calibration settings ISO3664 and ISO Profile settings LUT based (accurate) CIE Daylight standard D50 Tonal response Curve L* (recommended) Specify White and black luminance White 160 cd/m 2 Black Minimum neutral Profile type 16 bit LUT based Chromatic adaptation CAT02 After selecting the target the software asks for a monitor profile name and consequently the measurement process were started. According to the defined parameters the software (SpectraView Profiler) projects colour patches with certain RGB values that are measured using the X-rite EyeOne Pro spectrophotometer. Depending on the measurement data appropriate adjustments are made automatically until the difference between the measured parameters and the target parameters is achieved minimum. Consequently, an ICC monitor profile is created at the last stage of the calibration process defining device independent 191

203 Paper C values (CIEXYZ) that correspond to a given set of device dependent numbers and vice versa. After calibrating and creating the monitor profile, the profile was validated in order to check the objective quality of the generated profile in terms of calculating ΔE* ab colour difference between given reference values and the corresponding measured values. After successfully validating, the ΔE* ab colour difference is shown in form of a bar graph and the profile is automatically applied as the system profile. Subsequently, the viewing booth settings were checked. Check viewing booth option was selected in the review column Viewing booth and monitor comparison and measurements were performed. Figure 1 Schematic diagram of the position of the measuring instrument for viewing booth measurement To record the measurements, the measuring instrument was held parallel to the viewing booth facing towards the plane of viewing as shown in Figure 1. Instructions on the screen were followed. Viewing booth conditions were checked against the P2 viewing condition mentioned in ISO In order to reduce the difference between the measured values and the target values Adjust viewing booth tab was selected to make the necessary adjustments. In the Adjust viewing booth tab, for ISO reference values, viewing booth and monitor comparison was selected. It performs the necessary adjustments in the viewing booth (for light intensity, etc) to match the ISO 3664 P2 viewing condition requirements. The viewing booth measurements were performed using X-rite EyeOne Pro with the ambient light adaptor. 192

204 Paper C Ambient light condition setup and adjustment The room used for the soft proof setup has a neutral and matt colour according to ISO 3664 (sec 4.3.4). To setup the ambient lighting conditions according to ISO standards for soft proofing, daylight tubes named OSRAM L58W/950, LUMLUX de LUXE, Daylight were setup in the room. These tubes simulate Illuminant D50 and the light intensity can be varied as per the requirements. Therefore, in order to vary the intensity of these tubes a varying knob facility was introduced in the room with which the ambient light intensity in the room could be controlled. SpectraView Profiler software was used to perform the ambient light measurements. Ambient light option was selected in the review column and measurements were performed as per the instructions given by the software. To perform the measurements, X- rite EyeOne Pro spectrophotometer with the ambient light adaptor was used for measuring the ambient light in the room. The measurement instrument was held straight parallel to the display facing towards the room, with the monitor switched off. Furthermore, Figure 2 illustrates the schematic diagram of the position of the spectrophotometer to perform the ambient light measurements. Figure 2 Schematic diagram of the position of the measuring instrument for ambient light measurement Notice, the geometry of the measurement device is very critical in terms of the measured values. Small changes of the measurement instrument angle or distance to the light source can change the measurement results dramatically. The ambient light was measured for 193

205 Paper C light intensity and colour temperature. Depending upon the measurements obtained the light intensity was adjusted using the varying knob facility to be within the tolerance defined by ISO standards for soft proofing conditions. After varying the knob to change the light intensity, measurements were again conducted with a gap of 30 minutes (time required for the tubes to stabilize after adjusting the light intensity). This procedure was followed till the time appropriate light intensity was achieved to be within the ISO standards for soft proofing. As an alternative a number of software s like BabelColor CT&A, UDACT, etc can be used to perform ambient light measurements. BabelColor CT&A uses its Spectral Tool feature with EyeOne Pro spectrophotometer to measure and evaluate the ambient light conditions. This software, upon measurements, shows a small graph of u, v co-ordinates which tells if the measured colour temperature is within the ISO defined tolerance. BabelColor CT&A also provide values for Colour rendering index (CRI) and Metamerism index (MI) of the measured light for P1 condition of measurements. According to ISO 3664: 2009 CRI and MI are more important parameters to determine the quality of the illumination. 3. Results and discussions 3.1 Monitor display To analyse and evaluate the display for soft proofing environment the UDACT certifying software was used. UDACT enables objective, quality oriented and comparable evaluation for an individual soft proofing display (UGRA 2008). X-rite EyeOne Pro spectrophotometer was connected to the UDACT software. Instructions on the screen were followed. UDACT then shows 102 colour patches on the screen of which measurement values are recorded using the connected spectrophotometer. The measurements include 21 patches to verify the gray balance. To determine the profile quality 35 patches are measured. Finally, UDACT analyses the display for the specifications according to ISO standards for soft proofing measuring the 46 patches of the Ugra/Fogra Media Wedge. UDACT records the measurements and generates a report at the end that shows all the details of evaluation. ISO standards for soft proofing addresses the factors white point and black point, gray balance, colour gamut and uniformity of luminance determining the performance of the monitor display. In the following chapter, the results of the monitor display analysis will be presented. 194

206 Paper C White point and black point According to ISO 12646, white point of the display should be as close as possible to the calibration target. Table 3 below presents the target and the measured values for white and black point. Table 3 White and black point of the NEC display Target Measured Difference White point 5000 K 4991 K 0.5 ΔE ab Luminance 160 cd/m cd/m 2 Black point <1.6 cd/m cd/m 2 The comparisons between the measured values and the target values in terms of white point, black point and luminance show acceptable results. Gray balance The maximum allowed deviation shall not exceed max ΔC* ab 3 unit and preferably not max ΔC* ab 2 unit (ISO 12646, sec 4.10). For further details on calculating ΔC* ab and a general overview of CIE colorimetry see (Hunt 1998)). Table 4 shows the 21 patches for the gray balance measurement and the performances of the monitor display in terms of the corresponding measurement and the ΔC* ab deviation. It can be seen that although, the max ΔC* ab is 2.35 units, it is still within the given tolerance. Flare light could be considered influencing the measurement on dark colours (e.g. 0%, 5%, 10% and 15%). Table 4 Gray balance measurements performed and reported by UDACT % CIELAB (calculated) CIELAB (measured) ΔC* ab L* a* b* L* a* b* 100 (white) (black)

207 Paper C Average 0.80 Max 2.35 Colour gamut The colour gamut of the display should be such that it totally encloses the colour gamut produced by the inks specified in the appropriate part of ISO for which the display is required to provide a proof (ISO 12646, sec A.2). UDACT checks for the colour gamut by measuring the Ugra/Fogra Media Wedge 2.0 on the display simulating the ISOcoated_v2_300_eci.icc profile. The 46 Ugra/Fogra Media Wedge patches are visualized using the system profile and the corresponding measurements are recorded and compared with the reference values. The average colour difference between the reference values and the measurement values is ΔE* ab 0.8 units. The max ΔE* ab 2.8 and related to black (16 0 0) which is not unexpected due to the results of the gray balance task which show the largest colour differences in dark colours. The entire table including the reference values, measurement values and the calculated colour difference are provided in the appendix. For illustration purposes Figure 3 shows the top down projection of the colour gamut of the monitor profile (wireframe) and the ISOcoated_v2_300_eci.icc profile (solid) in the CIELAB colour space. In this presentation it can be seen that the boundary of the ISOcoated_v2_300_eci.icc colour gamut is clearly within the gamut of the monitor profile. 196

208 Paper C Figure 3 Top down projection of the colour gamut of the monitor profile (wireframe) and the ISOcoated_v2_300_eci.icc profile (solid) in the CIELAB colour space On the other hand Figure 4 illustrates the horizontal projection view of the four CIELAB planes a+, a-, b+ and b- (clockwise) of colour gamut of the monitor profile (wireframe) and ISOcoated_v2_300_eci.icc profile (solid). It can be clearly observed that the monitor colour gamut is big enough to simulate the ISO coated printer profile. Figure 4 Horizontal projection view of the four CIELAB planes a+, a-, b+ and b- (clockwise) of colour gamut of the monitor profile (wireframe) and ISOcoated_v2_300_eci.icc profile (solid) 197

209 Paper C Uniformity of luminance Finally the results of the uniformity of the luminance are presented. As previously mentioned the uniformity will be determined displaying white, grey and black images each filled the entire screen. Minimum 9 points of the image area of the screen shall be measured, for each level (Figure 5). According to the ISO standard requirement the white image consist of the maximum value in each channel Red, Green and Blue (255 for 8 bit). Then the grey image should have about half of the maximum value in each channel (127 for 8 bit), and finally the black should consist of approximately a quarter of the maximum value in each channel (which is e.g. 63 for 8 bit) but shall be greater than 10 % of the maximum digital code value (which is 26 for 8 bit). Table 5 shows the uniformity of luminance measurements performed on the display. Figure 5 Measurement positions on the monitor display Table 5 Luminance measurements performed on the display at 9 different positions Luminance measured in cd/m 2 Position 1 Position 2 Position 3 White Grey Dark Grey Position 4 Position 5 Position 6 White Grey Dark Grey

210 Paper C Position 7 Position 8 Position 9 White Grey Dark Grey Figure 6, Figure 7 and Figure 8 show the result of the uniformity check for white, grey and dark grey neutrals according to ISO standards for soft proofing. According to ISO 12646, sec 4.4, the luminance of the display (measured at 9 different locations figure 10) shall be within 10% of the luminance of the measurement made at the centre of the display. Position 5 (on X-axis) in the graphs below is the centre of the display. Figure 6 Uniformity of the display for white neutral Figure 7 Uniformity of the display for grey neutral 199

211 Paper C Figure 8 Uniformity of display for dark grey neutral From the graphs it can be observed that the luminance values varies at 9 different positions on the display. It can be noticed that the luminance at position 7, 8 and 9 (lower part of the monitor display) gives the highest values. However, all the measurements are within the 10% tolerance defined by ISO standards. 3.2 Viewing booth and ambient lighting conditions To analyse and evaluate the viewing booth and ambient lighting conditions, measurements were performed using the X-rite EyeOne Pro spectrophotometer with the ambient light adaptor. The measured value was then compared against the P2 viewing conditions defined in ISO 3664:2009 standards. Table 6 and Table 7 show the target and measured viewing booth and ambient light conditions. Table 6 Viewing booth target and the measured values Illuminant Viewing Illuminance Colour Chromaticity Booth lux Temperature Tolerance (u v ) Target 5000 K ± 125 Measured 4995 K

212 Paper C Table 7 Ambient light condition target and measured values Ambient Illuminant light Colour Illuminance condition Temperature Target 5000 K ¼ monitor white point luminance Measured 4853 K 103 lux 3.3 Soft proofing image data Implementation of the ISO standard alone does not guarantee that a displayed (soft copy) image will match the colour of the same image produced on the hard copy in the viewing booth (ISO 12646). To obtain a colour match, colour transformation is required to convert the colour data format from the printer colour space to the display colour space. This transformation is mainly done by means of a colour management system using the appropriate ICC profiles. ICC profiles define the relationship between the device specific colour space and the profile connection space (PCS) (ISO 12646:Annex A.1). Therefore, to obtain a colour match between hard copy and soft copy proofs using ICC colour management, the soft proof is obtained by applying the image data, the combination of the device to PCS transforms of the print output profile and the PCS to device transforms of the display profile. In this study, Adobe Photoshop CS4 software was used to perform the colour transformation. Visual Print reference (VPR) images were applied to simulate the ISOcoated_v2_300_eci.icc profile. These images were compared to evaluate the visual match between the VPR hard copy in the viewing booth and the soft copy on the display. To apply necessary simulation profiles a VPR image was opened in Photoshop and the menu View > Proof setup > Custom selected. In the Customize Proof Condition window the appropriate simulating profile was selected. In this case, simulating profile was the CMYK output profile that needed to be soft proofed (ISOcoated_v2_300_eci.icc profile). Image background was selected as neutral gray (similar to the viewing booth background). Simulate paper white option was tick marked to simulate the paper white and the images were then compared with the hard copy in the viewing booth. 201

213 Paper C For the soft proofing task Adobe Acrobat Pro software can also be used to display the soft copy image on the monitor display to compare with the corresponding hard copy image in the viewing booth. 4. Conclusion In this paper methods and procedures to setup a soft proofing station are discussed. ISO standards for soft proofing were successfully implemented. The NEC LCD2180WG-LED monitor display used in this paper has a very wide colour gamut and is performing visually uniform. The appropriate calibration and characterization procedure has been applied. Gamut plots in Figure 3 and Figure 4 showed that the monitor display is able to simulate printer profile colours. This was verified further by measuring the UGRA/FOGRA step wedge simulated with ISO coated printer profile on the display using the EyeOne Pro spectrophotometer. The maximum colour difference obtained was ΔE ab = 2.8. Measurement procedures to measure ambient light using an EyeOne Pro spectrophotometer with ambient light adaptor are implemented and discussed in this work. NEC SpectraView Profiler software was used to perform the ambient light measurements. A number of different software s can be used to perform ambient light, viewing booth and display measurements. As an alternative to NEC SpectraView profiler, Babel Color CT&A software can be used to perform these measurements. ISO standards were successfully implemented, as, all the measurements obtained were within the tolerance level defined. UDACT software was used to evaluate the display for soft proofing and printer profile simulation according to these soft proofing standards. ISO define a wide range of tolerance for ambient light and viewing booth light intensity measurements. It was observed that, in spite of being within the ISO standards tolerance level the two images (soft copy image on the display and the corresponding hard copy image in the viewing booth) might not show an exact visual match. Therefore, the ambient light intensity and the viewing booth light intensity can be adjusted (within the tolerances defined by ISO 12646) as per different job requirements to get the closest possible visual match between the soft copy on the display and the corresponding hard copy in the viewing booth. Finally, it is of interest to consider other potential directions for further work in the field soft proofing. Firstly, to verify the monitor display measurement results of dark colours a 202

214 Paper C spectroradiometer could be used. Furthermore, colour measurements of the monitor display and the viewing booth could be conducted with a spectroradiometer to confirm the appropriate set up. Moreover, a psychophysical experiment could evaluate the magnitude of visual differences between the soft proofed image on the monitor display and the hard copy in the viewing booth in terms of perceptibility and acceptability threshold. Factors and their magnitude affecting the appearance on the monitor display and the viewing booth could be investigated. Nevertheless, using more than one measurement instrument in the soft proofing workflow the inter-instrument reproducibility has to be considered. References CIE15 (2004), 'Colorimetry', 3rd ed (CIE Central Bureau, Vienna). Gatt, A, Westland, S, and Bala, R (2004), 'Testing the softproofing paradigm', 12 th Color Imaging Conference, p Gatt, A, et al. (2005), 'Testing the Softproofing Paradigm II', 10 th Congress of the International Colour Association, p Hunt, RWG (1998), Measuring colour (3 ed.: Fountain Press). ISO 3664 (2009), 'Graphic technology and photography Viewing conditions', third ed.: Geneva: ISO [www. iso. org]. ISO12646 (2008), 'Graphic technology Displays for colour proofing Characteristics and viewing conditions', Geneva: ISO [www. iso. org]. JUST (2008), 'Normlicht, Color communicator, Help manual'. NEC (2005) LCD2180WG-LED technical background and feature overview', Display Solution < >, accessed 2. July NEC (2008), 'SpectraView Profiler Version 4.1, User s manual'. 203

215 Paper C Nussbaum, P, Sole, A., and Hardeberg, Jon Y. (2009), 'Consequences of using a number of different color measurement instruments in a color managed printing workflow', TAGA Proceedings. UGRA (2008), 'UDACT Ugra display analysis and certification tool, Help manual'. UGRA (2009), 'PSO certification', < accessed 6. July Appendix Reference Measured ΔE ab L a b L a B ,

216 Paper C Avg 0.8 Measurements performed and recorded by UDACT for colour gamut 205

217 Paper C 206

218 Paper D Paper D Peter Nussbaum and Jon Yngve Hardeberg Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print Color Research & Application, Article first published online: March 8th 2011, DOI: /col Wiley. 207

219 Paper D 208

220 Paper D Print Quality Evaluation and Applied Colour Management in Coldset Offset Newspaper Print Peter Nussbaum and Jon Y. Hardeberg The Norwegian Color Research Laboratory Faculty of Computer Science and Media Technology Gjøvik University College P.O.Box 191 N-2802 Gjøvik Norway [email protected] [email protected] Keywords: Process control, print quality assessment, newsprint, psychophysical methods, colour measurement, colour management Abstract This paper aims to investigate print quality in newspaper print by considering the appropriate calibration standard and applying colour management. In particular, this paper examines the colorimetric properties of eight Norwegian newspaper printing presses, in order to evaluate the relevant colour separation approach, either by applying custom separation profiles or by using an industry standard profile. The key method underlying the work described here relies on obtaining colour measurements to determine the repeatability of each participant in terms of colour differences. Furthermore, the variation between the eight newspaper printing presses and the variation according to the colorimetric values of the ISO standard are important parts of the quantitative evaluation. Based on the colour measurements two custom ICC profiles were generated and an industry standard profile ISOnewspaper26v4.icc was also used. The first custom profile was generated using averaged colour measurement data set from a test print run, and the second using a data set averaged between measured data and the characterization data set IFRA26.txt provided by IFRA. These three profiles were applied to four test images, which were then printed by the eight newspaper printing presses. A psychophysical experiment was carried out to determine the pleasantness of the reproductions, which were produced using the three profiles. The results of the study show the performance of the appropriate profile, 209

221 Paper D which is applied to the eight newspaper printing presses to obtain significant best print quality. Eventually the results demonstrate the fact that the print variations in colours between the eight printing presses are larger than the difference between the custom and the standard profiles. Hence, the print variations and not the profile selection may have determined the visual print quality. Therefore the study reveals the importance of adopting international standards and methods instead of using insufficiently defined house standards to preserve equal results among different newspaper printing presses. 1 Introduction Although process control parameters for the production of half-tone colour separations, proofs and production prints, are clearly defined in ISO , often, newspaper printing processes show significant variations which affect the appearance of print. Essentially, there are two different approaches considering printing press convention, namely optimized or standardized press behaviour. A fully optimized press aims at maximizing its capability in terms of lowest possible dot gain, highest ink densities and best contrast that the individual printing press can achieve, without considerations of any external specifications or standards. Such individual parameters can create unique press conditions, which requires custom ICC profiles for creating the appropriate separations. On the other hand, an optimized press condition can be considered as a press consistency where certain predefined parameters (according to standards) over time (repeatability) and space (uniformity) are preserved. Another approach is to make the press conform to a certain reference or standard such as ISO By using the second approach and by standardising the behaviour of the press, industry standard ICC profiles can be used. In terms of aiming for a common print appearance across printing plants (e.g. preserving print appearance of an ad campaign) the second approach including consistency will be the most suitable one to ensure a predictable and equivalent print result. In 1995, the Norwegian Newspaper Publishers Association (NAL) founded its subsidiary NADA AS. NADA s primary goal has been to establish a standard for digital advertisement delivery in Norway, in order to simplify the process of creating and transmitting digital advertisements from producer to newspaper. Furthermore, NADA has also been responsible for the generation of three custom newspaper ICC profiles in the period of 2000 to NADA highly recommends application of these profiles in the national newspapers printing process. However, these custom profiles have two common 210

222 Paper D characteristics in terms of their parameters. Firstly, the number of colour measurements for generating the profiles is very small, and secondly, the degree of GCR (Grey Component Replacement) is rather low considering newspaper print. Therefore, the performance of the profiles and the corresponding print quality has been considered as not satisfactory and demonstrates the need for further revision. Recently, NADA and eight of the largest Norwegian newspaper printing plants started a project to evaluate their common print quality and print control. The aim of the presented work is to evaluate eight newspaper printing plants in terms of their conformance to specified values in accordance with the requirements of ISO Furthermore, the assessment of each individual printing press and the variation within the 8 participants are important parts of this study to evaluate the appropriate colour separation approach, either by applying custom separation profiles or by using industry standard profiles such as ISOnewspaper26v4.icc. 2 In order to obtain the defined goals, NADA in collaboration with the Norwegian Color Research Laboratory has carried out this print quality project. The following newspaper printing plants participated in the project: Dagblad-Trykk AS, Schibsted Trykk AS, Nr 1 Trykk AS, Fædrelandsvennens Trykkeri AS, Adressa Trykk AS, Bergens Tidende AS, Aftenbladet Trykk AS, and Halden Arbeiderblad AS. After the present Section introducing the work, Section 2 gives an overview of the experimental method including information about the quantitative evaluation and the psychophysical experiment. In Section 3 we present and discuss our results, before concluding in Section 4. 2 Experimental method Various studies and research have been done in the field of print quality and repeatedly it has been concluded as a very complex issue. 3-6 Principally there are two different approaches to assessing print quality. The first approach is by measurement, using instruments to determine values for the various quality factors. The second method is based on observation, using psychophysical experiments to gather the judgement of human observers. For instance, the rank order method is a robust approach where observers are asked to rank the image samples in order, from best to worst, along an attribute defined by the instructions, such as pleasantness. This method is based on Thurstone s «law of comparative judgement». 7,8 211

223 Paper D Considering the print quality evaluation in this study both approaches have been applied, using quantitative analyses based on colour measurements, and using psychophysical experiments (Figure 1). Figure 1: Overview of method. 2.1 Quantitative evaluation The purpose of the test procedure described here is to determine the short and long term repeatability of each participant of the project. Furthermore the variation between the eight newspaper printing processes and the variation according to the colorimetric values of the ISO are important parts of the quantitative evaluation. To compare the current printing conditions the test target ECI2002R CMYK.tif and the IT8.7-3 CMYK Target.tif have been printed and measured. The test target ECI2002R CMYK.tif is corresponding to the current characterization data set, which were used for generating the Norwegian newspaper ICC profile. Although a number of parameters (e. g. ink set colours, dot gain etc.) are required to define the appropriate printing conditions the only data available for the colorimetric comparison between the present press conditions and characterization data set were the colorimetric data as a part of the private tag in the present Norwegian newspaper ICC profile (NADA_avis_vjanu2004). On the other hand the test target IT8.7-3 CMYK Target.tif corresponds to the characterization data set IFRA26.txt. 9 The ISOnewspaper26v4.icc is the IFRA newspaper printing standard profile, which is based on the characterization table "IFRA26.txt and is valid for the following conditions in accordance with the international standard ISO

224 Paper D Furthermore, the colour measurement data measured on the IT8.7-3 CMYK Target from the first Test Print were the starting point for generating a number of ICC profiles with different separation settings. Considering the test print, each printing plant has been asked to use the Norwegian newspaper production method to perform the printing task. 10 In particular, the method requires 0.90 density for the solid primary colours cyan, magenta and yellow and 1.10 density for black. Furthermore, the setup of the Raster Image Processor (RIP) software has to be done according to linearized newspaper production agreement. For more details on Norwegian linearized newspaper production method see the guidelines of NADA 10. In general, a RIP converts information of a page layout such as digital structure in images, text, graphic elements, and positioning commands into a bitmap file. Furthermore, the significant process here is the screening and the generation of the appropriate data for addressing the output equipment, for example the computer to plate (CTP) device. 11 All participants used equal type of newspaper substrate of 45g/m 2 grammage. For the evaluation, the colour measurement conditions were according to ISO , the measurement geometry used was 45/0 with a 2º observer angle and CIELAB system. For documentation, all colorimetric measurements were made on white backing using Spectrolino/Spectroscan. Considering colour difference calculations, ΔE* ab values were computed between individual measurement data set and the characterization data set "IFRA26.txt. As the colorimetric variation tolerances in ISO are defined in CIELAB ΔE* ab, only this colour difference metric has been applied in this project. The arithmetic mean, 95 th percentile, and maximum of the resulting colour difference distributions were then computed. Although ISO is focusing on colorimetry, the current Norwegian newspaper production method is still exclusively density control oriented. Therefore the Norwegian Newspaper Publishers Association s required density measurements and density evaluation in this project. The density values mentioned above were adopted from the former ISO :1998 standard. However, the density measurement evaluation in this study were explicitly used for within sheet uniformity control according to the Association s requirements and only used for informative purposes. The IFRA Special Report 2.37 specifies the density tolerances with ±0.10 density. 13 The density measurements using SpectroEye were performed on black backing with DIN E and polarisation filter. The density deviation tolerance for densitometer measurements is ±0.01 according to DIN We measured the 50% patch 10 times and the variation was 213

225 Paper D less than 0.01 densities for all process colours. Therefore the repeatability analysis of the SpectoEye can be considered as satisfying. 2.2 Psychophysical Experiment The aim of the psychophysical experiment was to determine how pleasant the reproduction of a newspaper print was considered to be, when compared to the remaining newspaper reproduction prints. The observers were then asked to rank them in ascending order from best to worst. 7,8 Note that the two test print runs were carried out in this project with six months of time interval. According to the colour measurements of the first test print, a number of ICC profiles with different separation settings were generated. An expert panel determined the two appropriate colour separations, which were then used to carry out the psychophysical experiment using the test prints of the second test run. A total of two custom profiles and one industry standard ICC profile were applied to each test image (by using the relative colorimetric rendering intent) and printed in all the eight newspaper printing plants. 15 The two custom profiles characterize the specific Norwegian printing conditions whereas the standard ICC profile considers the conditions in accordance with the international standard ISO The three applied ICC profiles are: NADA X: Custom profile: Is the ICC profile based on the characterization data set Average80. The characterization data set Average80 is created using the measurement data, which is the average data of the measurement values of all eight newspaper printing plants which results in 80 measurements (8 printing plants x10 measurement each) of the IT8.7-3 CMYK testchart printed. The data have been averaged according to the weighted method, which ensures that less significance is attached to values with particularly large deviations when averaging. Profiling tool: GretagMacbeth ProfileMaker 5.0.8, Separation setting: Acromatic: MaxK, TIL: 240%, starting point: 5 NADA Y: Custom profile: Is the ICC profile based on two characterization data sets, which have been averaged according to the weighted method, which ensures that less significance is attached to values with particularly large deviations when averaging. The first characterization data set ( Average80 ) is based on the average data of the measurement values of all eight newspaper printing plants, 214

226 Paper D NADA Z: which results in 80 measurements (8x10 measurements) of the IT8.7-3 CMYK test chart printed. The second characterization data set is based on the characterization table IFRA26.txt 9 valid for the reference printing conditions relating to the international standard ISO Profiling tool: GretagMacbeth, ProfileMaker 5.0.8, Separation setting: Acromatic: MaxK, TIL: 240%, starting point: 5 Industry standard profile, ISOnewspaper26v4 2 : Based on the characterization table IFRA26.txt valid for the following reference printing conditions relating to the international standard ISO The images were simultaneously viewed side by side in a viewing cabinet, as shown in Figure 2 and no anchor stimuli were used as a reference. The viewing set up was based on the standard condition of the graphic art industry ISO 3664, vertical geometry 45, background grey, GretagMacbeth Judge II lighting cabinet with light source D50 and light intensity 100% (1035 lux). 16 The images were viewed in a darkened room (lights off) and the images, approximately 10cm x 15cm in size, were viewed from a distance of approximately 65 cm. Figure 2. Viewing arrangement for psychophysical experiment. Four test images were chosen to contain a range of different types of pictorial content, tonal and chromatic variety as shown in Figure 3. The chosen images represent typical image types used in daily coldset offset newspapers. srgb is the original colour space for all four images. Before the images were converted to CMYK using the different ICC profiles the provided images were verified on a calibrated and characterized monitor including soft proofing with the three printer profiles applied. 215

227 Paper D Figure 3. Four test images: camera, car, portrait and flag. The images camera and portrait are reproduced with the permission of Ole Jakob Bøe Skattum. The images car and flag are reproduced with the permission of Verdens Gang (VG). A total of 25 observers (14 experts and 11 naïve) with normal colour vision and an age range of 20 to 58, took part in the experiment. Each observer was required to assess four different images. Each image is reproduced by three different separation algorithms and printed in eight different printing plants. Note that the assessments were carried within each individual coldset offset newspaper press. The law of comparative judgement was applied and the method used was rank order. 17 The observer s task was to rank the three prints in the viewing cabinet in order, from best to worst, in terms of preferred pleasantness. In a study by Morovic 18, pleasantness is defined as the reproduction s correspondence with preconceived ideas of how a given image should look according to an individual, in terms of contrast, colour, sharpness, etc. 3 Results and discussion 3.1 Results of the quantitative evaluation We received 20 Test Prints (printed on one side only) from each of the eight participants. Subsequently we measured a total of ten test charts (IT8.7-3 CMYK Target) from each participant to evaluate their colorimetric properties. The test charts have been chosen according to the density control on solid CMYK bars. 5 density measurements have been performed on solids across the paper width (Figure 4). Subsequently the 5 densities across the paper width from all 20-test prints in each colour have been averaged. Table 1 shows the density values from each participant on Test Print 1 (column 1) and Test Print 2 (column 2). It can be seen that in Test Print 1 all participants showed an excellent performance within the IFRA density tolerance ±0.10 (except for Aftenbladet Trykk in yellow 0.77 density). However in Test Print 2, except for Dagblad-Trykk, Aftenbladet Trykk and Halden Arbeiderblad the density 216

228 Paper D performance was rather poor and the density values for some of the colours are clearly outside the IFRA density tolerance specifications. Notice, a high density value in black for Fædrelandsvennens Trykkeri has been observed. Figure 4. Test chart IT8.7-3 CMYK including solid bars across the paper width to perform density control. Table 1. Density performance from each participant in Test Print 1 (1.TP) and Test Print 2 (2.TP). The yellow marked densities are outside the IFRA tolerance value ±0.10 density. Participants Cyan Magenta Yellow Black 1.TP 2.TP 1.TP 2.TP 1.TP 2.TP 1.TP 2.TP Dagblad-Trykk 0, , ,95 1,0 1, Schibsted Trykk 0, , , , NR 1 Trykk 0, , , , Fædrelandsvennens Trykkeri 0, , , , Adressa Trykk 0, , , , Bergens Tidende 0, , ,98 1,0 1, Aftenbladet Trykk 0, , , , Halden Arbeiderblad 0, , , , In the next paragraph the results of the variations within the sheet, within the press and between the presses are reported. 217

229 Paper D Considering the solid ink uniformity within the sheet, Figure 5 shows the cyan ink variations on Test Print 1 and Test Print 2 for all participants. Although the solid ink uniformity in Test Print 1 (measured 5 times across the paper width) is within the IFRA density tolerance the variations in test print 2 are much larger and partly outside the IFRA tolerance. For almost all participants, a correlation can be seen between the large variations and the average density in test print 2. It is important to point out that the appropriate density and ink uniformity is an important issue of the printing press calibration and will determine the print quality. However, we emphasize again that density results are reported for informative purposes only. To determine the variation in terms of colour stability over time the quality factor repeatability is required. 19 Figure 6 shows the short-term repeatability performance of each participant individually according to the measurement results of Test Print 1. The colour differences were calculated between the mean of the measurements taken (10 press sheets) and each individual measurement (928 patches per sheet) called as Mean Colour Difference from the Mean (1). (1) Although some of the participants show larger variations, the results of all participants are still within an acceptable tolerance (Mean ΔE* ab < 1.0 units). Another property of evaluating the performance of a printing process is by analysing the long-term repeatability, the difference between Test Print 1 and Test Print 2 (time interval of six months). The colour differences were calculated between the mean of the measurements taken from Test Print 1 and the mean of the measurements taken from Test Print 2. It can be observed that Dagbladet, Fædrelandsvennens Trykkeri and Adressa Trykk show the largest colour differences in terms of the mean (ΔE* ab 5.09, ΔE* ab 4.92 units and ΔE* ab 4.93 units respectively). This can be explained due to different RIP settings between Test Print 1 and Test Print 2 as reported by the printing plants. On the other hand due to the excellent density performance between Test Print 1 and Test Print 2 it is expected that Halden Arbeiderblad shows a rather low colour difference (mean ΔE* ab 2.38 units) between the two test prints. A further approach to assess the colorimetric properties is to analyse the variations between the eight newspaper printing processes. Figure 8 presents the colour difference 218

230 Paper D between the average measurements of all eight newspaper printing plants and each individual participant in Test Print 2. Notice, that participant Fædrelandsvennens Trykkeri shows the largest difference (mean ΔE* ab > 4.8 units). Furthermore the max difference has been calculated for the solid black as expected due to the tremendous high density value (black density 1.49). The participant Dagblad-Trykk has the smallest difference (mean ΔE* ab 2.1 units). Figure 5. Solid cyan ink uniformity in Test Print 1 (left chart) and Test Print 2 (right chart) measured 5 times across the paper width. Figure 6. Short-term repeatability performance of each participant in Test Print

231 Paper D Figure 7. Long-term repeatability performance of each participant (six months of time interval). Figure 8: Variations between the eight participants according to Test Print 2. The next paragraph reports the results of Tone Value Increase and the colour measurement evaluation between the presses and the aim values defined in the ISO standard. The size of the halftone dots increases during the printing process. This is known as Tone Value Increase (TVI) or dot gain. It is important to know the TVI characteristics of the printing process to achieve high print quality. The RIP, if necessary, can adjust a TVI curve. Ideally, a press calibration starts with aiming the target values on the solid tones in 220

232 Paper D the primary colours, e.g. according to the defined CIELAB values given in the ISO standard. Subsequently, the midtone has to be determined. For example, ISO has defined a tone value increase curve of 26%. However, assuming that the tone value increase curve in the midtone range indicates, for example, only 22% instead of 26%, the curve must be raised by 4% in the RIP. Table 2 shows the TVI for the 50% control patch measured for Test Print 1 and Test Print 2. Note that the measurement results demonstrate large variation in terms of tone value both within the colours cyan, magenta, yellow and black, between the printing presses and between the two test prints. As mentioned previously, each printing plant was required to use the appropriate RIP settings according to the predefined Norwegian linearized newspaper production method including equal RIP setting for each primary colour. 10 However, in order to obtain a common specified dot gain (e.g. 26%) in all printing plants the RIP setting has to be adjusted individually for each printing press and for each colour. It can only be speculated what caused the large variations. A further important part of the evaluation is to analyse the variation between all eight participants and the colorimetric values of the characterization data set "IFRA26.txt. According to the results given in Table 3, the participant Dagblad-Trykk shows a small colour difference (mean ΔE* ab 4.12 units) while, the participant Adressa Trykk shows the largest colour difference (mean ΔE* ab 7.67 units). This result agrees with the tone value increase result for 50% value seen in Table 2 (Test Print 2) which have shown a very small difference between Dagblad-Trykk and ISO and a rather large difference between Adressa Trykk and ISO respectively. Table 2: Tone value increase for the 50% control patch measured on paper substrate with DIN E with polarisation filter on a black background. Participants: Tone value % Tone value % Test print 1 Test print 2 C M Y K C M Y K Dagblad-Trykk Schibsted Trykk Nr 1 Trykk Fædrelandsvennens Trykkeri Adressa Trykk Bergens Tidende Aftenbladet Trykk Halden Arbeiderblad NADA_avis_vjanu

233 Paper D ISO : Table 3. Colorimetric colour differences between each newspaper printing plant and the characterization data set "IFRA26.txt. ΔE* ab Mean 95 th Percentile MAX Dagblad-Trykk Schibsted Trykk Nr 1Trykk Fædrelandsvennens Trykkeri Adressa Trykk Bergens Tidende Aftenbladet Trykk Halden Arbeiderblad Looking at the CIELAB values for the primary colours cyan, magenta, yellow and black specified in ISO , the differences between the actual values and the nominal values must not exceed the tolerances shown in Table 5. Table 4 presents the CIELAB values for the primary colours from all eight participants, ISO and NADA average 2007 (Test Print 2). Moreover the table shows colour differences calculated between the participants and ISO The green marked values indicate that the colour differences are within the ISO tolerance of ΔE* ab 5 units. The violet marked values are outside the ISO tolerance. It is interesting to note that according to Test Print 1 almost all participants show colour differences within the ISO tolerance. On the other hand it can be noticed that except for Dagblad-Trykk and Halden Arbeiderblad the colour difference in Test Print 2 exceed noticeable the ISO tolerance for most of the participants. Although it might not be adequate to relate colorimetric results to density values, a correlation between the large colour differences presented in Table 4 and large density variations seen in Table 1 can be observed. 222

234 Paper D Table 4: CIELAB coordinates for newspaper according to measurements on white backing (ISO ). Table 5: CIELAB tolerances (ΔE* ab ) for the solid tones of primaries according to ISO Cyan Magenta Yellow Black Deviation Tolerances Note that all participants have used substrates that meet the tolerances for the colour of the print substrate (ISO ). Figure 9 shows the average colour gamut projection of the Test Print 1 and Test Print 2 and the ICC profile ISOnewspaper26v4 onto the a*b*-plane at to the level of lightness L*=50. The data used in this task is based on colorimetric measurements, which have been analysed and visualised by the icc3d application. 20 It can be seen that the average measurement of the Test Print 1 has the largest colour gamut. On the other hand the average measurement of all the 8 printing presses in Test Print 2 result in a smaller colour gamut. The measurement data of the ISOnewspaper26v4 profile lies between the two test prints. 223

235 Paper D Figure 9. 2D colour gamut comparison of Test Print 1 and Test Print 2 and the ICC profile ISOnewspaper26v4 onto the a*b*-plane according to the level of lightness L*= Results of the psychophysical experiment The following are the results in terms of z-scores, which have been obtained in the psychophysical experiment. The goal of the psychophysical experiment was to determine pleasantness of the reproductions made according to different newspaper prints. For each image and newspaper printing plant, the 3 x 1 matrices of ranking order results for each observer were arranged over the 25 observers and the raw data from the experiments were treated statistically to obtain z-scores. The precisions of the experimental results are described in terms of 95% confidence interval (CI), which is calculated using equation (1) which consists of the mean (R), standard deviation ( ) and the number of observations (N). Using case V of the method proposed by Thurstone, the standard deviation of the z- scores is assumed to be. 21 (1) 224

236 Paper D Figure 10. Rank order z-scores for each newspaper printing press and separation (The error bars represent the confidence interval 95% of population distribution). Figure 10 presents the results in terms of z-scores for each newspaper printing plant. For the number of images and observations, CI was calculated to be ±0.13. Overall it can be noticed that NADA X performs worst for all participants, except for Adressa Trykk. NADA Z performs significantly best for participant Fædrelandsvennens Trykkeri. On the other hand NADA Y was ranked significantly best for the participants Schibsted Trykk, Adressa Tykk, Bergens Tidende and Aftenbladet Trykk. Note that the profiles NADA Y and NADA Z do not indicate a significant difference either for Dagblad-Trykk nor Halden Arbeiderblad in terms of preferred pleasantness. Considering Figure 11 the experimental results identify the profile NADA Y as the candidate, which scored best for the images Portrait, Camera and Flag. NADA Z performs significantly best for the image Car. However, NADA X was ranked significantly worst in terms of preferred pleasantness. Table 6 gives the ranking of the performance of the three separation profiles for the four images. As it can be seen the profile NADA Y was significantly best except for the image Car and the profile NADA Z was second. Figure 12 presents the results in terms of z-scores for all four images and all eight newspaper participants. 225

237 Paper D For the total number of images, newspaper printing plants and observations, CI was calculated to be ±0.05. Notice, overall, the NADA Y profile performed significant best and the NADA Z profile second best. Figure 11. Ranking order z-scores for each type of image and separation (The error bars represent the confidence interval 95% of population distribution). Table 6. Ranking of separation profiles for each image (1=best, 3=worst). Image Profile NADA X Profile NADA Y Profile NADA Z Portrait Camera Car Flag Overall Figure 12. Overall z-scores for the three separation profiles. 226

238 Paper D 4 Conclusions and perspectives In this work, we have investigated eight newspaper printing presses in terms of their colour variations within the sheet, within the press and between the presses. Although all participants show an acceptable performance in the short-term repeatability (within the press), the within sheet variations are rather large especially in the second test print. Except for one printing press the long-term repeatability is inconsistent and has to be improved. The investigation between the presses also demonstrates some inconsistency. Considering the TVI, the results show rather large variations between the primary colours cyan, magenta, yellow and black, between the printing presses and between the two test prints. A number of factors can have affected the differences. In the course of our work, we have found that a common TVI specification in relation to dot gain is missing in the Norwegian newspaper production method. This can explain the inconsistency in terms of dot gain between the primary colours and between the printing presses. The RIP settings among the participants can be questioned. Eventually the print variations observed between Test Print 1 and Test Print 2 have affected the dot gain too. Another possible factor, which can contribute to the uncertainty of the printed results is the measurement technology used in the newspaper production process recommended by NADA. The Norwegian method proposes to use image-based dot meters not only for measuring and verifying the screening dots on the printing plates but also for measuring the dot gain on newspaper substrates. 10 A study by Wroldsen et al. 22 investigated the measurement performance of dot meters on newspaper substrate. The results of the repeatability analysis demonstrated very low confidence using dot meters in newspaper print, compared to using a colorimetric measurement approach. Considering the conformance to specified values in accordance with the requirements of ISO , the inconsistency of the density values on the solid primary colours, especially in Test Print 2, has affected the colorimetric values, CIELAB coordinates. Hence, for most of the participants the colour difference between the measurements of Test Print 2 and CIELAB coordinates in ISO has exceeded the ISO tolerance. Although the quantitative evaluation has demonstrated some obvious shortcomings there is a large potential for improving the target values of the ISO to obtain a better coherence between the newspaper printing presses. Nevertheless, to preserve the daily printing conditions and to match the colorimetric requirements of the adopted 227

239 Paper D standard profile it is highly recommended to perform press control according to a well defined standard e.g. ISO As seen from the results of the psychophysical experiment the ICC profile NADA Y performs significantly better than the other two profiles, NADA Z and NADA X. However, it is observed that the measurement parameters of NADA Z do not match the calibration parameters of most of the participants in Test Print 2. Therefore it might not be expected that NADA Z will perform better then NADA Y. Moreover, the colorimetric difference between the measurement data of NADA Y and NADA Z results in a colour difference ΔE* ab 2.5 which will be classified as rather small. On the other hand the print variations within the eight newspaper printing presses in Test Print 2 are between ΔE* ab 2.2 and ΔE* ab 5 which means that the print variations in colours between the eight printing presses are larger than the difference between the custom and the standard profile. Hence, the print inconsistency and not the profile selection may have determined the visual print quality. Therefore the outcomes in this study demonstrate the importance of adopting international standards and methods instead of using insufficiently defined house standards to preserve equal results among different newspaper printing presses. Standardising newspaper printing means improved communication with customers, and reducing or eliminating the costs of dealing with complaints. A result of that will make newspaper advertising more attractive and therefore improving the market conditions for newspapers in their competition with other media. Finally, it is of the interest to consider potential directions for further work in the field of process control and print quality assessment. Colour difference metrics for image quality assessment have been widely used for various applications. However, there is not very often a strong correlation between the objective evaluation and the visual assessment. Furthermore, the interpretation of the complete image quality assessment considering the colour difference calculation is dependent on the application and the acceptance. Eventually the acceptability threshold considering print quality assessment is defined as a vague concept and one that depends strongly on application and industry; further research is indeed needed to determine the requirements for acceptable print quality. Acknowledgments The author thanks the Norwegian Newspaper Publishers Association for their kind permission to use this material. Furthermore, the author wishes to extend its thanks to the participating members of the project for their grateful collaboration. 228

240 Paper D References 1. ISO Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 3: Coldset offset lithography on newsprint. ISO; IFRA. ISO Profiles Download Available: &CTDL&E& 3. Hardeberg J, Skarsbø S. Comparing color image quality of four digital presses. Proceedings of the 11th International Printing and Graphic Arts Conference, 2002; Bordeaux, France. 4. Nussbaum P, Hardeberg JY. Print quality evaluation and applied colour management in heat-set web offset. In Proceeedings of the 33rd International Research Conference of IARIGAI, 2006; Leipzig, Germany. p Nussbaum P, Hardeberg JY, Skarsbø SE. Print quality evaluation for governmental purchase decisions Proceeedings of the 31st International Research Conference of IARIGAI; 2004; Copenhagen, Danmark. Acta Graphica. p Pedersen M, Bonnier N, Hardeberg J, Albregtsen F. Attributes of image quality for color prints. Journal of Electronic Imaging 2010;19: Handley J. Comparative analysis of Bradley-Terry and Thurstone-Mosteller paired comparison models for image quality assessment. Proceedings PICS conference, p Gescheider G. Psychophysics, Method, Theory, and Application: Lawrence Erlbaum Associates; IFRA26 data set. Characterization data for standardized newspaper Coldset-Offset printing conditions Available: NADA Support Lineær avisproduksjon kort og godt. [Guideline]. Available: Kipphan H. Handbook of print media: technologies and production methods: Springer Verlag; ISO Graphic technology Spectral measurements and colorimetric computation for graphic arts images. ISO; IFRA. Revision of ISO : IFRA; 2005 Report nr Special Report

241 Paper D 14. DIN Prüfung von Drucken und Druckfarben der Drucktechnik Farbdichtemessung an Drucken. Teil 2: Anforderungen an die Messanordnung von Farbdichtemessgeräten und ihre Prüfung: Deutsches Institut für Normung; Sharma A. Understanding Color Management. New York: Thompson Delmar Learning; ISO Graphic technology and photography Viewing conditions. Geneva: ISO; Engeldrum P. Psychometric scaling: a toolkit for imaging systems development: Imcotek Press, Winchester, Mass.; Morovic J. Colour gamut mapping. In: Green P, MacDonald L, editors. Colour Engineering Achiving Device Independent Colours: John Wiley & Sons Ltd; p Morovic J, Nussbaum P. Factors affecting the appearance of print on opaque and transparent substrates. Journal of Imaging Science and Technology 2003;47(6): Farup I, Hardeberg J, Bakke A, Kopperud S, Rindal A. Visualization and interactive manipulation of color gamuts. Proceedings of the Color Imaging Conference, 2002; Scottsdale, Arizona. The Society for Imaging Science and Technology. p Thurstone L. A law of comparative judgment. Psychological review 1927;34(4): Wroldsen M, Nussbaum P, Hardeberg JY. Densitometric and Planimetric Measurement Techniques for Newspaper Printing TAGA 2008;4. 230

242 Paper E Paper E Maria Wroldsen, Peter Nussbaum and Jon Yngve Hardeberg A Comparison of Densitometric and Planimetric Measurement Techniques for Newspaper Printing Published: In TAGA Journal of Graphic Technology, Technical Association of the Graphic Arts, Volume 4, pp ,

243 Paper E 232

244 Paper E A Comparison of Densitometric and Planimetric Measurement Techniques for Newspaper Printing Maria S. Wroldsen, Peter Nussbaum, Jon Y. Hardeberg The Norwegian Color Research Laboratory, Faculty of Computer Science and Media Technology, Gjøvik University College, P.O.Box 191, N-2802 Gjøvik, Norway [email protected], [email protected], [email protected] Keywords: densitometer, dot meter, newsprint, process control First manuscript submitted to TAGA Journal, May 6, 2007 Revised manuscript submitted to TAGA Journal, November 20, 2007 Abstract Two types of measurement technologies are used for process control in newspaper printing, namely densitometric and planimetric technologies. Densitometric measurements are done with densitometers or spectrophotometers, while planimetric measurements are typically done with CCD image sensor-based instruments called dot meters. Although these two technologies are fundamentally different, they are often used interchangeably in print calibration and process control. In this paper we investigate the statistical relationship between densitometric and planimetric measurements on newspaper print. The aim of our project was to investigate whether it was possible to estimate halftone values measured by a densitometer, from the halftone values measured by different dot meters. The applied model is based on regression analysis using second order polynomials. The results are given as estimates of the polynomial parameters, i.e. the polynomials give the relation between halftone measurements with one of the dot meters and halftone measurements with the densitometer. Our statistical analysis showed that due to the large uncertainty of the estimated parameters, the model does not accurately describe the relationship between the two measurement technologies. This can be explained in part by the poor repeatability performance for dot meters applied to newspaper print. Moreover the measurement results also have shown significant variations within the three dot meters used in this experiment. Factors affecting the repeatability and determining the performance of the model are considered and discussed in this work. 233

245 Paper E 1. Introduction In newspaper printing essentially two types of measurement technologies are used for process control, namely densitometric and planimetric measurements. In densitometric measurements, the optical density is measured, and if needed converted to halftone values, typically using the Murray-Davies equation. In planimetric measurements, it is attempted to directly measure the halftone values, that is, the dot area coverage, typically using devices containing a CCD imaging sensor; such devices are often called dot meters. Although these two technologies are fundamentally different, they are often used interchangeably in print calibration and process control, in particular in the Norwegian newspaper industry (Aasen et al., 2002; NADA, 2007). This motivated us to investigate whether there is a statistically significant relationship between halftone measurements on newspaper print done with densitometers (converted into halftone value with the Murray-Davies-equation) and halftone measurements done with dot meters, The objective of this study is thus to find out whether it is possible to convert planimetric halftone measurements into densitometric halftone measurements and vice versa. Since these technologies are used interchangeably, it is important to know how to convert from planimetric measurements into densitometric measurements, to keep the printing process under control and to achieve high print quality. This study is limited to newspaper printed in coldset offset lithography using AMscreening. The test-target is printed in three different printing devices with different process parameters. Factors that possibly could have affected the final print (like printing parameters) are considered as noise and will not be discussed. Our focus is to find whether there is a statistical relationship between densitometric and planimetric measurements on newspaper print independent of factors that possibly could have affected the printing process. This study also includes repeatability analysis for the measuring devices; three dotmeters and one densitometer. It is necessary to know the repeatability of the measuring devices, to indicate the validity of a possibly relationship. After giving a brief overview of the different measurement technologies in Section 2 and the state of the art of research discussing their properties and relations in Section 3, we present our experimental setup and preliminary pre-tests in Section 4. In Section 5 we 234

246 Paper E present and dicuss our experimental results, before concluding and proposing ideas for further research in Section Measurement technologies The size of the halftone dots increase during the printing process. This is called dot gain. It is important to know the dot gain characteristics to achieve high print quality. The dot gain is divided into mechanical and optical dot gain. Mechanical dot gain is the result of growth during the printing process (Malmqvist et al, 1999). Optical dot gain appears due to absorption and light scattering in the ink and the paper. This makes the dots seem larger and darker than they really are. The sum of mechanical dot gain and optical dot gain is called total dot gain. Density is the light absorbtion ability of a material. The measurement of density is done with an instrument called a densitometer and is used to control colours in the printing process. Density is given by: D ink = log 10 I i I m, (1) where I m is the reflected light intensity and I i is the intensity of the incident light (Bergman, 2005). High density corresponds to high absorption. A reflection densitometer measures the amount of reflected light from a surface. It consists of a light source to illuminate the sample, optics to focus the light, filters to define the spectral response of the sample and a detector to monitor the reflected light. The sample is viewed at 45 degrees from the surface. The reflected light is then converted to density with a logarithmic amplifier and displayed digitally. The densitometer sees the dot almost like the human eye and provides an optical density value (Tobias Associates, 2007) Murray (1936) expressed the relationship between the reflection density of halftone prints and the dot area R, known as the Murray-Davies equation: R = D D R H 100%, (2) where D R is the density for a sample and D H is the solid ink density. Traditionally halftone dot area measurements have been done in laboratories with an instrument called a planimeter (Romano, 1996). This is the same as an image analyzer. In planimetric measurements, the dot area coverage is measured by using devices containing 235

247 Paper E a CCD imaging sensor. Such devices, designed for measuring printing plates, are often called dot meters. A dot meter combines a microscope and a CCD imaging sensor. According to Romano (1996) the major variables in a system like this are image capture, aperture selection and thresholding. The dot meter analyzes the digital image and decides what is a part of the dot and what is not based on a threshold. The camera takes a snapshot of the area being measured and literally counts the number of black and white pixels in the image. The dot meter is actually measuring the dot area and provides an absolute value of dot coverage (Colthorpe and Imhoff, 1999). The focus of the camera is an important factor. The depth of focus is typically less than 0.2 mm for any such system (Colthorpe and Imhoff, 1999). The dot meter uses the image histogram and a threshold to calculate dot area. The threshold defines how dark a pixel should be to be taken into account. 3. Literature review In the past, several studies regarding dot meter and densitometer measurements have been done; considering their reliability for different materials (Colthorpe and Imhoff, 1999; Hsieh et al., 2003), comparisons of the two measurement technologies (Spotts et al., 2005; Hsieh et al., 2003) and the use of an image analysis system to measure density (Malmqvist et al., 1993; Brydges et al., 1998). Most of these studies deal with measurement of printing plates. However, lately densitometers and dot meters are used interchangeably and not only for printing plates, but also on newspaper print (Aasen et al., 2002; NADA, 2007). Yule and Nielsen (1951) studied whether halftone values from density measurements corresponded to real dot areas. They found that halftone value calculated from density values with the Murray-Davies equation did not correspond to the real dot area coverage, because the effect of the penetration of light into the paper is usually neglected. Especially for uncoated papers the density of middle tones increases and multiple internal reflections from the paper surface increase it still further, so that the usual simple equation relating dot area to density is not accurate. Their general conclusion is that the relationship between dot area and halftone density is not nearly as simple as it appears. Arnaud (2001) compared three methods of image analysis to determine the physical area of dots on five different substrates, among these uncoated paper and printing plates. The three devices were all based on an optical microscope. His conclusion was that the dot gain (mechanical and optical dot gain) is a parameter at least as important as the solid ink 236

248 Paper E density in process control. Arnaud states that an image analysis software needs to be created specifically for the printing industry and that this software should be able to accurately measure dot area on any substrate (including papers). Romano (1996) states that measuring halftone dot areas on printing plates with a video image analyzer is a simple procedure, but tends to be rather subjective. It is very important to obtain a high quality image. According to Romano, the image quality is dependent of two criterias; the distance between the histogram s peaks (contrast) and the depth of the histogram valley (sharpness) of the 50% tint. Illumination is also an important factor. When it comes to size of aperture (field of view), Romano states that this is a critical factor to make accurate measurements. With small aperture size (enclosing only a few dots), errors can occur when the aperture is randomly placed. 4. Experimental approach As mentioned previously the aim of this study is to find out whether it is possible to convert planimetric halftone measurements into densitometric halftone measurements. Hence, a specific test target consisting of 16 patches (patch size 8x8mm) in different halftone values for each process colour (CMYK), was designed (Table 1 and Figure 1). The target was printed in coldset web-offset lithography using AM-screening, in three different printing plants (three different Norwegian newspapers), namely Bladet Sunnhordland, NR1 Trykk, and Orkla Trykk. In the following, the test targets are referred to with the name of the printing plant and a number indicating the sequence number of the copy. For instance, the test-target used as training-set is referred to as NR1 Trykk Table 1. Digital halftone values (as specified in the image file) for the test target used in this study (see Figure 1). C100% C97% C89% C81% C73% C65% C58% C55% C50% C45% C40% C32% C24% C16% C8% C3% M100% M97% M89% M81% M73% M65% M58% M55% M50% M45% M40% M32% M24% M16% M8% M3% Y100% Y97% Y89% Y81% Y73% Y65% Y58% Y55% Y50% Y45% Y40% Y32% Y24% Y16% Y8% Y3% K100% K97% K89% K81% K73% K65% K58% K55% K50% K45% K40% K32% K24% K16% K8% K3% 237

249 Paper E Figure 1. The test target. As densitometer, a GretagMacbeth Spectrolino spectrophotometer was used, under the following setup: Physical filter: Pol, White base: Paper, Illuminant: D65, Observer angle: 2º, Density standard: DIN NB. Three commercially available dot meters were used in this study (brand names withheld for anonymity). Considering the aperture size and the treshold method used to segment the digital image into ink and substrate areas the manufacturer s default settings have been used in this project. Given the halftone values measured by one of the dot meters, the aim was to predict halftone values of the densitometer. The applied prediction model is based on regression analysis using second order polynomials. The results are given as estimates of the polynomial parameters, i.e. the polynomials give the relation between halftone measurements with one of the dot meters and halftone measurements with the densitometer, as follows: 2 ydensitomet er = axdotmeter + bxdotmeter + c (3) In polynomial regression, it is important to avoid over-fitting. Graphs with measurement data indicated that the relation could be described with second order polynomials; the scatter plots showed slowly decreasing graphs. Third order polynomials were also investigated, but the third order terms were extremely small. Hence, the polynomials used are second order to avoid over-fitting, for details, refer to Wroldsen (2006). Two limitations of the model were introduced; if the predicted densitometer value exceeds 100% or is below 0%, the value is clipped to 100% and 0%, respectively. Empirical correlation coefficients were calculated to indicate whether a statistical relationship exists between the measurement datasets, as follows: 238

250 Paper E r = Correl( X,Y ) = ( x ( x 2 x ) x )( y y ) 2 ( y y ) (4) Because of significant measurement differences between the process colours, it was necessary to study each of them individually. Furthermore the measurement data were divided into two sets; a training set to establish the model and a test set to evaluate its performance. The residuals between the predicted and measured halftone values with the densitometer (with test set = training-set and test set training set, respectively) were used to judge the performance of the model. Because of significant measurement differences between the dot meters and also between the process colours for each densitometer-dotmeter-combination, it was necessary to study both the instrument combinations and the process colours individually (Wroldsen, 2006). The modelling and data analysis were therefore conducted separatly for each dot meter. The following method was used in this study to build and test the model (describing a possible relationship between densitometric and planimetric measurements) for each combination of instruments: First, three series of measurement data from one test target were used to establish the model (one model for each process colour). This measurement data constitutes the training set. Then, the residuals between predicted and measured halftone values with the densitometer (with the test set being part of the training set and with the test set totally independent of the training set, respectively) were used to judge the performance of the model. Some of the test targets in the test set were printed in another printing plant than the test target of the training set. 4.1 Preliminary repeatability tests To justify that the densitometer could be used as a reliable representative for all densitometers, we did a preliminary test with two different densitometers. This was done to verify whether different densitometers give the same result (in contrast with the dot meters which are based on thresholds). For this test we used the test-target named Bladet Sunnhordland The following patches were measured for each process colour: 100%, 81%, 50% and 24% (white base: paper). The density values were converted to halftone values using the Murray-Daviesequation, and the densitometer pretest showed the largest deviations for halftone values below 24%. This is probably caused by the conversion from logarithmic density values 239

251 Paper E into halftone values. Low density values converted to halftone values using the Murray- Davies equation result in larger variations than high density values. This effect is getting even more obvious with low solid ink densities, like in newspaper printing. Another critical factor is the number of decimals used for density measurement. Even though the densitometer pretest showed some deviations between the two densitometers, only one of them is used in the analysis. This was necessary to limit the analysis. The repeatability analysis of the densitometer was satisfying. We measured the 50% patch 10 times. The variation was less than 0.01 density for all the process colours. The tolerance density deviation for densitometer measurements is ±0.01 according to DIN (1995). A repeatability analysis was also conducted for the dot meters. On newspaper print the 50%-patch for each process colour (CMYK) was measured 10 times with each dot meter. Based on these measurements, we calculated the average; range (absolute value of maximum halftone value minus minimum halftone value) and standard devation were for the three dot meters on newspaper print (note: not printing plates). The repeatability analysis showed low repeatability for all three dot meters, as shown in Table 2, 3 and 4. Table 2. Repeatability analysis dot meter 1. Dot meter 1 Test-target: NR1 Trykk C50% M50% Y50% K50% Average 49.70% 46.35% 47.50% 41.05% Range 3.00% 3.00% 3.50% 1.00% Standard deviation 0.92% 1.11% 1.00% 0.28% Table 3. Repeatability analysis dot meter 2. Dot meter 2 Test-target: NR1 Trykk C50% M50% Y50% K50% Average 48.86% 48.25% 52.53% 39.69% Range 5.80% 6.10% 7.00% 2.80% Standard deviation 1.90% 1.67% 1.91% 0.95% Table 4. Repeatability analysis dot meter 3. Dot meter 3 Test-target: NR1 Trykk C50% M50% Y50% K50% Average 49.15% 42.25% 47.80% 40.40% Range 4.00% 3.50% 2.50% 2.00% Standard deviation 1.49% 0.98% 0.86% 0.57% 240

252 Paper E DIN (1995) states the tolerance variation of density measurements to be ±0.01. However, there is no standard dealing with acceptable variations for dot meter measurements. In accordance with ISO (2005) the optical density for CMY should be 0.9 and for black 1.1. Outside U.S the 26% tonal value curve is used. This means that the tone value increase at 40% or 50% should be 26%. Optical solid density 0.9 and 26% tone value curve make ±0.01 correspond to approximately 2% tone value for the middle tones (see Table 5). None of the three dot meters fulfilled this requirement. Table 5. Tolerance tone value for the middle tones. Density Murray-Davies D H D R Halftone value % % % % According to the presented results, it is not possible to decide whether this low repeatability is caused by the measuring devices and/or inhomogeneous halftone values within one patch. Print irregularities cause noticeable differences in measured halftone values and reduce the repeatability when the aperture is small. Large screen dots used in newspapers in combination with small aperture is therefore unfavorable. It is not unambiguous to decide what is substratum and what is part of a screen dot, especially for middle halftone values, due to high optical dot gain (see Figure 2). Figure 2. Image of a 50% halftone value printed on newspaper with magenta ink taken with a Zeiss Axioplan 2 imaging microscope ( Maria S. Wroldsen). 241

253 Paper E As mentioned previously the dot meter analyzes the digital image and according to a certain defined threshold operation the colour image will be converted into a high-contrast, black-and-white image. All pixels lighter than the defined threshold are converted to white; all pixels darker are converted to black. Figure 3 illustrates variations in terms of defining an appropriate threshold to determine the size of the dot area. Taking a porous and scattering substrate such as newsprint into consideration the main difficulty is to determine an ideal threshold. Figure 3. Treshold variations and the corresponding % dot area. Image A with treshold 100 results in 25% dot area, Image B with treshold 128 results in 42% dot area and Image C with treshold 150 results in 52% dot area. 5. Experimental Results and Discussion The calculation of empirical correlation between the densitometer and dot meter values indicated relationship (see Table 6). Due to the fact that the correlation coefficients were close to 1, it can be assumed that there must be a correlation between the halftone values measured by the various instruments. Table 6. Empirical correlation coefficients Dot meter 1 Dot meter 2 Dot meter 3 Densitometer, C Densitometer, M Densitometer, Y Densitometer, K Based on three measurement series of one test target, second order polynomials estimating the relationship between halftone measurements with the dot meters and the corresponding halftone measurement with the densitometer were established. Figures 4, 5, 6 and 7 show measurement data with belonging trendlines for the three measurement series with dot meter 1, of the test target NR1 Trykk ; this constitutes the training set for our model. These graphs show that the trendlines highly fit the measurement datas. To test this model, we calculated residuals between predicted and measured halftone value. 242

254 Paper E Cya n 100,00 Halftone value in %, (densitometer) 90,00 80,00 70,00 60,00 50,00 40,00 30,00 20,00 10,00 y = -0,006x 2 + 1,6953x - 4,7409 R 2 = 0,9937 0,00 0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 Halftone value in % (dot meter 1) Figure 4: Relation for cyan ( NR1 Trykk 24000, dot meter 1) Magenta 100,00 Halftone value in %, (densitometer) 90,00 80,00 70,00 60,00 50,00 40,00 30,00 20,00 10,00 y = -0,0042x 2 + 1,4406x + 4,4598 R 2 = 0,9953 0,00 0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 Halftone value in % (dot meter 1) Figure 5: Relation for magenta ( NR1 Trykk 24000, dot meter 1) 243

255 Paper E Yellow 100,00 Halftone value in %, (densitometer) 90,00 80,00 70,00 60,00 50,00 40,00 30,00 20,00 10,00 y = -0,0049x 2 + 1,468x + 3,606 R 2 = 0,9965 0,00 0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 Halftone value in % (dot meter 1) Figure 6: Relation for yellow ( NR1 Trykk 24000, dot meter 1) Black 100,00 Halftone value in %, (densitometer) 90,00 80,00 70,00 60,00 50,00 40,00 30,00 20,00 10,00 y = -0,0052x 2 + 1,5301x + 1,5519 R 2 = 0,9967 0,00 0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 Halftone value in % (dot meter 1) Figure 7: Relation for black ( NR1 Trykk 24000, dot meter 1) The obtained regression polynomials for dot meter 1 are given in Equations 5-8. Cyan : 2 y densitometer = 0.006x dotmeter x dotmeter (5) Magenta : 2 y densitometer = x dotmeter x dotmeter (6) 244

256 Paper E Yellow : Black : 2 y densitometer = x dotmeter x dotmeter (7) 2 y densitometer = x dotmeter x dotmeter (8) The residual between predicted and measured halftone value with the densitometer was calculated as in the following example with cyan 65%: Measured halftone value with dot meter 1: 59.50% Measured halftone value with the densitometer: 78.12% Predicted halftone value with the densitometer: 74.89% (relation for cyan; y = x x ) The residual between measured and predicted halftone value with the densitometer: 3.23% Table 7 shows the residuals when the test-set is part of the training-set. Because the test set is part of the training set, it is expected that the differences between the predicted and measured halftone values are rather small. As it can be seen in Table 7 the variations are colour and halftone value independent and does not follow a certain trend although cyan shows the largest variations. However, there is no significant trend for the obtained variations. Although the model performs well it is important to test the model with another test-set. Table 8 shows the residuals when the test-set is not part of the training-set. It can be seen that the model does not perform that well applying another test-set. Consequently the differences between the predicted and measured halftone values are larger. Although the residuals have increased, the variation still does not follow a certain trend. Table 7. Dot meter 1, densitometer: Residuals in % when the test set is part of the training set. (Test set: NR1 Trykk 24000, 1. measurement series.) Residuals Patch Cyan Magenta Yellow Black 100% % % % % % % % % % % %

257 Paper E 24% % % % Table 8. Dot meter 1, densitometer: Residuals in % when the test set is different from the training set. (Test set: Orkla Trykk 5000.) Residuals Patch Cyan Magenta Yellow Black 100% % % % % % % % % % % % % % % % Residuals between predicted and measured densitometer values were calculated for several test sets different from the training set. Some of these test-sets where printed in different printing processes than the training-set. As stated earlier, 2% can be considered the tolerance variation for dot meters. Some residuals are larger than 2% and this indicates that the model is not good enough to describe a relation. The low repeatability is an unfavorable factor. We have so far presented and the detailed results from the first dot meter only. In the following we summarize briefly the results for the two other dot meters, for more detailed information refer to Wroldsen (2006). For dot meter 2 the obtained regression polynomials are shown in Equations The residuals are larger when test-set is different from the training-set (see Table 10) than for test-set part of training-set (see Table 9) and indicates that our second order polynomials 246

258 Paper E do not satisfactory describe a possibly relationship between halftone measurements with dot meter 2 and the densitometer. This is partly caused by the low repeatability for dot meter 2. 2 Cyan : ydensitometer = xdotmeter xdotmeter (9) 2 Magenta : ydensitomet er = xdotmeter xdotmeter (10) 2 Yellow : ydensitomet er = xdotmeter xdotmeter (11) 2 Black : ydensitomet er = xdotmeter xdotmeter (12) The polynomials for dot meter 3 is are given in Equations 13-16, and the residuals are given in Table 9 and 10. We see the same trend as for the two other combinations of instruments. The average residuals are larger when test-set is not part of training-set. As for the two other combinations, the low repeatability of the dot meter is one factor that makes our model unsatisfactory. Cyan : y 2 =. 0049x x (13) densitomet er 0 dotmeter3 dotmeter3 +. Magenta : y 2 =. 0044x x (14) densitomet er 0 dotmeter3 dotmeter3 +. Yellow : Black : y y 2 =. 0043x x (15) densitomet er 0 dotmeter3 dotmeter =. 0047x x (16) densitomet er 0 dotmeter3 dotmeter3 +. Table 9. Residuals in % between dot meter and densitometer when the test set is part of the training set. (Test set: NR1 Trykk 24000, 1. measurement series.) Residuals Cyan Magenta Yellow Black Dot Max Meter 1 Average Dot Max meter 2 Average Dot Max meter 3 Average

259 Paper E Table 10. Residuals in % between dot meter and densitometer when the test set is different from the training set. (Test-set: Orkla Trykk 5000.) Residuals Cyan Magenta Yellow Black Dot Max Meter 1 Average Dot Max meter 2 Average Dot Max meter 3 Average Conclusions and Perspectives Our statistical analysis showed that due to large uncertainty of the estimated parameters, the model does not accurately describe the relation between the two measurement technologies. This can be explained by the poor repeatability performance for dot meters applied in newspaper print. The repeatability analysis provided, already in the first part of this study, low confidence using dot meters in newspaper print. None of the three dot meters fulfilled the requirement of 2% tolerance deviation (note: these are requirements which are not defined in an official standard). Dot meters are originally developed for measuring printing plates only. The residuals between predicted and measured half tone values with the densitometer increased when the test set was different from the training set, as would be expected. Moreover, the measurement results have shown significant variations within the three dot meters. Some factors affecting the repeatability and determining the performance of the model are listed in this section. Important factors that impair the use of dot meters on newspaper print: Print irregularities cause noticeable differences in measured halftone values and reduce the repeatability when the aperture is small. Small aperture in combination with large halftone dots used in newspapers are unfavourable. Due to high optical dot gain (especially for the middle tones and in newspaper print) it is ambiguous to decide what is substratum and what is part of a screen dot. Large residuals between predicted and measured halftone values for the middle tones could partly be explained by the high optical dot gain and problems due to determination of threshold (what is substratum and what is not) in the image analysis. 248

260 Paper E Important factors that impair the use of densitometers on newspaper print (when using the Murray-Davies equation to calculate halftone values): The convertion from logarithmic density values into halftone values with the Murray- Davies equation causes a slowly decreasing graph that makes low density values converted to halftone values result in larger variations than high density values. The effect of this conversion is even more obvious when used with low solid ink densities like in newspapers. Based on these results, dot meters are not recommended for halftone measurements on paper substrates in newspaper printing. Throughout this project some ideas of further research to investigate a possibly relation between densitometric and planimetric measurement emerged. The test target was printed in different printing processes with different solid ink densities, even though the instructions for printing said K 1.10 and CMY 0.9 in accordance with ISO (2005). It is difficult to control this in newspaper printing. It would have been interesting to do the same experiment with a print medium where accurate solid ink densities are possible. The uncertainty of dot meters and densitometers for use in newspaper printing is too high. Another type of paper (with lower optical gain) would also be preferable. To increase the repeatability it is advantageous with coated paper, accurate solid ink density and finer screen ruling (the aperture size would not be so critical). Moreover, FM-screening could be used. The reason why we did this experiment with newspaper in the first place is the increasing use of dot meters in the newspaper industry. More than one copy of each instrument could have been included in the repeatability analysis to make any variations between copies become visible. Image analysis of halftone images would be interesting to investigate the decision of threshold (what is part of a screen dot and what is not) and to illustrate the percent of optical dot gain for different halftone values. Different thresholds could be set and the result (dot area coverage) could be compared to measured values for the different dot meters. Different size of aperture could also be simulated to observe the influence of calculated dot area coverage. This experiment could perhaps lead to a recommandation of optimal size of aperture for different screen rulings; what size of aperture is necessary to avoid systematical errors? 249

261 Paper E Acknowledgments First, we would like to thank senior lecturer Sven Erik Skarsbø for providing supervision to this master thesis project (in collaboration with professor Jon Y. Hardeberg and lecturer Peter Nussbaum). Thanks for your interest, motivation and help throughout this project! Secondly, we would like to thank associate professor Are Strandlie for valuable help with the statistical analysis. Finally, we would like to thank Mediebedriftenes Landsforening for financial support and the newspapers (Bladet Sunnhordland, Romerikes Blad and Drammens Tidende) for offering to print the test targets. References Arnaud, S. (2001). Measurement of dot area, In TAGA Proceedings, pp Bergman, L. (2005). Using Multicoloured Halftone Screens for Offset Print Quality Monitoring, Licentiate Thesis No. 1147, Linköping University. DIN (1995), Prüfung von Drucken und Druckfarben der Drucktechnik Farbdichtemessung an Drucken. Teil 2: Anforderungen an die Messanordnung von Farbdichtemessgeräten und ihre Prüfung. Deutsches Institut für Normung. Colthorpe, S. and Imhoff, G. (1999). CTP Why densitometers do not work, White paper, Centurfax Ltd and Grip Digital Inc. Hsieh, Y., Wu, Y. and Lin, W. (2003). An Expermental Research to Compare Devices for Measuring Aluminium Lithographic Printing Plates., In TAGA Proceedings, pp ISO (2005). Graphic technology Process control for the manufacture of halftone colour separations, proofs and production prints. Part 3: Coldset offset lithography on newsprint. International Organization for Standardization. Malmqvist, K., Verikas, A. and Bergman, L. (1999). Consistency of mechanical dot gain - a hidden quality parameter, TAGA 51st Annual Technical Conference, pp Murray, A. (1936). Monochrome Reproduction in Photoengraving, J. Franklin Institute, vol. 221, pp

262 Paper E NADA (Visited 2007), Support Lineær avisproduksjon kort og godt. Online: In Norwegian. (NADA is a company working with digital ad delivery for Norwegian newspapers, see also Romano, D. (1999). The Image Analyzer A True Dot Area Meter? In TAGA Proceedings, pp Tobias Associates, Inc. (Visited 2007), Reflection Densitometry, Online: Yule, J. A. C. and Nielsen, W. J. (1951). The penetration of light into paper and its effect on halftone reproduction, Taga Proceedings, pp Wroldsen, M. S (2006). Densitometriske og planimetriske målinger av rasterstrukturer, Master s thesis, Gjøvik University College, Gjøvik, Norway. In Norwegian. Aasen, E. Danielsen, Ø., and Bovolden, A. J. (2002), Halftone measurements in newspaper print, Bachelor thesis, Gjøvik University College, Gjøvik, Norway. In Norwegian. 251

263 Paper E 252

264 Paper F Paper F Peter Nussbaum, Aditya Sole and Jon Yngve Hardeberg Analysis of Color Measurement Uncertainty in a Color Managed Printing Workflow Accepted for publication in Journal of Print and Media Technology Research, The International Association of Research Organizations for the Information, Media and Graphic Arts Industries, (iarigai). 253

265 Paper F 254

266 Paper F Analysis of Color Measurement Uncertainty in a Color Managed Printing Workflow Peter Nussbaum, Aditya Sole, Jon Y. Hardeberg The Norwegian Color Research Laboratory, Faculty of Computer Science and Media Technology, Gjøvik University College, P.O.Box 191, N-2802 Gjøvik, Norway [email protected], [email protected], [email protected] Keywords: Color measurement, calibration, color differences, print quality assessment, color management Abstract Since the recent revision of ISO and ISO , specifiying the requirements for systems that are used to produce hard-copy digital proof prints, the use of color measurement instruments is even more than before required in the printing industry. Currently, there are many different makes and models of color measurement instruments used in the industry. Therefore, in a modern color managed printing workflow, most of the printing houses use more than one color measurement instrument, typically one instrument in each department (pre-press, press, and post-press). In this paper, a total of nine commercial spectrophotometers are compared in terms of measurement uncertainty, precision and accuracy, repeatability and reproducibility. The BCRA series 2 ceramic gloss tiles are used to confirm the accuracy and repeatability of these measuring instruments according to the manufacturer s standards. We focus especially on inter-instrument and inter-model reproducibility and discuss the effect of instrument calibration and certification. For our experimental setup, four different materials are used, one proof print, one commercial print, and one reference print, along with the BCRA series 2 ceramic gloss tiles. In a color managed printing workflow the use of more than one instrument can impair and complicate the color process control due to the color differences between different measurement devices. The effect of the colorimetric measurement errors due to large interinstrument and inter-model variability between instruments used in different parts of the workflow (e,g, in the printing house, at the customer s site for inspection, and for certification) is discussed and demonstrated in this paper. 255

267 Paper F 1. Introduction Recently ISO [1] and ISO [2], have defined the colorimetric parameters for process control in the graphic arts, and tolerances for their acceptance. However, the color measurement devices used in a production workflow (from the costumer, designer, prepress, to printing house) may show variations in terms of precision (repeatability, reproducibility) and accuracy of the measurements made. Furthermore, in the context of PSO (Process-Standard Offset) certification, test prints and proofs are printed according to certain aim parameters. The prints and proofs are measured twice, firstly in the printing house with the instrument of the company, and secondly by the certification body to ensure that prints are made within the predefined tolerances. In a practical application, if both measurement devices result in values which qualifies the print or proof as approved, then nobody will question the reliability (precision and accuracy) of the instruments. Similarly if both instruments give values that are outside the given ISO production tolerance. However, if only one of them qualifies the results to pass then you might think of the other one as a false-positive. Moreover, it lies in the Human s nature to believe that the instrument, which does not qualify the print or proof as approved is not performing appropriate - without taking into consideration that it might be the proof or print which is in deed produced outside the defined tolerances. However, by improving the instrument accuracy and reducing the inter-instrument and inter-model agreement will contribute in reduce inappropriate considerations. The aim of the presented work is to evaluate the performance of nine color measurement instruments in terms of precision (repeatability, reproducibility) and accuracy. Furthermore, the effect of color measurement variability in a color managed printing workflow will be demonstrated. In particular the result of inter-model agreement measuring colors on paper substrate will be reviewed. After this brief introduction, we give some more background information in Section 2, by illustrating the problem, defining key concepts, and discussing central references. Then, in Section 3, we describe our experimental setup. In Section 4 we present and discuss our results, before concluding in Section 5 256

268 Paper F 2. Background To illustrate how measurement uncertainty in a printing production workflow may cause unexpected discussions, Figure 1 shows a simplified diagram of a practical scenario in which two instruments are used to measure the same target in a color workflow. Given a certain color patch reference and measuring the patch with the color measurement instrument of the customer will result in ΔE* ab 3.5. Although the inter-instrument reproducibility between the customer and print house measurement devices is ΔE* ab 3.0, the color difference measured with the print house measurement device on the color patch is almost twice the one measured with the instrument of the customer. Furthermore, assuming having a certain color difference tolerance of e.g. ΔE* ab 5.0 the result of the print house measurement of ΔE* ab 6.5 would not be accepted. Figure 1. Simplified diagram of a practical scenario using two instruments in a workflow measuring the same target/reference. According to Berns [3], measurement uncertainty can be divided into two main categories, precision and accuracy (Figure 2). Precision describes the dispersion of the measurements taken. On the other hand accuracy describes the distance between the measurements taken by the color measurement instruments and the actual target value (Figure 3). Accuracy is affected by systematic errors, which are errors due to different geometry, detector linearity errors resulting from wavelength. Precision can be further divided into repeatability and reproducibility. ASTM Standard E 284 [4] defines the repeatability as the closeness of agreement between the results of successive measurement of the same test specimen, or test specimens taken at random from a homogeneous supply, carried out in a single laboratory, by the same method of measurement, operator, and measurement instrument with a repetition over a specified 257

269 Paper F period of time. On the other hand, changing conditions such as the operator, measuring instrument, laboratory, or time, gives a measure of reproducibility. Figure 2. Overview over measurement uncertainty. Figure 3. Accuracy describes the distance between the measurement and the target and precision the dispersion of the measurement taken (after Berns' figure on page 95 [3]). In the past, various studies and research regarding measurement uncertainties and comparative studies of color measurement instrumentations have been presented. Already 40 years ago Billmeyer [5] presented a work where he studied the comparative performance of fifteen different color measurement instrument according to their precision and accuracy. Some years later another research conducted by the same author [6] concludes that the three instruments tested over en period of seven weeks are essentially equivalent in the precision and accuracy, measuring a large variety of samples, including textile. A slightly different approach has Rodgers [7] used in his comparative study of color measurement instrumentation were he compared not only the inter-instrument agreement but also the user friendliness of the software and computer interface, vendor amenability to a long term logistical and maintenance relationship and finally the price. A quality improvement team considered the most critical parameter to be the inter-instrument 258

270 Paper F agreement, followed by the software, the logistical/service relationship, and at last the price. Another comprehensive set of work was carried out by Wyble addressing the evaluation of methods for verifying the performance of color measurement instrument such as integrating sphere and bidirectional devices. Part I is covering the repeatability issue [8] and part II is addressing inter-instrument reproducibility [9]. However, our main contribution is the analysis of color measurement variability of using a number of different color measurement instruments in a color managed printing workflow. The specifications, methods and procedures to evaluate the performance of color measuring instruments in terms of repeatability and accuracy are defined in ASTM E2214 [10]. Because, the assessment of the instrument s performance and the analysis of the color measurement uncertainty is an important aim of this work, a brief evaluation of the methods is provided here. Accuracy describes the conformance of a series of readings to the accepted or true value. Repeatability defines how well an instrument repeats it s reading of the same target over a certain period of time. The assessment of the instrument s consistency can be tested over three periods of time. First is the short-term repeatability which is based on measurements made in succession, second is the medium-term repeatability which can be based on measurement s made over a period of hours and finally the long-term repeatability which is based on measurements made over weeks or longer. The short-term measurements can be performed either with or without replacement of the measuring instrument from the color tile/patch to be measured. When measuring without replacement, the tile/patch is left in place at the instrument s aperture. This approach might be dependent on the instrument technology and the user interface. To obtain the most reproducible results, measurements have been restricted to the central region of the tiles. The color differences are calculated between the mean of the measurements taken and each individual measurement called as Mean Color Difference from the Mean [3]. (1) Reproducibility is a form of repeatability in which one or more of the measurement parameters have been systematically changed such as the target is different, the time frame of measurements are very long, the procedures or instrument are different or the operator has changed. 259

271 Paper F Inter-instrument agreement describes the reproducibility of two or more instruments of identical design (In this study e.g. instrument 1-3 have identical design). Inter-model agreement expresses the reproducibility of two or more instruments of different design (e.g. instrument 4 and instrument 5). In other words, reproducibility determines the variations between instrument s readings. Instruments, which have the same design the random amount of bias is reduced compare to instrument with different design. The two types of reproducibility can be tested in a similar way. The most common way of testing is pairwise color difference assessment of a series of specimens. Various color difference parameters are used in the literature [10] including the mean color difference, maximum, the root mean square (RMS) color difference or the MCDM. RMS color difference ΔE is calculated as, (2) which is similar to average standard deviation, for N values of ΔE. 3. Experimental approach 3.1 Methods and procedures To determine the accuracy of the used instruments the color difference has been calculated between the measurement average value and the corresponding true value. For each instrument 15 measurements of each tile has been taken in a sequence and the average has been calculated. In this work, the true value is related to the reference values provided by CERAM who is a high-accuracy laboratory and the manufacturer of the used 14 BCRA tiles. Due to practical reasons only the short-term repeatability including 15 measurements in sequence and the long-term repeatability with 10 weeks interval of the instruments have been assessed. The sample used for this test is the White BCRA tile and the color differences were calculated according to MCDM. It is worth mentioning that before the 15 actual measurements have been conducted to assess the short-term repeatability and after 10 weeks to determine the long-term repeatability a warm up procedure including 25 measurements in a row on its own white standard has been performed. To determine the inter-instrument agreement 14 BCRA tiles have been quanified and the mean from each set has been calculated and consequently compared instrument pairwise. To assess the inter-model agreement the color difference between the average of 15 measurements of 260

272 Paper F each BCRA sample has been calculated. Thus, the pairwise contrast was compared between the different types of instruments. 3.2 Instruments In this paper, nine commercial spectrophotometers (one bench-top and 8 hand-held) typically used in the graphic arts industry have been analysed. Table 1 presents the instruments and their manufactor s specifications. Because one of the manufacturers is requesting not to publish their name the names were anonymised by identifying the instruments with instrument 1, instrument 2 instrument 9. All instruments represent bidirectional measurement geometries and uses similar light sources. The instrument s aperture size varies from 2 mm up to 4,5 mm. Note, that only two instruments had valid certification (VC) at the time of testing and for seven instruments the manufactory s certification have expired (EC). Nevertheless, the authors were aware of the situation obtaining measurements from 7 instruments, which were not re-certified at the time of the experiment. The 7 (uncertified) instruments are between three and eight years old and not used very often. However, due to the fact that the presented study is intended to demonstrate the real situation and not reproducing laboratory conditions the choice of the used instruments can be justified. According to the authors experiences from the field it is not uncommon to use instruments in the printing industry, which the re-certification have expired. Therefore the instruments used in this study represent the realistic situation, which is common in the practical production environment. As stated in ICC [11] when comparing, instruments can be divided into product families which are instruments of the same model from the same manufacturer using equal parameters (e.g. in this work instrument 1-3 can be considered as one family and named Family A and instrument 6-9 Family B ). In terms of repeatability, or reproducibility, instruments with identical design (inter-instrument) or different design (inter-model) can also be compared. 261

273 Paper F Instrument 1 (EC) Instrument 2 (EC) Instrument 3 (EC) Instrument 4 (VC) Instrument 5 (EC) Instrument 6 (EC) Instrument 7 (EC) Instrument 8 (EC) Instrument 9 (VC) Family A Family B Measuring without replacement Aperture in mm Yes 4 No 2 Yes Yes x:0 45 x:0 45 x:0 45 x:0 Light source Tungsten lamp, gas filled, type A light Type A Tungsten lamp, gas filled, type A light Tungsten lamp, gas filled, type A light Measuring geometry Manufacturer s claimed precision Mean ΔE* ab 0.3 Max ΔE* ab 0.8 by 12 BCRA tiles Mean ΔE* 94 < 1.0 by 12 BCRA tiles Mean ΔE* ab 0.3 by 12 BCRA tiles Mean ΔE* Max ΔE* by 12 BCRA tiles Spectral range and interval 380nm to 730nm at 10nm 380nm to 780nm at 10nm 380nm to 780nm at 10nm 380nm to 780nm at 10nm Manufacturer s claimed Short term repeatability ΔE* ab 0.02 (Standard shift from 10 measurements at 10 sec. interval on white) ΔE* 94 < 0.2 ΔE* ab 0.02 (Standard shift from 10 measurements at 10 sec. interval on white) ΔE* 94 < 0.1 (From 10 measurements at 3 sec. interval on white) Table 1. Overview of the nine instruments (two with valid certification [VC] and seven with expired certification [EC]) used in this work and the corresponding manufacturer s specifications. Different components (such as light source, detector and dispersing element) and their properties consisting of the measurement instrument determine the value of a measured sample. However, it is not the scope of that work to investigate these properties and their contribution to the measurement results. A comprehensive overview of color measurement fundamentals and different instrument components is given by Battle [12]. 3.3 Calibration procedure Before conducting the measurements, normal warm up and calibration procedures were followed. To warm up the instrument, 25 measurements in a row on its own white standard were made. Consequently, each instrument has been calibrated on its own white reference 262

274 Paper F tile supplied by the manufacturer along with the instrument (absolute white calibration). The absolute spectrum of the white reference tile is stored in the instrument and during the calibration the obtaining spectral response is adjusted so that it matches the stored spectrum. Thus, the instrument s internal software is calculating the spectral reflectance of the measured samples [13]. Instrument manufacturers typically recommend a calibration procedure at least once a day or if numerous measurements are undertaken in rapid succession. Furthermore, if thermal oscillation occur perhaps due to changes in the room temperature or due to the instrument s measurement lamp which can be frequently switched on recalibration is recommended to keep the measurements constant [14]. 3.4 Test procedure on BCRA tiles To evaluate the performance of the instruments in terms of accuracy, repeatability and reproducibility a series of British Ceramic Research Association (BCRA) Ceramic Color Standards Series II (CCS II) ceramic tiles have been employed [15]. In this paper, 14 BCRA ceramic gloss tiles including one Black and one White BCRA ceramic gloss tile were measured and compared to the true value of the BCRA tiles to determine the instruments performance. The measurement procedures were done according to ISO [16]. 3.5 Test procedure on printed substrates In order to analyse the measurement uncertainities of the color measurement instruments on commercial printed substrates, measurements were conducted on the UGRA/FOGRA Media Wedge CMYK [17], which includes 46 color patches (Figure 4). The Media Wedge was printed on three different paper substrates. The first paper substrate was a hard-copy digital proof print, printed according to the ISO graphic art standards for paper type 1 simulation by a commercial printing house. The second paper substrate was paper type 1 printed by the same commercial printing house aiming the ISO graphic art standards. And the third paper substrate was paper type 5 Altona testsuite reference print [18]. Before measuring the color patches of the Media Wedge, warm up and calibration procedure was followed as discussed previously. The Media Wedge was measured three times in a sequence with each instrument. White backing material in accordance with ISO [16] was used. 263

275 Paper F Figure 4. Figure 14, UGRA/FOGRA Media Wedge CMYK. 3.6 Data collections All instruments used in this paper reported spectral reflectance factor values from 380nm to 730nm with 10nm interval. Spectral measurements taken from the BCRA tiles and the paper prints were converted to CIEXYZ tristimulus values according to the CIE observer and the CIE Standard illuminant D50 using the method proposed by ASTM 308, Table 1 [19]. Further details on these calculations are well documentet in e.g. Hunt [20] too. Furthermore, CIELAB (D50 as the reference white) values were calculated according to CIE 15 specifications. Colorimetric difference ΔE* ab and ΔE* 94 (as some manufacturers quoted the color difference in ΔE* 94 ) values were computed between the BCRA reference data and the measurements data obtained using each instrument [21]. All measurements in this study have been conducted in the same location and the identical room temperature conditions. 4. Experimental results and discussion As mentioned before the presented article is aiming to evaluate the performance of color measurement instruments in terms of precision and accuracy and is demonstrating the effect of color measurement variability in a color managed printing workflow. 4.1 Measurement accuracy As stated earlier ISO defines accuracy as the conformance of a series of measurements to the accepted value for a given sample. In other words how closely an instrument can conform to a certain reference or true values. In this work the reference values have been provided by CERAM who is the manufacturer of the used 14 BCRA tiles. Figure 5 shows the color difference between each instrument s reading on the BCRA tile and the corresponding true value. Overall, it can be observed that the chromatic BCRA tiles (e.g. Red, Orange and Yellow) produces larger color differences than the achromatic BCRA tiles. Furthermore, all the instruments demonstrate a smaller color differences for the Black tile than the White tile and the dark to light grey tiles show very similar behavior. 264

276 Paper F Figure 5. Color difference of nine instruments according to the 14 BCRA tiles reference. The better performance of the instruments on the Black tile suggests good black trap calibration. On the other hand the larger differences on the White tile may indicate that the instruments do not agree very well on the definition of white, which could be traced to the instruments calibration or missing certification (Table 2). Black White Instrument 1 (EC) Family A Instrument 2 (EC) Instrument 3 (EC) Instrument 4 (VC) Instrument 5 (EC) Instrument 6 (EC) Family B Instrument 7 (EC) Instrument 8 (EC) Instrument 9 (VC) Table 2. Color difference ΔE* ab results of nine instruments according to the Black and White BCRA tiles reference. According to the White tile results shown in Table 2 the instruments of the Family A range from 0.16 to 0.48 ΔE* ab, for the instruments of Family B the differences are even larger, 0.18 to 1.11 ΔE* ab. This again confirms that some of the instruments didn t comply within the calibration on white reference tile supplied by the manufacturer or this could be a consequence of out of date re-certification. On the other hand, the newest and certified 265

277 Paper F instrument (instrument 4) shows a rather high color difference on the White tile (0.87 ΔE* ab ) whereas the instrument 8 performs the least color difference on the White tile although, the re-certification has been expired a long time ago. Similar results with rather small color difference can be seen from instrument 2 (0.22 ΔE* ab ). According to the presented results on the White tile, there is no significant evidence whether an expired certification effect the measurement results. Questions like how frequently an instrument is used and how an instrument is treated and maintained by the operator determine the precision of the instrument. However, the authors recommend an appropriate maintenance of the instrument including regular instrument re-certification to approve the obtained measurements. The poorest performance for almost all instruments results from the measurements of the Red, Orange and Yellow tiles as shown in Figure 6. Except for the instrument 4 the Orange tile produces the largest color difference. Instrument 7 shows the least color differences for these three tiles. Considering the color differences within the product families (instrument family A includes instruments 1-3, and family B includes instruments 6-9) on the tiles Red, Orange and Yellow, there is no obvious trend visible. Figure 6. Color difference of nine instruments according to the BCRA tiles Red, Orange and Yellow reference. Note, it is important to consider the inherent physical properties of the BCRA tiles. According to a previous work by Fairchild and Grum [22] they stated that the BCRA tiles Red, Orange and Yellow can exhibit appreciable thermochromism due to sharp changes in their spectral reflectance curves. Therefore, based on this findings a study by Berns [23] proposed against using the tiles Red, Orange and Yellow unless the temperature of the tiles at the time of calibration was known and this temperature was maintained both at the location where the tiles would be used and during their measurements. However, according 266

278 Paper F to the results shown in Figure 6 there is no clear evidence of thermochromism for the instrument 4 and instrument 7 except for the yellow tile. In contrast, the master instrument demonstrates larger color differences due to possibly generating significant heat in the measuring process. According to Fairchild and Grum [22], it is important to make sure that the temperature of calibration standards remains constant during their use. On the other hand, there have no significant color changes be observed with small temperature changes around room temperature. Another way of examining the measurement distribution is to assess the dispersion of the measurements on the CIELAB a*- b*plane. Figure 7 illustrates the measurements of nine instruments on the Orange tile including the distance to the reference itself. Although the results of all measurements show a rather low accuracy, a relatively high precision of the instruments can be considered due to the measurement dispersion, which lies almost in one quadrant in the CIELAB system. Figure 8 shows the measurement value distribution of all instruments on the Orange tile including reference displayed on CIELAB L*, C* plane. It can be seen that except for instrument 4, the C* color differences comparing to the reference can be considered as rather large. On the other hand, the L* differences can be considered as low. Addressing the product families, it can be noticed that instrument 4 performes best on Orange comparing to the other instruments with different designs. Figure 7. Measurements of nine instruments on BCRA tile Orange including reference displayed on CIELAB a*, b*plane. 267

279 Paper F Figure 8. Measurements of nine instruments on BCRA tile Orange including the reference displayed on CIELAB L*, C*plane. Figure 9 illustrates the spectral reflectance of all nine instruments measured on the Orange tile. It is also clearly noticeable that in the area between 600nm and 730nm the dispersion of the measured spectral reflectance by the instrument is rather large. It can be only speculated what the reasons can be for the large dispersion. One reasonable explanation can be some degree of thermochromism in combination with some white point error due to expired instrument certification. In addition, the reflectance factors of most of the instruments are far below the reference reflectance factor. Finally, it can be noticed that instrument 4 shows the closest spectral reflectance curve to the reference and again confirms the least color difference on the Orange tile as seen in Figure

280 Paper F Figure 9. Spectral reflectance measurements of nine instruments on BCRA tile Orange including the reference. 4.2 Short-term and long-term repeatability Table 3 shows the manufacturer s agreement and the corresponding results in terms of the short-term and long-term measurements. Note, for the long-term repeatability evaluation instrument 1, instrument 6 and instrument 7 were not accessable. Although the manufacturers do not specify any particular measurement agreements for the long-term repeatability it might be obvious that it is the degree to which the instrument makes identical measurements over a long time. Instruments Manufacturer s agreement Short-term repeatability Long-term repeatability Instrument 1 (EC) ΔE* ab 0.02 (standard shift from 10 Fail n.a. Instrument 2 (EC) Pass Pass measurements at 10 sec. interval on white) Instrument 3 (EC) Pass Pass Instrument 4 (VC) ΔE* 94 < 0.2 Pass Pass Instrument 5 (EC) ΔE* ab 0.02 (standard shift from 10 measurements at 10 sec. interval on white) Fail Pass Instrument 6 (EC) Pass n.a. Instrument 7 (EC) ΔE* 94 < 0.1 from 10 measurements at 3 Pass n.a Instrument 8 (EC) sec. interval on white) Pass Pass Instrument 9 (VC) Pass Pass Table 3. Overview over short-term and long-term repeatability performance. 269

281 Paper F Except for the instrument 1 and instrument 5 on short-term repeatability, all instruments perform results, which qualifies them to pass for the short-term repeatability according to the manufacturer s agreement. Note, that the certification for the instrument 1 and instrument 5 has expired. On the other hand, all available instruments have passed the long-term repeatability test. Figure 10 shows the performance of the short-term and long-term repeatability of the instrument 9 and the manufacturer s agreement which is defined with ΔE* with respect to the mean CIELAB value of 10 measurements on white. The x-axis indicates the short-term and long-term repeatability variations whereas on the y-axis the color difference is represented. The closer the horizontal mean-lines (Oct08 E94 Mean and Jan09 E94 Mean) are, the more identical are measurements and hence better the long-term performance can be considered. Figure 10. Short-term and long-term repeatability on white including manufacturer s agreement ΔE* for instrument 9. It can be seen that both the short and long-term repeatability performs almost equally and within the manufacturer s agreement. Furthermore, the graph shows that the largest variations are in the beginning of the measurement sequence. Hence, increasing the number of measurements in the warm up time procedure would increase the total performance of the repeatability for this instrument. It is worth mentioning that instrument 9 has been recently re-certified by the manufacture. Therefore it can be speculated that the recertification of the instrument can be the reason for the excellent performance of the short-term and long-term repeatability. For the instrument 5 the manufacturer reduces the short-term repeatability to ΔE* ab 0.02 units. In Figure 11 it can be clearly observed some minor short-term measurement variations and that the overall repeatability measurements for the instrument 5 exceed the 270

282 Paper F manufacturer s agreement. Hence, instrument 5 does not conform to the manufacturers specifications and need to be re-certified. On the other hand, although the manufacturer is not providing any long-term repeatability specification the long-term repeatability can be considered as very good due to the almost identical measurements between the 10 weeks interval. Figure 11. Short and long-term repeatability on white including manufacturer s agreement ΔE* ab 0.02 for the instrument 5. In our paper, the manufacturer of the instrument 4 has defined the largest repeatability agreement with ΔE* 94 of 0.2. Although the mean measurements are strongly inside the manufacturer s agreement as can be seen in Figure 12, the short-term measurement variations are rather large. The variation is very apparent regardless that instrument 4 is new and recently certified by the manufacturer. On the other hand, the long-term repeatability illustrates almost the same variations. Hence, the long-term repeatability can be considered as acceptable. However, comparing the short-term and long-term repeatability performance with another instrument family the variations are rather large, e.g. the short and long-term variation of the instrument 4 is much larger then the manufacturer s short-term agreement for e.g. the instrument 5. Figure 13 demonstrates the L* versus the measurement number for the three instruments 4, 5 and 9. Although, instrument 5 and instrument 9 show reasonable repeatability performance in L*, the drift in instrument 4 is very apparent, especially when considering the very short time scale of the measurements. It can be speculated in which direction the drift would have continued by increasing the number of measurements. The rather large variations might be explained due to instrument technology and the user interface of the instrument 4. The short-term measurements have been performed with replacement/updown settings, which means that the tile is not left in place at the instrument s aperture when measuring. Furthermore, it has 271

283 Paper F to be noted that the aperture of this instrument is very small, only 2 mm in diameter. The aperture size in combination with the physical measurement settings, replacement/updown may contribute to the obtained measurement variations. Figure 12. Short and long-term repeatability on white including manufacturer s agreement ΔE* for the instrument 4. Figure 13. L* versus measurement number for instrument 4, instrument 5 and instrument Inter-instrument agreement The instrument manufacturers define certain inter-instrument agreements within their instrument families [24]. For the instrument family B (instruments 6-9) the manufacturer 272

284 Paper F has defined an inter-instrument agreement of mean ΔE* 94 of 0.4 and Max ΔE* 94 of 1.0 for single measurement mode on 12 BCRA tiles (D50, 2 ). Figure 14 demonstrates the pairwise contrast of the inter-instrument agreement within the instrument family B. It can be seen that instrument 6, 8 and 9 meet the manufacturer s requirements both in terms of mean ΔE* 94 < 0.4 and Max ΔE* 94 < 1.0. On the other hand for the instrument family A the manufacturer has defined an inter-instrument agreement of mean ΔE* ab 0.3 and Max ΔE* ab 0.8 for single measurement mode on 12 BCRA tiles (D50, 2 ). Figure 15 shows the pairwise contrast of the inter-instrument agreement within the instrument family A (instrument 1-3). It can be seen that even though the direct comparison between instrument 2 and instrument 3 is within the inter-instrument agreement given by the manufacturer regarding max and mean ΔE* ab have been slightly exceeded. The pairwise comparisons between instrument 1 and instrument 2, and between instrument 1 and instrument 3 exceed the manufacturer s requirements distinctly in terms of mean and max ΔE* ab. Instrument 7 exceeds the requirements noticeably in both the mean value and Max value. Figure 14. Pairwise contrast of the inter-instrument agreement within the instrument family B, compared to the manufacturer s specifications. 273

285 Paper F Figure 15. Pairwise contrast of the inter-instrument agreement within the instrument family A, compared to the manufacturer s specifications. 4.4 Inter-model agreement In order to determine the performance of the inter-model agreement instruments from different families have been compared (instrument 2, instrument 4, instrument 5 and instrument 9). Table 4 shows one color difference ΔE* ab for each 14 BCRA sample for each pair of instruments. From each set of color differences, the mean, maximum and RMS color differences and representing standard deviation have been computed. The pairwise contrast where instrument 4 is involved is very apparent with maximum color differences between ΔE* ab 2.76 and ΔE* ab 4.46 in the BCRA tiles Red, Orange and Yellow which again is not surprising having seen the accuracy results previously. On the other hand, looking at the pairwise contrast between instrument 5 and instrument 9 the inter-model agreement can be considered as rather good with a maximum ΔE* ab < 1. In a previous work conducted by Wyble [9] the RMS results from a very similar test using three bidirectional instruments show significant larger color differences. Again, the tiles Red, Orange and Yellow were responsible for the largest color differences. Although, the instrument models are unknown, it can be assumed that the difference between the instrument design was larger compared to the instrument design presented in the present work. The dark to light grey tiles show very similar behavior as already seen previously in the results of the accuracy. Generally, it can be observed that dark achromatic tiles result in a significant better pairwise instrument performance than measurement from chromatic tiles with respect to the inter-model agreement. 274

286 Paper F According to ASTM E2214 [10] the difference between inter-instrument and inter-model agreement can be as large as an order of magnitude which can be confirmed by the presented results with certain pairwise instrument combinations. Note, filter based colorimeter providing tristimulus values only were not part of this study. It can be assumed that a pairwise comparison between instruments obtaining spectral data versus tristimulus data the inter-model agreement can be rather poor. Previously, reports by Rich et al. [25] have reported rather large color differences considering inter-instrument reproducibility. Furthermore, in his article, the inter-model agreement between the colorimeters and spectrophotometers used for emission measurement has shown a very large color difference. 2 vs 4 2 vs 5 2 vs 9 4 vs 5 4 vs 9 5 vs 9 Pale grey Mid Grey Diff Grey Deep Grey Deep Pink Red Orange Yellow Green Diff Green Cyan Deep Blue Black White MEAN MAX RMS STDEV Table 4. Pairwise contrast of instruments using BCRA tiles 275

287 Paper F 4.5 Results of print measurements The following are the results from the measurements performed with seven instruments on three types of substrates (instrument 6 and instrument 7 were not accessible in this task of the work). Firstly, the measurement results on substrate proof will be presented followed by the results for paper type 1 and paper type 5 respectively. To recap, the proof has been created in a commercial printing house, simulating ISO paper type color patches of the UGRA/FOGRA Media Wedge CMYK [17] have been measured three times in sequence and consequently the average were calculated. The mean value (of the three measurements per patch) have been used to calculate the color difference between the target values and each single instrument. The CIELAB target values of the UGRA/FOGRA Media Wedge CMYK are based on print conditions as stated in ISO and the appropriate characterisation tables for different paper types are provided by Fogra [26]. Table 5 shows the calculated color difference values compared with the CIELAB ΔE* ab tolerances according to ISO It can be seen that five instruments (instrument 1, instrument 2, instrument 3, instrument 5 and instrument 8) have performed measurements, which are within the acceptable tolerances. The measurements of the instrument instrument 4 and instrument 9 show results, which are far outside the defined tolerances. At first glance, the verdict might be justified. 276

288 Paper F Substrate Mean Max Primaries Composed grey ΔE* ab 3 ΔE* ab 3 ΔE* ab 6 ΔE* ab 5 ΔH* (ΔE* ab 2,5) ΔH* (ΔE* ab 1,5) C M Y K C M Y Average Instrument Instrument Instrument Instrument Instrument Instrument UV cut Instrument Table 5. Color differences on proof of seven instruments including the CIELAB ΔE* ab tolerances according to ISO :2007 (Orange marked values are outside the tolerance). Although instrument 4 performs satisfactorily for most of the colors, the color difference between the instrument s measurement and the reference on the primary color yellow is ΔE* ab > 7, which is a considerably large color difference. Instrument 9 is the only device, which is using an UV cut filter. Therefore it is obvious that the measurement on the proof substrate exceeds the tolerance due to the concentration of optical brighteners which effects the CIE b* value most (from reference b* -2 to measured b* +4). Looking at the above measured values, if the proof would have been measured initially in the print shop (where the proof is generated) with e.g. the instrument 1 and then measured by the customer with e.g. the instrument 4 or instrument 9 (which contains the UV Cut filter), then, only the measurement performed by the instrument 1 would have been considered as within the tolerance. However, the customer would not have accepted the proof as approved in a first attempt, as the measurements made by his instrument exceed the tolerances. Using an instrument with a UV Cut filter and an instrument without measuring the proof is inappropriate can be considered as an obvious operator error. It has been observed previously that the instrument 4 results in a large color difference in the primary color yellow when compared with the reference. Figure 16, which shows the 277

289 Paper F measurements of seven instruments on proof substrate on the primary color yellow including reference displayed on CIELAB a*, b* plane can confirm this finding. However, looking at the precision of the other instruments, the graph illustrates a very small dispersion of the measurements taken. Furthermore, the instrument family A (instrument 1, instrument 2 and instrument 3) can clearly be recognised as the one with the highest precison. Instrument 9, on the other hand, shows a larger difference in the CIE b* value as seen earlier due to the concentration of optical brighteners in the proof substrate and the measurement with a UV cut filter. Therefore, this large difference can t be considered as a systematic error. Figure 16. Measurements of seven instruments on proof substrate primary color yellow including reference displayed on CIELAB a*, b*plane. Figure 17. Measurements of seven instruments on proof substrate primary color yellow including reference displayed on CIELAB L*, C*plane. Looking at the measurement results on CIELAB L*, C* plane (Figure 17) the precision within the instrument families can be considered as good. Although, the difference in L* 278

290 Paper F value between the instrument 4 and the other instrument families is rather small, the difference in C* value can be recognised as very large. Figure 18. Measurements of seven instruments on proof substrate primary color cyan including reference displayed on CIELAB a*, b*plane. Similar measurement patterns can be observed in other primary (cyan and magenta) and secondary colors (red, green and blue). Figure 18 shows the measurements of seven instruments on proof substrate on the primary color cyan including reference displayed on CIELAB a*, b*plane. Overall, it can be observed that the color difference in CIE b* is larger than on CIE a*. It is apparent that the concentration of the optical brighteners in the proof substrate again has effected the cyan measurement with instrument 9. Although the dispersion of the measurements within the instrument families is slightly larger compared to the primary color yellow, the variations can still be considered as acceptable. It has to be emphasised that in this task the dispersion of the instrument s measurements should be taken into account and not the color difference between the instrument measurements and the reference. Another way of assessing the measurement results on the solid primary colors is by comparing the inter-model agreement. The observation made in Figure 18 can be confirmed with the CIE ΔE* ab values in Table 6 which shows the color differences ΔE* ab on the solid cyan and magenta between each instrument. It is apparent that the instrument 4 and instrument 9 result in the largest color differences on cyan when compared with the other instruments (e.g. Color difference of ΔE* ab 6.12 between instrument 4 and instrument 9). This rather poor inter-model performance has already been observed in Table 4 by the pairwise comparison of the two instruments e.g. on BCRA tile Cyan. The 279

291 Paper F differences between instrument 4 and the other instruments range between ΔE* ab 2.27 and On the other hand, the inter-instrument performance between the instrument family A (instrument 1, instrument 2 and instrument 3) can be considered as acceptable with differences ranging between ΔE* ab 0.6 and Looking at the results of the instrument 5 and instrument 8 the ΔE* ab is less than 0.8. The upper triangle in Table 6 shows the results on solid magenta, where again, the instrument pair instrument 4 and instrument 9 show the largest ΔE* ab of The least color differences are not within the instrument family A (instrument 1, instrument 2 and instrument 3) itself but between instrument 8 and instrument 5 (ΔE* ab < 0.5) and instrument 8 and instrument family B (ΔE* ab < 0.9). Table 7 shows the color differences ΔE* ab on the solid yellow and black. Regarding measurements on yellow again, the instrument 4 shows the most significant color differences compaired to the other devices with ΔE* ab > 5.12 which, already has been seen on CIELAB a*, b*plane in Figure 16. On the other hand, instrument 9 shows a much better precison on yellow than what we have seen on the color cyan and magenta. The instruments performance on solid black, however, shows measurement results, which are almost ΔE* ab < 1.0 across all instrument combinations including instrument 4 and instrument 9. A very similar measurment performance on the BCRA tile black has been observed previously in Table 2. Hence, black seems to be the least critical color considering the precision on inter-instrument and inter-model agreement. Table 6. Inter-instrument agreement on proof substrate in solid cyan (lower left half of the table) and magenta (upper right half of the table) between all instruments. Table 7. Inter-instrument agreement on proof substrate in solid yellow (lower left) and black (upper right) between all instruments. 280

292 Paper F Below the results on substrate paper type 1 according to the ISO standard will be presented. The mean values (of the three measurements) have been used to calculate the color difference between the reference given by ISO paper type 1 (white backing) and each single instrument. The calculated color difference have been compared with the CIELAB ΔE* ab tolerances according to ISO It can be seen in Table 8 that only three instruments (instrument 2, instrument 5 and instrument 8) give measurement results, which will qualify the print as approved due to the values obtained, which are within the printing ISO tolerance values for all primary colors and the substrate. There is evidence of optical brighteners being present in the paper type 1 substrate which affects the CIE b* value when measuring with instrument 9. Therefore using instrument 9 will exceed the measurement value of the substrate above the tolerance value (ΔE b*±2). Moreover, substrates containing optical brighteners affect not only the measurement results on the substrate but also colors in the blue regions when measuring with instrument including UV cut filter. For that reason, the cyan measurement is rather high too. This observation has been confirmed previously by a work conducted by Radencic [27] where he concluded that colors which produce extremely high color differences regardless of the instrument were generally recorded on the substrates containing optical brightener and were generally blue in shade. Instrument 1 and instrument 3 show measurement values on black, which just exceeds the color differences tolerances too, as well as instrument 4 in yellow. Substrate Primaries ΔE L*±3 ΔE a*±2 ΔE b*±2 ΔE* ab 5 C M Y K Instrument Instrument Instrument Instrument Instrument Instrument Instrument Table 8. Color differences on substrate paper type 1 of seven instruments including the CIELAB ΔE* ab tolerances according to ISO (Orange marked values are outside the tolerance). 281

293 Paper F Figure 19 shows the measurements of seven instruments on substrate paper type 1 on the primary color cyan including reference on CIELAB a* - b*plane. Overall, a very similar pattern considering the dispersion of the measurement as seen previously on proof can be observed. Also, the color difference in CIE b* is larger then on CIE a*. Figure 19. Measurements of seven instruments on substrate paper type 1 primary color cyan including reference displayed on CIELAB a*, b*plane. Table 9 shows the color differences ΔE* ab on the solid cyan and magenta between each instrument on substrate paper type 1. It can be recognised that the inter-instrument and inter-model agreement on substrate paper type 1 is almost identical to the inter-instrument and inter-model agreement on substrate proof. The same can be stated for solid yellow and solid black for paper type 1 as seen in Table 10. Table 9. Inter-instrument agreement on substrate paper type 1 in solid cyan and magenta between all instruments. Table 10. Inter-instrument agreement on substrate paper type 1 in solid yellow and black between all instruments. 282

294 Paper F And finally the results on substrate paper type 5 according to the ISO standard will be presented. Table 11 shows the color difference results on substrate paper type 5 of seven instruments. It can be seen that all instruments performed measurements, which will qualify the print as approved due to the values obtained, which are within the printing ISO tolerance values for all primary colors. Although this paper type 5 should not contain any concentration of optical brighteners (as stated by the paper manufacturer) instrument 9 (UV cut) shows a CIE b* value (2.8) which exceeds just the given tolerance. Instrument 5 gives the closest readings compare to printing ISO tolerance values. Figure 20 shows the measurements of seven instruments on substrate paper type 5 on the primary color cyan including reference on CIELAB a* - b* plane. The dispersion of the measurements is almost identical again with the dispersion of measurements seen on substrate proof and substrate paper type 1. The inter-instrument and inter-model agreements on the solid primary colors cyan, magenta, yellow and black on substrate paper type 5 are almost identical to the interinstrument and inter-model agreement on substrate proof and paper type 1 respectively. Substrate Primaries ΔE L*±3 ΔE a*±2 ΔE b*±2 ΔE* ab 5 C M Y K Instrument Instrument Instrument Instrument Instrument Instrument Instrument Table 11. Color differences on substrate paper type 5 of seven instruments including the CIELAB ΔE* ab tolerances according to ISO (Orange marked values are outside the tolerance). 283

295 Paper F Figure 20. Measurements of seven instruments on substrate paper type 5 primary color cyan including reference displayed on CIELAB a*, b*plane. 5. Conclusions Nine commercial spectrophotometers typically used in the graphic art industry were evaluated in terms of precision and accuracy and the effect of color measurement variability was demonstrated. As stated in ASTM E2214 [10] the most important specification in color measuring instrument is repeatability. According to our results, except of two instruments all others did pass the manufacturer s agreement in the shortterm repeatability test. All available instruments did conform to the long-term agreement. On the other hand, as discussed above, there are large differences between inter-instrument and inter-model reproducibility. On one side, the obtained results from the inter-instrument test have demonstrated a reasonable performance among instruments within the same family, as agreed by the manufactures. Nonetheless, the inter-model measurement results have shown, as expected, much larger color differences, especially using instruments from different manufactures. The results of the print measurements did confirm the interinstrument and inter-model agreement observed before by measuring the BCRA tiles. As stated previously accuracy describes the averaging of grouping compared to the centre of a certain true value. Considering the accuracy results, overall, it can be observed that the chromatic BCRA tiles (e.g. Red, Orange and Yellow) produces larger color differences than the achromatic BCRA tiles perhaps due to possible thermochromism. Furthermore, all the instruments demonstrate a smaller color differences for the Black tile than the White tile and the dark to light grey tiles show very similar behaviour. The larger differences on 284

296 Paper F the White tile may indicate that the instruments do not agree very well on the definition of white, which could be traced to the instruments calibration or certification status. However, there has no obvious consistency been observed between certified and noncertified instruments in terms of their performance on the White tile. Consciously, instruments with valid certification and instruments with expired certification have been used in this study to be conforming to the common situation in the printing workflow. I might be expected that instruments with valid certification perform better than instrument with expired certification. Although missing instrument re-certification did not show a significant effect on the measurements it is highly recommended to maintain the instruments according to the manufactures requirement including appropriate recertification procedures. In conclusion, beside of applying only calibrated and certified instruments a further obvious consequence will be the use of only one certain instrument family (same model, same design of instrument from the same manufacturer with the same parameters) in a color managed printing workflow to preserve reasonable color differences. Finally, prevent the use of instruments with different filters (e.g. UV-cut filter) in the same workflow to avoid large errors in measurements. However, in order to improve the colorimetric performance and inter-instrument and inter-model agreement a method of characterizing measurement instruments using colorimetric regression technique has to be considered. It might be of interest to consider other potential directions for further work in the field color measurement uncertainties. The performance of a number of color measurement instruments (and measurement technologies), in particular for emission purposes (display) in terms of precision and accuracy could be evaluated and the possible consequences of the inter-instrument reproducibility in color managed workflow addressed. Another area within the standardization process in the graphic art industry is the viewing condition set up according to the parameters defined in ISO 3664 [28] and ISO [29]. To perform the calibration and to verify the appropriate parameters of the ambient light conditions the same hand-held instruments are used attaching a diffuse light measurement head. Therefore, the precision and accuracy for ambient light measurements have to be investigated. 285

297 Paper F Acknowledgments Firstly, we would like to thank Carinna Parraman, Senior Research Fellow, Centre for Fine Print Research, University of the West of England, Bristol, for providing her instruments to this project. Secondly, we would like to thank Giorgio Trumpy, Institute of Applied Physics of the National Research Council, Italy, for providing his instrument. References [1] ISO , Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 2: Offset printing processes, ISO, 2004 [2] ISO , Graphic technology Process control for the production of half-tone color separations, proof and production prints Part 7: Proofing processes working directly from digital data, ISO, 2007 [3] R. Berns, Billmeyer and Saltzman's principles of color technology: Wiley New York, [4] ASTM E284-08, Standard Terminology of Appearance, American Society for Testing and Materials West Conshohocken, PA, USA, 2008 [5] F. Billmeyer Jr, "Comparative performance of color-measuring instruments," Applied Optics, vol. 8, pp , [6] F. Billmeyer Jr and P. Alessi, "Assessment of color-measuring instruments," Color Research & Application, vol. 6, pp , [7] J. Rodgers, K. Wolf, N. Willis, D. Hamilton, R. Ledbetter, and C. Stewart, "A comparative study of color measurement lntstrumentation," Color Research & Application, vol. 19, pp , [8] D. Wyble and D. Rich, "Evaluation of methods for verifying the performance of color-measuring instruments. Part I: Repeatability," Color Research & Application, vol. 32, pp , [9] D. Wyble and D. Rich, "Evaluation of methods for verifying the performance of color-measuring instruments. Part II: Inter-instrument reproducibility," Color Research & Application, vol. 32, pp , [10] ASTM E , Standard Practice for Specifying and Verifying the Performance of Color Measuring Instruments, American Society for the Testing of Materials, West Conshohocken, PA,

298 Paper F [11] ICC Precision and Bias of Spectrocolorimeters,. ICC white paper 22. Available: pdf,12. September 2008 [12] D. Battle, "The measurement of colour," in Colour physics for industry, R. McDonald. 2nd: Society of Dyers and Colourists, 1997, pp [13] G. Beretta, Spectrophotometer Calibration and Certification vol. 2: HP Laboratories Palo Alto, [14] D. R. Wyble. (2010, 24. August 2010). Color Measurement - Introduction, Historical Perspective, Definitions and Terminology, Components of a Spectrophotometer, Light Source, Detector, Dispersing Element. Available: [15] BCRA, tiles are produced by CERAM Research, Queens road, Penkhull, Stoke-on- Trend, ST4 7LQ, England, [16] ISO13655, Graphic technology Spectral measurement and colorimetric computation for graphic arts images ISO, 1996 [17] U. Schmitt and F. Dolezalek, Ugra/FOGRA Media Wedge CMYK V 2.0, Munich, 2004 [18] Print & Media Forum AG, Altona Test Suite Anwendungspaket, Bundesverband Druck und Medien, 2004 [19] ASTM E308-08, Standard Practice for Computing the Colors of Objects by Using the CIE-System, American Society for Testing and Materials, West Conshohocken, PA, 2008 [20] R. Hunt, Measuring colour, 3 ed.: Fountain Press, [21] CIE15, Colorimetry, CIE Central Bureau, Vienna, 2004 [22] M. Fairchild and F. Grum, "Thermochromism of ceramic reference tile," Applied Optics, vol. 24, pp , [23] R. Berns and K. Petersen, "Empirical modeling of systematic spectrophotometric errors," Color Research & Application, vol. 13, pp , [24] R. Danuser, Standards Laboratory, Personal communication, 2009 [25] D. C. Rich, Y. Okumura, and V. Lovell, "The Effect of Spectrocolorimeter Reproducibility on a Fully Color-Managed Print Production Workflow," in 4th European Conference on Color in Graphics, Imaging, and Vision (CGIV), Barcelona, Spain 2008, pp

299 Paper F [26] A. Kraushaar. Characterization data for standardized printing conditions Available: [27] G. Radencic, "A comparison of spectrodensitometers and their agreement on common graphic arts materials," in TAGA, San Francisco, CA, [28] ISO3664, Graphic technology and photography Viewing conditions, Geneva: ISO [www. iso. org], 2009 [29] ISO12646, Graphic technology Displays for colour proofing Characteristics and viewing conditions, ISO,

300 Paper G Paper G Peter Nussbaum, Jon Yngve Hardeberg and Fritz Albregtsen Regression based Characterization of Color Measurement Instruments in Printing Applications In Electronic Imaging: Color Imaging XVI: Displaying, Processing, Hardcopy, and Application, SPIE Proceedings, 7866, San Francisco, CA,

301 Paper G 290

302 Paper G Regression based characterization of color measurement instruments in printing applications Peter Nussbaum a, Jon Y. Hardeberg a and Fritz Albregtsen b a Gjøvik University College, P. O. Box 191, N-2815 Gjøvik, Norway b Department of Informatics, University of Oslo, P. O. Box 1080 Blindern, N-0316 Oslo, Norway [email protected], [email protected], [email protected] Keywords: Color measurement, color measurement instrument characterization, ISO standards, print quality, process control, polynomial fitting technique, measurement uncertainties, Inter-instrument agreement. ABSTRACT In the context of print quality and process control colorimetric parameters and tolerance values are clearly defined. Calibration procedures are well defined for color measurement instruments in printing workflows. Still, using more than one color measurement instrument measuring the same color wedge can produce clearly different results due to random and systematic errors of the instruments. In certain situations where one instrument gives values which are just inside the given tolerances and another measurement instrument produces values which exceed the predefined tolerance parameters, the question arises whether the print or proof is approved or not accepted with regards to the standard parameters. The aim of this paper was to determine an appropriate model to characterize color measurement instruments for printing applications in order to improve the colorimetric performance and hence the inter-instrument agreement. The method proposed is derived from color image acquisition device characterization methods which have been applied by performing polynomial regression with a least square technique. Six commercial color measurement instruments were used for measuring color patches of a control color wedge on three different types of paper substrates. The characterization functions were derived using least square polynomial regression, based on the training set of 14 BCRA tiles colorimetric reference values and the corresponding colorimetric measurements obtained by the measurement instruments. The derived functions were then used to correct the colorimetric values of test sets of 46 measurements of the color control 291

303 Paper G wedge patches. The corrected measurement results obtained from the applied regression model was then used as the starting point with which the corrected measurements from other instruments were compared to find the most appropriate polynomial, which results in the least color difference. The obtained results demonstrate that the proposed regression method works remarkably well with a range of different color measurement instruments used on three types of substrates. Finally, by extending the training set from 14 samples to 38 samples the obtained results clearly indicate that the model is robust. 1. INTRODUCTION In general, process control is the basic requirement for ensuring satisfactory print and proof quality in the graphic art industry. To preserve the standardization concept colorimetric parameters and tolerance values for print and proof productions are defined in ISO [14] and ISO [15] respectively. Therefore, to ensure the quality control colorimetric values have to be obtained using color measurement instruments. Currently, there are many different models of color measurement instruments used in the printing industry, and this has been found to have significant consequences on print and proof quality [22]. In a modern color managed and standardized printing workflow, most of the printing houses use more than one color measurement instrument, typically one instrument in each department (pre-press, press, and post-press). Moreover, in the context of a Process Standard Offset (PSO) [21] certification process often the same color control wedge of a print or proof is measured first by the instrument of the company, that is to be certified, and secondly with the instrument of the certification body, to determine and confirm whether the colorimetric values are within the defined ISO tolerances. However, measuring a control wedge with two different color measurement instruments will obviously result in different colorimetric data sets due to the nature of the instrument s uncertainties [22]. In a certification context assuming that both instruments give values that are within the given color difference tolerances according to the ISO standard, both measurements will be approved and the print or proof will be accepted. On the other hand, if one of the instruments gives values that exceed the tolerances, the question arises which of the measurement values are correct and which one has failed, even though both measurement instruments are certified. Depending on the applications and the customer s requirements the predefined ISO standard tolerances have been defined narrower to 292

304 Paper G increase the print quality. Consequently, the color measurements performed with more than one instrument are even more critical in terms of the instrument uncertainty. In the past, a number of studies have addressed the issues of color measurement instrument accuracy and uncertainties. For more details on assessment of color measuring instruments in general and inter-instrument reproducibility in particular see the works of Billmeyer [7], Briggs et al. [9], Billmeyer and Alessi [8], Rodgers et al. [25] and Wyble and Rich [29]. In a study by Rich et al. [24] the authors have observed that the differences between pairs of instruments can be quite significant, with maximum differences of up to ΔE* ab of 4.0. In a previous work by Nussbaum et al. [22] the authors conclude that in order to reduce the measurement errors in a color managed printing workflow the use of only one instrument product family (instruments of the same model from the same manufacturer using equal parameters) is recommended. However, due to a number of different reasons this advice seems to be rather difficult to implement in the daily printing production environment. A further technique to reduce the color differences obtained by measuring the same sample using more than one measurement instrument is applying a correction method to the obtained color measurements. Therefore, the aim of the present work is to propose a method to reduce the variations in color measurement performed with more than one instrument measuring the same color target. In particular, the main contribution of this study is in characterizing measurement instruments using a colorimetric regression technique. Finally, the appropriate correction model applied to the measurement data sets will reduce the color errors between the measurements obtained by a master instrument and the measurements performed by a second instrument used (Figure 1). Consequently, the model will improve the colorimetric performance and inter-instrument and inter-model agreement. 293

305 Paper G Un-corrected Measurements by Instrument 1 Correction Corrected Measurements by Instrument 1 Color Control Wedge Before ΔE* ab After ΔE* ab Un-corrected Measurements by Instrument 2 Correction Corrected Measurements by Instrument 2 Figure 1. A schematic diagram of the color measurement workflow using two measurement instruments measuring a color control wedge, and applying a correction model to reduce the errors between instrument 1 and instrument 2. In order to determine the performance of measurement instruments there are a number of parameters to consider. According to ASTM E2214 [3] the most important specification is the repeatability which defines how well an instrument repeats its reading of the same target over a certain period of time. Reproducibility is a form of repeatability in which one or more of the measurement parameters have been systematically changed, such that the target is being different, the time frame of measurements are being very long or the operator has being changed. Inter-instrument agreement describes the reproducibility of two or more instruments of the same design and inter-model agreement describes the reproducibility of two or more instruments of different design. Finally, accuracy describes the conformance of a series of readings to the accepted or true value. The measurement variations between instruments can be divided into systematic and random errors. According to Berns [6], repeatability is affected by random errors including drift, electronic noise and sample presentation. Random variations are difficult to avoid. On the other hand the accuracy is affected by systematic errors, which among other characteristics may be due to different measurement geometry, or detector linearity errors resulting from a change of wavelength. In the past several attempts have been made to reduce the systematic errors by characterizing color measurement instruments. The study by Berns [6] proposes the correction of various systematic errors applying to the spectral measurements data using multiple linear regression based on modeling the results to improve the colorimetric performance. In this work, however, the aim is to correct the instrument s systematic errors by applying a regression technique directly to the measured CIELAB data, and hence improve the inter-instrument and inter-model agreement. 294

306 Paper G Following this brief introduction including the definition of the aim in this work and discussing central references, we provide some more methodology information in Section 2, by illustrating the regression model, defining key concepts, and the experimental procedure including data collection. Then, in Section 3 we present and discuss our results, before concluding in Section METHODOLOGY The method we propose in this work is based on color image acquisition device characterization, which has been applied by implementing polynomial regression with a least square technique [12]. The purpose of the characterization model of a color measurement instrument is to predict color measurement data from a given set of reference data (training set). Essentially, the derived model according to polynomial fitting technique describes the colorimetric relationship between a given sample set of reference data and the corresponding measurements taken by an instrument. Consequently, the derived model is applied to another set of measurements (test set) obtained by the same instrument. The obtained new data set is a corrected version according to the used regression model, as depicted in Figure 2. In this work, for each instrument a separate model has been derived and consequently applied to the test set to correct the measurement data set. It is assumed that by modeling the systematic errors of the measurement instruments the results will improve the colorimetric performance and hence the inter-instrument and inter-model agreement. In other words, the color difference between two corrected measurement data sets will be reduced. BCRA tiles Reference data Training set BCRA tiles Measurements Instrument 1 Regression model Color wedge Measurements Instrument 1 Test set Corrected Measurements Instrument 1 Figure 2. Schematic diagram of the regression method using a training set and a test set. 295

307 Paper G A set of 14 British Ceramic Research Association (BCRA) Ceramic Color Standards Series II (CCS II) ceramic gloss tiles including one Black and one White BCRA tile and printed substrates were measured using six spectrophotometers, according to the procedures outlined by ISO [16]. The instruments used are commercial industrialoriented spectrophotometers typical utilized for daily production control in prepress and press applications. A spectrophotometer measures the ratio of reflected to incident light (the reflectance) from a sample at many points across the visible spectrum [5]. Table 1 presents the instruments employed including the corresponding specifications. Some of the instruments are typically from the same model and some of them represent different models from different manufacturers. According to Nussbaum et al. [22] the instruments used in this study show an acceptable performance in terms of repeatability and reproducibility. Table 1. Overview of the six instruments used in this work and the corresponding specifications. Measuring without replacement mm Measuring geometry Aperture Manufacturer s claimed precision Spectral range and interval Manufacturer s claimed short term repeatability Master instrument Secondary E Secondary A Secondary D Secondary B Secondary C Yes x:0 No x:0 Yes x:0 Yes x:0 Mean ΔE* ab 0.3 Max ΔE* ab 0.8 on 12 BCRA tiles Mean ΔE* 94 < 1.0 on 12 BCRA tiles Mean ΔE* ab 0.3 on 12 BCRA tiles Mean ΔE* Max ΔE* on 12 BCRA tiles 380nm to 730nm at 10nm 380nm to 780nm at 10nm 380nm to 780nm at 10nm 380nm to 780nm at 10nm ΔE* ab 0.02 (Standard shift from 10 measurements at 10 sec. interval on white) ΔE* 94 < 0.2 ΔE* ab 0.02 (Standard shift from 10 measurements at 10 sec. interval on white) ΔE* 94 < 0.1 (From 10 measurements at 3 sec. interval on white) 296

308 Paper G One particular measurement instrument has been defined as a reference and is called the master instrument. It is worth pointing out that the chosen reference instrument is not meant to represent the best or ideal instrument but instead to be a state to which we compare the measurements obtained by other instruments. The corrected measurements obtained from the regression model were then used as the starting point with which the corrected measurements from other instruments were compared to find the polynomial which gives the least color difference. The other five instruments are called secondary s. Note that because one of the manufacturers is requesting not to publish their name the devices were anonymized by identifying the instruments as master instrument, secondary A, secondary B secondary E. 2.1 Experimental procedure To create the characterization model a training set is required. This should consist of a reference data set and the corresponding measurements performed by a measurement instrument. As seen above, ASTM [3] defines accuracy as the conformance of a series of measurements to the accepted value for a given sample. In other words how closely an instrument can conform to a certain reference. In this study the reference values have been provided by the manufacturer of the BCRA tiles. All measurement instruments have performed first a normal calibration procedure according to the manufacturer s recommendations before measuring the tiles 15 times in a sequence. Note that not only the precision between the instruments has to be improved but also the accuracy in terms of the appropriate colorimetric values. Therefore, the BCRA tile reference values, which are traceable, have been used to establish the model. The idea of using the 14 BCRA tiles for the training set is to make the appropriate adjustments in terms of the accuracy. Eventually the derived model has been tested using color measurements of 46 patches of the UGRA/FOGRA Media Wedge CMYK [26] on three different types of printed substrates (Figure 3). The first paper substrate was a hard-copy digital proof print, printed according to the ISO graphic art standards for paper type 1 simulation by a commercial printing house. The second paper substrate was paper type 1 printed by the same commercial printing house aiming at the ISO graphic art standards. And the third paper substrate was paper type 5 Altona Test Suite reference print [23]. The CIELAB target values of the UGRA/FOGRA media Wedge CMYK are based on print conditions as stated in ISO 297

309 Paper G to and the appropriate characterization tables for different paper types are provided by Fogra [19]. Figure 3. UGRA/FOGRA Media Wedge CMYK. 2.2 Polynomial regression Essentially, polynomial regression is needed whenever you have two data sets, which are related and you are aiming to predict one of them if you know the other, typically device dependent colors from device in-dependent color and vice versa. Polynomial device characterization technique with least squares fitting for different application has been adequately explained and applied in a number of studies by Kang [18], Sharma [27], Hong et al. [13] and Johnson [17]. However, a brief description is given to demonstrate this technique applied to color measurement instruments. The first data set, on which the regression model is based, is referred to as the training set and the second data set, which was not involved in deriving the model is used for evaluating it, and is referred to as the test set. As depicted in Figure 2 certain BCRA tiles sample colors are selected and defined as the reference values and the measurement instruments used were measuring the corresponding color specifications. For simplification purposes let s call the BCRA reference values REF and the corresponding measurement values from the instrument INS. Assume that the reference target has N samples. For each color sample the corresponding reference values R, E and F are represented by a 1 x 3 vector p i (i = 1 N) and their corresponding I, N and S color measurement values obtained by the measurement instrument are represented by a 1 x 3 vector x i (i = 1 N). Suppose that only I, N and S values are used in p, the transformation between INS and REF is a simple linear transform. However, the reason for using polynomials is that vector p i can be extended by adding more terms such as I 2, N 2, S 2 etc. which may improve the accuracy of the model in terms of reducing the color differences over all the color samples [13]. While higher order polynomials will give a perfect fit to the data of the training set, this may result in over fitting as well as causing oscillations between the points, Runge's phenomenon [28]. Hence, in this work we applied only second order polynomial even for a 3 x 11 matrix. 298

310 Paper G The polynomials applied and analyzed in this work have the following form: 1. p i = [I N S] 2. p i = [I N S 1] 3. p i = [I N S INS 1] 4. p i = [I N S IN IS NS] 5. p i = [I N S IN IS NS INS 1] 6. p i = [I N S IN IS NS I 2 N 2 S 2 ] 7. p i = [I N S IN IS NS I 2 N 2 S 2 INS 1] Suppose R denotes an 3 x N matrix of vectors p i and X the predicted matrix of vector x i. The mapping from INS to REF can be expressed by X = M*R (1) M is the unknown transformation matrix that determines the accuracy of the model, which means minimizing the color differences over all color samples. The differences between Y and Y can also be expressed as the Sum of the Squares of the Differences (SSD): (2) where Y = MX (3) Depending on the polynomial being solved the size of the matrix M in this work varies from 3 x 3 up to 3 x 11. On the following 1 st order sample it is shown how the model can be derived. Forward model: [R E F] = [I N S IN IS NS INS 1]*M In this case M is an 8x3 transformation matrix that contains the model parameter calculated from the training set by the equation: M = (R T *R) -1 *R T *X (4) where R T denotes the transpose of R, and R -1 denotes the inverse. In this example R is an n x 8 matrix which contains values of the I N S samples as well as corresponding IN, IS, NS, INS and 1 values calculated from them for each sample. X is an n x 3 matrix which contains the number of samples n used in the training set and the columns accommodate R, E and F values of all the samples. 299

311 Paper G The regression model is based on the training set containing 14 BCRA reference values and the corresponding measurements obtained by the measurement instruments from the BCRA tiles. The performance and accuracy of the characterization model has to be evaluated using an independent data test set, which in this work is represented by measurements of the 46 patches (UGRA/FOGRA Media Wedge) on different substrates. The best results with the least color differences are obtained by experimentation. Finally, the most appropriate transformation matrix M is the one that results in the least color difference between the corrected measurements of the master instrument and the corrected measurements of the secondary instruments. Note that in this study CIELAB values are directly used in the characterization and evaluation procedure because CIELAB values have been reported from the spectral reflectance data initially measured by the instruments. Furthermore, the ISO tolerances given in the standards are communicated in CIELAB color space as well. Moreover and most important, Euclidian distance in CIELAB color space is corresponding quite well to the perceptual color differences [12]. 2.3 Data collections All instruments used in this study measured spectral reflectance factor values from 380nm to 730nm with 10nm intervals. Spectral measurements were converted to CIEXYZ tristimulus values according to the CIE observer and the CIE Standard illuminant D50 using the method proposed by ASTM 308, Table 1 [1]. Furthermore, to use a visually meaningful color space CIELAB (D50 as the reference white) values were calculated according to CIE 15 [10] specifications. Consequently, CIELAB data have been used for the regression model and colorimetric difference ΔE* ab values were computed between the master measurement instrument and the secondary instruments. Furthermore, the obtained results will be compared with the ISO tolerances. Because the colorimetric production control tolerances in the ISO standard and ISO standard are defined with ΔE* ab only, no further color difference metrics are used in this work. 3. RESULTS AND DISCUSSIONS As mentioned previously the aim is to find a method to reduce the color difference between instruments measuring the same color patches. Furthermore the applied model shall improve the colorimetric performance and inter-instrument and inter-model agreement on three different types of substrates. 300

312 Paper G Figure 4 shows the color difference results between the BCRA tiles reference values and the master instrument. Moreover, the color difference results between BCRA tiles reference values and the secondary instrument A and between the master instrument and the secondary instrument A. Notice, that the master instrument and the secondary instrument A are not from the same instrument family (which in a practical application very often can be the case). We see that the color difference between the BCRA tiles reference values and the master instrument has the highest values in the red (ΔE* ab 2.8 units), orange (ΔE* ab 4.2 units) and bright yellow (ΔE* ab 2.4 units) tiles. Comparing the BCRA tiles reference values with the secondary instrument A only the bright yellow tile shows a rather high color difference value (ΔE* ab 2.4 units). On the other hand, comparing the measurement results between the master instrument and the secondary instrument A the results on the red, orange and bright yellow tiles again show very large color differences. Moreover, although the master instrument and the secondary instrument A show almost identical color difference compared to the BCRA tiles reference values on the bright yellow tile (approximately ΔE* ab 2.4 units), the direct comparison shows the largest color difference of ΔE* ab 4.5 units. This indicates that the accuracy of both measurement instrument, master instrument and the secondary instrument A on the bright yellow tile can be considered as very similar, However, the color difference between the master instrument and the BCRA tiles reference values and between the secondary instrument A and the BCRA tiles reference values points in different directions. It is important to consider the inherent physical properties of the BCRA tiles. Fairchild and Grum [11] stated that the BCRA tiles red, orange and yellow can exhibit appreciable thermochromism due to sharp changes in their spectral reflectance curves. Based on this finding Berns [6] argued against using the tiles red, orange and yellow unless the temperature of the tiles at the time of calibration was known and this temperature was maintained both at the location where the tiles would be used and during their measurements. However, according to the results shown in Figure 4 there is no clear evidence of thermochromism for the secondary instrument A except for the yellow tile. In contrast, the master instrument demonstrates larger color differences due to possibly generating significant heat in the measuring process. According to Fairchild and Grum [11], it is important to make sure that the temperature of calibration standards remains constant during their use. On the other hand, no significant color changes have been observed with small temperature changes around room temperature. Furthermore, all 301

313 Paper G measurements in this study have been conducted in the same location and the same room temperature conditions. Figure 4. Color difference between the 14 BCRA tiles and the master instrument, BCRA tiles and the secondary instrument A and between the master instrument and the secondary instrument A. The impact of such a systematic error due to the nature of different instrument properties is illustrated in Table 4 which shows the color difference results between the measurements performed by the master instrument and the secondary instrument A. The measurements were conducted by measuring the control media wedge on proofing substrate. Additionally, the table indicates the defined ISO tolerances according to ISO It can be seen that the master instrument gives values which qualify the proof as approved. On the other hand, the secondary instrument A gives values which are outside the given tolerance (orange marked values) and therefore the proof might be not accepted. Moreover, the maximum color difference is clearly seen in the yellow color, which exceed the color difference tolerance of ΔE* ab 5 units. This is perhaps not unexpected due to the large difference in terms of the accuracy performance between the two measurement instruments on the bright yellow BCRA tile, seen previously. Nevertheless, the question may arise, whether the proof is really generated very close to the standard values, or it is just outside the colorimetric tolerance obtained by the two measurement instruments, which varies in terms of accuracy conformance. In a real practical application this situation can be considered as very inappropriate, where due to systematic errors of the measurement instruments the one instrument results in approved and the other one not. Therefore in a first attempt the measurement instrument 302

314 Paper G accuracy has to be ensured by modeling the relationship between the BCRA tile reference values and the actual measurements performed by the measurement instruments. Table 2. Results from the regression using CIELAB values according to 14 BCRA tiles references values and the corresponding values of the master instrument and the secondary instrument A. Trainings data set 14 BCRA tiles Master instrument versus Secondary instrument A BCRA (reference) versus Master instrument BCRA (reference) versus Secondary instrument A Average ΔE* ab Max ΔE* ab Average ΔE* ab Max ΔE* ab Average ΔE* ab Max ΔE* ab Real ΔE* ab Matrices 3 x x x x x x x Table 2 shows the results of the training set with 14 samples using polynomial regression minimizing the difference between reference values and corresponding measurements. As expected, the higher the degree of the polynomial the more reduction in the difference in terms of average ΔE* ab and maximum ΔE* ab. The regression is a function of CIELAB reducing differences in systematic errors to extremely low levels. The derived training functions are used to correct the colorimetric values of the UGRA/FOGRA Media Wedge measured by the master instrument and the secondary instrument A on three different type of substrates. To determine the appropriate function in terms of the least color difference between the corrected measurement data conducted by the master instrument and the secondary instrument A, all calculated training functions have been applied to the test data of the 46 color patches of the UGRA/FOGRA Media Wedge. Figure 5 shows the results (Mean and Max ΔE* ab ) on proof substrate using different correction matrices on a 3-D surface. The horizontal axis and the depth axis 303

315 Paper G indicate the polynomials used for the correction of the master instrument and the secondary instrument A. The vertical axis point out the corresponding corrected color differences. It is important to note, that the uncorrected color difference between the master instrument and the secondary instrument A on proofing paper results in Mean ΔE* ab 2.59 units and Max ΔE* ab 6.92 units respectively. Figure 5. Results from the regression applying to the test set of the master instrument and the secondary instrument A on proofing substrate. Further, it can be seen from Figure 5, that the systematic error of the master instrument and the secondary A can be corrected and the color differences reduced by applying almost all functions to the uncorrected measurement data. The goal is to find a connection between modeling the least color difference and the number of terms in the matrices. As can be seen there are a number of functions, from simple 3 x 3 to 3 x 9 polynomials, which reduce the color differences with more than half compare to the uncorrected data. Although, as seen in Table 2, higher polynomial functions perform excellent in the training set, testing the functions on the test set, the results from the test set indicate that using the higher polynomials (e.g. 3 x 11) on the training set has resulted in over fitting. Therefore, it is generally recommended to use the smallest number of the polynomial terms which adequately fits the function while still smoothing out the noise [4]. Though, there is no single function performing significant best the 3 x 4 or 3 x 5 polynomials can be considered as most appropriate for both the master instrument and the secondary instrument A. To find the relationship of the color differences in terms of different color attributes between the master instrument and the secondary instrument A of the uncorrected and 304

316 Paper G corrected data set, ΔE* ab versus lightness (L*), ΔE* ab versus chroma (C* ab ), and ΔE* ab versus hue-angle (h* ab ) are plotted in Figure 6. Figure 6. Color difference distribution between the master instrument and the secondary instrument A of the uncorrected and corrected data set: plot of lightness vs. ΔE* ab ; plot of chroma vs. and ΔE* ab ; plot of hue angles vs. ΔE* ab. The results of the corrected data set were obtained when a 3x5 matrix was used. Figure 6 illustrates the color difference distribution between the uncorrected and the corrected data set, which gradually affect the color attributes lightness, chroma and hue angle. It can be seen that larger color differences, especially in highly saturated colors, have been reduced significant by the correction model. As expected, the largest color difference in the uncorrected data set is in the yellow color (ΔE* ab 6.92 units). After the correction is applied to the measurements of the master instrument and the secondary instrument A the color difference is reduced to ΔE* ab 2.24 units. To further evaluate the performance of the proposed model the same procedure has been applied for substrate paper type 1 and paper type 5. Moreover, the proposed technique has been used and tested for all the secondary instruments. Table 3 presents the results from the regression and the corresponding color difference between the master instrument and the secondary instruments (A-E) on different types of paper substrates. In addition, the color differences between the master instrument and the secondary instruments (A-E) of the uncorrected measurement data are presented. Except for secondary instrument A the color differences of the uncorrected measurement data are smaller between the other secondary instruments (B-E) and the master instrument. This is due to the different instrument product family. However, in addition to the proofing substrate the model is performing very well in terms of reducing the color differences on the two other substrates. As discussed previously, there is no polynomial which performs significantly best. Very small differences in the results among the functions can be observed. Similar performance 305

317 Paper G patterns in terms of the size of the function can be found from the other secondary instruments (B-E). Note, that the polynomials for both the master instrument and the secondary instrument (A-E) have almost the same size for all three substrates. For the proofing substrates, the applied correction has reduced the color difference both in terms of mean and maximum for all the secondary instruments (A-E). The same is true for the results from the substrate paper type 5. For substrate paper type 1 and the secondary instrument B, secondary instrument D and secondary instrument E the correction model is not performing as good (slightly higher maximum values). Table 3. Results from regression and the corresponding color difference between master instrument and the secondary instruments. Master instrument Type ΔE* ab Un-corrected Polynomial ΔE* ab Corrected versus Substrate Max Mean Master Secondary Max Mean Proofing x 4 3 x Paper type x 4 3 x Secondary A Paper type x 5 3 x Proofing x 5 3 x Paper type x 5 3 x Secondary B Paper type x 5 3 x Proofing x 4 3 x Paper type x 4 3 x Secondary C Paper type x 8 3 x Proofing x 5 3 x Paper type x 6 3 x Secondary D Paper type x 5 3 x Proofing x 4 3 x Paper type x 5 3 x Secondary E Paper type x 6 3 x As indicated previously instruments can be divided into product families which are instruments of the same model from the same manufacturer using equal specifications. Hence, the least inter-instrument color differences can be expected within a product family. In this work, the secondary instrument E is the same model as the master instrument. Regardless of the very small color differences of the uncorrected 306

318 Paper G measurement data, applying the regression model has further minimized the color differences. It is noticeable that, taking the inter-instrument agreement from Table 1 into account which specifies the equipment accordance with mean ΔE* ab 0.3 units and maximum ΔE* ab 0.8 units on 12 BCRA tiles ceramics the proposed method is performing reasonable. On the other hand, the model is not able to handle the instrument s repeatability issues such as drift over time. The secondary instrument B and secondary instrument C are considered as the same instrument model too. However, the secondary instrument C shows rather large color differences, in particular the values given for proofing substrate and substrate paper type 1. The reason is an UV cut filter attached to the instrument which causes the color differences due to the concentration of optical brighteners to affect the CIE b* values in the measurements. Such variations are not considered as systematic errors. Therefore, the applied model is not performing as expected, in terms of reducing the color differences. In the context of quality control using more than one measurement instrument in the workflow, the proposed method can improve the inter-instrument and inter-model agreement significant. Table 4 shows the color differences on proofing substrate according to the uncorrected measurement data of the master instrument and the secondary instruments A-E. Furthermore, the orange marked numbers demonstrate the values, which are outside the tolerances defined by ISO According to the results presented in Table 4 the measurements from the secondary instrument A and the measurements from the secondary instrument C the proof would not be qualified as approved. On the other hand, measurements conducted with the master instrument, the secondary instruments B, D and E the proof would be qualified as approved. Again, the question may arise which of the instrument gives the appropriate results? Note, that the instruments random errors including repeatability performance has been tested in a previous study and concluded as acceptable [22]. 307

319 Paper G Table 4. Color difference results on proofing substrate obtained by six instruments with respect to CIELAB ΔE* ab tolerances according to ISO (Orange marked values are outside the ISO tolerance). Un corrected measurements Master Substrate Mean Max Primaries Composed grey ΔE* ab 3 ΔE* ab 3 ΔE* ab 6 ΔE* ab 5 ΔH* ab 2,5 ΔH* ab 1,5 C M Y K C M Y Average instrument Secondary A Secondary B Secondary C Secondary D Secondary E Table 5 presents the results of all instruments after applying the regression model to the uncorrected data set. Although the secondary instrument A results in values which now qualify the proof as approved it is important to emphasize that it is not the intention of the proposed method to get the values as close as possible to the ISO standard values but to reduce the color difference between the instruments. Furthermore, it can be seen that the secondary instrument C with the UV cut filter still results in values which qualifies the proof far from approved. The variations between the secondary instrument C and the other instruments are large, especially the results obtained on the substrate and the composed grey. As stated previously, this effect of variation is considered as a systematic error and therefore the regression method is not handling this. 308

320 Paper G Table 5. Corrected measurement data set (based on 14 training samples) of seven instruments with respect to CIELAB ΔE* ab tolerances according to ISO training Composed Substrate Mean Max Primaries samples grey Corrected Measurement ΔE* ab ΔE* ab ΔE* ab ΔE* ab ΔH* ab ΔH* ab ,5 1,5 data set C M Y K C M Y Average Master (3x4) Secondary A (3x5) Secondary B (3x5) Secondary C (3x6) Secondary D (3x5) Secondary E (3x3) So far the training set for building the model was limited to 14 samples (14 BCRA tiles). To test if the model will improve the performance in terms of reducing the color difference between the master instrument and the secondary instrument (A-E) the sample number of the training set has been increased with 24 patches from the ColorChecker, which is a color rendition chart including traceable reference values [20]. Consequently, the regression method has been applied again for all the measurement on all three substrates. Although there is no significant improvement in terms of reducing the mean color difference, the maximum ΔE* ab could be reduced substantially in all instrument combinations and all three substrates. Moreover, also functions with second order polynomial (such as 3x11) give reasonable results reducing the color difference significantly, in particular the maximum color difference. This indicates clearly that the model with 38 sample points is more robust. 309

321 Paper G Table 6. Corrected measurement data set (based on 38 training samples) of seven instruments with respect to CIELAB ΔE* ab tolerances according to ISO training Composed Substrate Mean Max Primaries samples grey Corrected Measurement ΔE* ab ΔE* ab ΔE* ab ΔE* ab ΔH* ab ΔH* ab ,5 1,5 data set C M Y K C M Y Average Master (3x4) Secondary A (3x3) Secondary B (3x4) Secondary C (3x5) Secondary D (3x3) Secondary E (3x6) Table 6 shows the results using the training set with 38 samples. It can bee seen that, except for the secondary instrument C on composed grey tolerance, all corrected measurement values obtained will qualify the proof as approved. In other words, increasing the number of training samples in the model and applying the function to the test set (in our case to the measurements of the proofing substrate) will correct the measurements and reduce further the maximum color differences. It is important to note that by increasing the number in the training set, the model behaves more robustly in terms of using higher polynomials in correcting the test data set. 310

322 Paper G Figure 7. Range of variations with respect to uncorrected and corrected measurement data sets from all six instruments on proofing substrates. Another way to describe the range of variation between the uncorrected and the corrected measurements is standard deviation σ. As mentioned previously the aim of the presented work is to reduce the variations related to color differences between measurement instruments in measuring the same color patches. Figure 7 illustrates the range of variations between the uncorrected and corrected measurement data sets (including 14 training samples and 38 training samples) on proofing substrates in terms of the standard deviation (according to the results given in Table 4, Table 5 and Table 6). It can be seen that the applied regression model is reducing the range of color difference variations among the six instruments compared to the uncorrected data sets. In particular the maximum values could be reduced significantly and therefore the mean results have been affected too. Looking at color difference results for the substrate it can be seen that the range of variations between the uncorrected data set and the corrected data set remains almost the same. This is again due to the secondary instrument C with the UV cut filter, as explained above. Except for cyan and black, the correction model based on the training set with 38 samples is performing better compared to the training set with 14 samples. To further verify the proposed method, two other commercial measurement instruments from different manufactures representing different models have been used. The first is considered as a spectrophotometer and the second as a spectrocolorimeter. ASTM E1347 defines a spectrocolorimeter as a spectrometer that provides colorimetric data, but not the underlying spectral data [2]. For the training set, 38 samples (14 BCRA tile and

323 Paper G ColorChecker) including the reference values and the corresponding measurements have been used to derive the model. The test data set has been extended from 46 patches (UGRA/FOGRA Media Wedge CMYK, version 2.2) to the new 72 patches (UGRA/FOGRA Media Wedge CMYK, version 3.0) and the measurements have been conducted on proofing substrates. Table 7 presents the results correcting the measurements of both instruments according to the training set for each device. As can be seen from the results, the color difference between the two instruments has been reduced by approximately 30% in both the mean and maximum respectively. The least difference has been obtained by using a 3x5 polynomial for both instruments and indicating that the applied model performs reasonable. Table 7. Color difference results from the uncorrected and corrected measurements between two different instruments on proofing substrate. Spectrophotometer versus Spectrocolorimeter ΔE* ab ΔE* ab Type Un-corrected Polynomial Corrected Substrate Max Mean A B Max Mean Proofing x 5 3 x It has to be mentioned that only one single measurement with each instrument on the training samples (14 BCRA tiles and 24 ColorChecker) has been conducted. Presumably, averaging multiple measurements per sample will reduce the noise, increase the performance of the model, and further reduce the color differences between the two corrected test data sets from each instrument. 4. CONCLUSIONS It is known that the accuracy and inter-instrument and inter-model agreement of measurement instruments are limited. In this work we have described a method to correct the instrument s systematic errors by applying a regression technique directly onto the measurement output values in the CIELAB color space to improve the colorimetric performance and hence the inter-instrument and inter-model agreement. The study compares different terms of polynomials derived using least-squares regression to determine the appropriate correction for six different measurement instrument s measured on three different types of substrates. One of the measurement instrument used 312

324 Paper G has been defined as the reference instrument. Reference data from 14 BCRA tiles and the corresponding obtained measurements from each instrument has been used to derive a model. The model has been applied to a test set containing 46 measurements from the UGRA/FOGRA Media Wedge on three different substrates. To determine the most appropriate polynomial color differences have been calculated between the corrected measurements of the master instrument the corrected measurement of the secondary instruments. We conclude that first order polynomials (more precise 3x5 polynomial) in most cases produce the best results in terms of reducing the color differences between the instruments on different substrates. Although there is no significant difference in the performance of the model on the three different types of substrates, the proofing substrate results in the least color differences. Moreover, with instruments from different product families the inter-model agreement can be significantly improved by applying the characterization model, reducing the color differences between the measurement instruments by more than 50%. Increasing the size of the training set from 14 to 38 samples is slightly reducing the maximum color differences, but much more important, the model s behavior is more robust in terms of different applied polynomials. To justify whether thermochromism affected the model, the BCRA tiles red, orange and yellow could be left out in the training sample. As seen, the proposed regression method works remarkably well with a range of instruments used on the three types of substrates. However, for future work, the proposed method could by further investigated using different paper substrates (e.g. glossy paper, newspaper) and material (e.g. plastic, textile, aluminum, glass). To improve the performance of the model further extension of the sample number including different types of sample surfaces could be considered to derive the model. Furthermore, the method can be extended and tested on emission measurements using different models of spectrophotometers, spectrocolorimeters and colorimeters. Perhaps, a combination of Bern s [6] proposed method correcting various systematic errors in the spectral domain and the presented technique adjusting the output CIELAB data set may further improve the colorimetric performance and the inter-instrument and inter-model agreement. Finally, considering a relevant application, the proposed model could be implemented into a measurement software system where the correction model is directly applied to the obtained measured values from the instruments. 313

325 Paper G REFERENCES [1] ASTM E308-08, Standard Practice for Computing the Colors of Objects by Using the CIE-System,American Society for Testing and Materials, West Conshohocken, PA, 2008 [2] ASTM E , Standard Test Method for Color and Color-Difference Measurement by Tristimulus Colorimetry,American Society for the Testing of Materials, West Conshohocken, PA, 2006 [3] ASTM E , Standard Practice for Specifying and Verifying the Performance of Color Measuring Instruments,American Society for the Testing of Materials, West Conshohocken, PA, 2008 [4] R. Bala, "Device characterization," in Digital color imaging handbook. vol. CRC Press LLC, S. Gaurav [5] R. Berns, Billmeyer and Saltzman's principles of color technology: Wiley New York, [6] R. Berns and K. Petersen, "Empirical modeling of systematic spectrophotometric errors," Color Research & Application, vol. 13, pp , [7] F. Billmeyer Jr, "Comparative performance of color-measuring instruments," Applied Optics, vol. 8, pp , [8] F. Billmeyer Jr and P. Alessi, "Assessment of color-measuring instruments," Color Research & Application, vol. 6, pp , [9] J. Briggs, D. Forrest, M. Tse, and I. QEA, "Reliability Issues for Color Measurement in Quality Control Applications," in IS&Ts NIP 1998, 14 pp [10] CIE15, Colorimetry,CIE Central Bureau, Vienna, 2004 [11] M. Fairchild and F. Grum, "Thermochromism of ceramic reference tile," Applied Optics, vol. 24, pp , [12] J. Hardeberg, Acquisition and Reproduction of Color Images, Colorimetric and Multispectral Approaches Dissertation.com, ISBN [13] G. Hong, M. Luo, and P. Rhodes, "A study of digital camera colorimetric characterization based on polynomial modeling," Color Research & Application, vol. 26, pp ,

326 Paper G [14] ISO , Graphic technology Process control for the production of half-tone colour separations, proof and production prints Part 2: Offset printing processes,iso, 2004 [15] ISO , Graphic technology Process control for the production of half-tone color separations, proof and production prints Part 7: Proofing processes working directly from digital data,iso, 2007 [16] ISO13655, Graphic technology Spectral measurement and colorimetric computation for graphic arts images ISO, 1996 [17] T. Johnson, "Methods for characterizing colour scanners and digital cameras," Displays, vol. 16, pp , [18] H. Kang, Color technology for electronic imaging devices: SPIE-International Society for Optical Engineering, [19] A. Kraushaar. Characterization data for standardized printing conditions Available: [20] C. McCamy, H. Marcus, and J. Davidson, "A color-rendition chart," J. App. Photog. Eng ; now called The Society for Imaging Science and Technology (IS&T) ), vol. 2, pp , [21] K.-G. Nickel. (2010, November 10th ). Process Standard Offset, German Printing and Media Industries Federation (bvdm) and Graphic Technology Research Association. Available: [22] P. Nussbaum, A. Sole, and J. Y. Hardeberg, "Consequences of using a number of different color measurement instruments in a color managed printing workflow," in TAGA, [23] Print & Media Forum AG, Altona Test Suite Anwendungspaket,Bundesverband Druck und Medien, 2004 [24] D. C. Rich, Y. Okumura, and V. Lovell, "The Effect of Spectrocolorimeter Reproducibility on a Fully Color-Managed Print Production Workflow," in 4th European Conference on Color in Graphics, Imaging, and Vision (CGIV), Barcelona, Spain 2008, pp [25] J. Rodgers, K. Wolf, N. Willis, D. Hamilton, R. Ledbetter, and C. Stewart, "A comparative study of color measurement lntstrumentation," Color Research & Application, vol. 19, pp , [26] U. Schmitt and F. Dolezalek, Ugra/FOGRA Media Wedge CMYK V 2.0,Munich,

327 Paper G [27] G. Sharma, Digital color imaging handbook: CRC, Press, New York, [28] Wikipedia. (2010, retrieved 12. November 2010). Runge's phenomenon. Available: [29] D. Wyble and D. Rich, "Evaluation of methods for verifying the performance of color-measuring instruments. Part I: Repeatability," Color Research & Application, vol. 32, pp ,

328 Paper G 317