POWER SAVINGS Using eecolor

Transcription

1 POWER SAVINGS Using eecolor Rodney L Heckman, Professor Munsell Color Science Laboratory Chester F Carlson Center for Imaging Science Rochester Institute of Technology James R Sullivan, CEO Entertainment Experience, LLC ABSTRACT A key feature of all LCD displays particularly mobile displays is the power usage. Mobile displays are being used increasingly for all types of entertainment media including long form media like television and movies in diverse lighting environments and the usage lifetime is a critical characteristic to consumers. For LCD displays the biggest power drain is caused by the backlight. There have been many attempts to improve lifetime by reducing the backlight brightness and power usage with adaptive techniques that measure the environmental lighting and usage. For environmental lighting the backlight brightness can be decreased in dark or dim lighting significantly without a loss in overall brightness perception and light meters can be integrated into the products to adaptively drive those changes. Also, during usage, the backlight brightness and power can be reduced when the user is not actively viewing with measurable gains in lifetime. This paper adds to those options by introducing the concept of using color processing to boost quality in any lighting and designing the red, green, and blue color primaries to provide increased efficiency and brightness that can then be used to reduce power back to the initial brightness levels. It is well know that when the brightness is reduced in dark and dim lighting, the color is also reduced and there are no methods to improve that until eecolor and the redesigns of color primaries described in this report are accompanied by significant losses in color saturation that can only be recovered with a 3D color visual table like eecolor as will be demonstrated. 1. BACKGROUND and OBJECTIVE eecolor is a 3D color processing technology that uses visual models with completely independent output color design capability to compensate for color losses in different ambient lighting that has been shown to be effective to lighting compensation technical reports, gamut mapping SID 2011 conference by Sullivan and Heckman, Rendering Digital Cinema and Broadcast TV Content to Wide Gamut Display Media :, and chromatic visual adaptation. In this paper it will be applied to power savings by compensating for color losses when display brightness is reduced and for smaller gamut 1

2 displays when color primaries are adjusted to be more efficient, i.e., brighter, for power savings. There are a variety of ways to modify the color primaries of a display to make it more efficient and increase brightness. Since power and brightness are directly related for a backlight in an LCD display those brightness increases can be used directly to reduce power. For example a 100% increase in brightness can be used to decrease power 50%, with the equation PR = (1-1/B), with PR being power reduction and B being ne decimal brightness. For a 100% brightness increase, B=2.0 and PR= 50%. A number of methods to modify the color primaries have been suggested for mobile devices but they have all been difficult to implement because they cause a significant loss in color saturation and color hue errors without methods to restore the colorfulness of the display. eecolor is the key technology of this paper to restore the colorfulness for modified chromaticity color gamuts and desaturated color primaries. eecolor is s 3D color mapping method that uses visual models and colorfulness increases based on visual compensation for ambient lighting losses in colorfulness. Using visual models and compensation for colorfulness loss in various ambient lighting is a preferred approach to restore the colorfulness with color primary changes because the color increases are very natural perceptually since they represent what your eye would see in colorfulness with better lighting. They create smooth increases in colorfulness throughout the full color gamut volume at all brightness levels. In building the 3D table, eecolor also uses a gamut mapping method that avoids loss of detail due to brightness and hue changes that are common in more standard gamut mapping approaches. This is critical because there are significant color gamut issues with modified chromaticities as described by Daly, et al in Gamut Mapping in LCD backlight compensation, 16 th Color Imaging Conference, May 31, The objective of this paper to show that significant brightness increases and power savings can be achieved by modifying the color primaries of a standard backlight LCD display and that the corresponding losses in colorfulness can be restored by eecolor. 2. EXECUTIVE SUMMARY The analyses of this technical report shows the value in using variations in color primaries with the 3D visual color processing method of eecolor to compensate for losses in color saturation and implement chromatic white point adaptation in power savings for displays. The first step in altering the color primaries is to find optimal initial primaries based on optimal block dye filter regions for a standard white LED light source in a backlight LCD application. That result alone is quite useful because it shows that a more saturated green with a filter bandwidth near the visual luminance response is optimal and better in brightness and color gamut than srgb standard filters with this white LED light source. Since it is rare to get both a brightness increase and color gamut increase this is an important result and all additional brightness increases and eecolor color saturation results started from these optimal block dyes. The final results are shown in Table A for the best display design options of this analysis which included optimal block dye filters, adding 10% static white to blue, 20% static white to red and 30% static white to green, using chromatic adaptation to adapt the image data to the resultant white point, processing all gamut mapping and eecolor in IPT color space to restore the color saturation and preserve hue and brightness after adding white. The brightness increases for this case 105% and 114%, which corresponds to a power savings of 51%. With eecolor the calculated colorfulness from the loss in color gamut due to adding white compared to the srgb increased from 22.6 to 43.3 or an increase of 2

3 91%. As discussed later, this colorfulness measure is linear with the colorfulness ranking of users for a large sample set of images. Power savings can be achieved by adding white or open filter regions to each pixel with good color saturation using eecolor and chromatic white point adaptation. Figure T from Section 10 of this report and repeated in this Executive Summary shows the final image simulation result for a very colorful test image restored by eecolor for this 51% power savings. The original srgb image is on the top-left, the image without eecolor processing for 10% white added to Blue, 20% white added to Red and 30% white added to Green is on the bottom-left,and the eecolor processed image for this case on on the middle-right. This is the final result of this analysis. The color restoration is significant over the unprocessed image in Figure U, illustrating the power of eecolor. Although the specific analyses in the report were for backlight LCD displays with a white led light source, the methods apply to any display technology such as Oled, laser, rgb LED display, and displays with inherently large initial color gamuts such as those with nanotechnology color primaries. In each case the specific results will need to be modified depending on the display physics but the simulated results of this report will remain true. The larger the starting color gamut of the display, more power savings can be achieved. All of these analyses were based on adding a static amount of white for each pixel so the loss in color saturation is throughout color space. Future analyses will consider adding a dynamic amount of white for each pixel so that there is a 4 th white controllable subpixel which will allow for adding white only where it is needed in pixels that have a neutral component. This will allow for saturated colors to have no added white and produce higher colorfulness for a given power savings than that included in this report. Comparison of colorfulness values before and after eecolor processing shows that eecolor processing can produce the same overall CieCam02 perceptual colorfulness volume for a color gamut that is greater than 50% smaller in CieLuv. This is how the power savings can be achieved and it also shows that with eecolor processing the cost of an acceptable display can be reduced significantly with a much smaller color gamut. Color Primaries U % Open White Lightness (lux) % Brightness Increase Power Savings Colorfulness Measure without eecolor Colorfulness Measure with eecolor srgb Optimal Block Dyes 10% Blue 20% Red 30% Green % 51% Table A: Comparison of Brightness, Power Savings and Colorfulness for srgb and static white sub pixel amounts using Optimal Block Dyes and chromatic adaptation to the new White Point with differing amounts of added White for each primary 3

4 Figure T: Image examples for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation for U=.1 Blue, U=.2 Red and U=.3 green 3. METHODS AND RESULTS The methods of the paper to increase the efficiency of display color primaries are listed in the next Section 4. These were chosen to be simple changes that are relatively straightforward to implement and applicable to a variety of display technologies, not limited to backlight LCD. More sophisticated changes in panel lighting can be considered as derivatives of these simple changes. All analyses were by image simulation and visual analysis. No actual panel hardware was modified. Various standard optimization methods were used to determine variable settings that optimize brightness and color gamut in CieLuv color space. Brightness increase and colorfulness (defined in Section 5) was calculated for all results to determine the highest brightness increase possible without loss in colorfulness due to applying eecolor transformations. The base light source analyzed was the Nichia NSSW206B white light LED which is quite common in mobile displays. They methods of this paper apply to other panel technologies such as independent rgb light sources or Oled with simple modifications in the optimization analysis. The results show that the brightness can be increased up to 105% without a loss in colorfulness. Image results are 4

5 included to illustrate the affect of eecolor on colorfulness for a variety of sample images. A single image was used that was chosen because it contained most saturated colors which would illustrate the color restoration of eecolor. Results for other eecolor studies have shown that other images show similar results although they were not included to keep the report size manageable. To be as physically close to actual display hardware as possible in these analyses for backlight LCDs, the light spectrum that was assumed for the backlight and all spectral analyses was the white LED from Nichia that is used for many LCD backlight displays shown below 4. COLOR PRIMARY MODIDFICATION METHODS To keep the analysis simple the modification of color primaries considered were (a) Find the optimal block dye filter characteristics that are different than the current panel color primary standard srgb, i.e., 72% NTSC, that provide the most color volume in CieLuv with the highest brightness increase. This is presented in Section 7. (b) Add static white light to the optimal filter set for each rgb pixel to increase britghtness with static meaning that the same amount of white is added to each pixel. This can be accomplished by reducing the size of the rgb filter areas and adding a clear segment for the backlight white light to transmit or by diluting the rgb filters to add broadband white light or by various other methods. This analysis will include the use of eecolor 3d tables to performed chromatic adaption to compensate for any shift in the white point from D65. In so doing, the amount of white light added to rg and b will be different and optimized using image simulation results. This will allow for more power savings than the modification of method b) because it will allow for a white point that is physically slightly different than D65 and brighter with visual adaptation to that white point. This is presented in Section 8. (c) Add static white light to the optimal filter set for each rgb pixel to increase the brightness with static meaning that the same amount of white is added to each pixel but compared to method (b), this method does not force the white point to be D65 but determining the primary concentration values that cause D65 but rather takes advantage of visual chromatic adaptation built into eecolor tables to adapt the image values to the white point that results from the brightness combination of the block dyes with added static white. In this method the gamut mapping and eecolor processing was implemented in a uniform hue color space called IPT to ensure that hue was preserved. This is presented in Section 9. (d) Add static white light to the optimal filter set for each rgb pixel to increase the brightness as in the method of but in this case use different amounts of white for each of red, green and blue to optimize power savings and quality. This is presented in Section 10. (e) Add dynamic white light to the optimal filter set for each rgb pixel to increase brightness with dynamic meaning that the each pixel contains a controllable 4 th white subpixel that can produce a calculable amount of white for each pixel. This allows for less loss of color saturation at a given brightness increase because the method is adding white only where it is needed and not to the most color saturated pixels. The algorithm for determining the white to add to each pixel can be quite 5

6 sophisticated, but the most simple approach was taken in this analysis which is to create a 4 th white subpixel value that is the minimum rgb value, e.g. the neutral vector component of the pixel. In that manner the white is only added for pixels that have a neutral component so that more purely saturated pixels have no added white,subsequently reducing the loss in color saturation for images. This dynamic white method, of course, requires more calculation for the image display and the electronics for adding a 4thcontrollable subpixel, so the additional gains in brightness and power savings must be traded off with that advantage. This analysis is presented as part of future analyses in Section SIMULATON IMAGE ANALYSIS AND COLORFULNESS MEASURE To judge the effectiveness of restoring the lost color saturation for each method, images will be processed with the new color gamut and eecolor and shown in this report. Secondly, a colorfulness measure will be calculated that has been shown to be linear with visual rankings across a large number of image types by the Munsell Color Science (MCS) Laboratory in the Rochester Institute of Technology. This is a critical factor in this analysis because the chromaticity s of the colorants or primaries are clearly being reduced and it is critical to replace that color gamut measure with a measure that is more related to how users see the color quality of the modified panels. The colorfulness measure used in this paper is not a color gamut area, but rather a statistical measure of how colorful a set of images would be to users throughout the full color gamut volume. It has been shown to be a good measure of visual rankings with a nearly linear relationship from a long history of empirical psychophysical data consolidated from many researchers in the field and under many viewing conditions. This was developed by the CIE in CIECAM02 which in turn was based on the work of many people and discussed in Color Appearance Modelling for Color Management Systems CIE Technical Conference 9-01, 2002 by Mark Fairchild and RW Hunt. It is also listed in Wikipedia as a color appearance standard. With 3D methods like eecolor that can boost the color saturation using visual models for midrange color saturation levels, this average colorfulness can be increased for any color gamut. Keys for this colorfulness measure are that it is linear with visual rankings and that it covers all of color space, not just a planar slice showing primary vector projection boundaries. It is expected to much better represent the user s color experience for a system than the chromaticity gamut boundary. This report shows that including eecolor processing for a small color gamut display produces image results that are more colorful than a 100% larger color gamut display. 6. GAMUT MAPPING METHOD AND HUE PRESERVATION The methods of these analyses can result in hue shifts in the resultant eecolor processed images unless the gamut mapping, increase color saturation processing and the addition of white to increase brightness are carried out in a non-uniform hue color space. The gamut mapping component of the lack of hue preservation were reported by Daly, et al in Gamut Mapping in LCD backlight compensation, 16 th Color Imaging Conference, May 31, Others who have attempted to use added white for extra brightness have also faced hue shifts. This is because processing in XYZ chromaticity color space for added white can modify hue significantly because the l hue path in XYZ chromaticity color space is not a straight line between the color primaries and backlight white, in fact it is not a line at all but rather a 6

7 curved path. For this reason all the gamut mapping, white addition and eecolor saturation increases were done in a uniform hue color space call IPT that was discussed by Sullivan and Heckman, Rendering Digital Cinema and Broadcast TV Content to Wide Gamut Display Media and first defined by Fairchild and Ebner at the 1998 CIC Conference Development and Testing of a color space (IPT), with Improved Hue Uniformity. All eecolor tables are defined in this IPT color space. This is a critical part of this analysis that allows eecolor to restore the lost colorfulness with added white without artifacts. 7. OPTIMAL COLOR FILTERS AND DISPLAY CHROMATICITIES The standard display chromaticities or primaries have been unchanged for a number of years. The original set of display color primaries developed for color television is called NTSC. Since they were the original set, all alternative color primaries are reference to NTSC in terms of chromaticity area, again not a good visual measure for typical image data. For example the standard computer, broadcast, bluray and digital camera primary values are now the same. For computers they are defined as srgb and for television CCIR709 which are the same and approximately 72% of the NTSC chromaticity area. Displays will be specified in terms of %NTSC because as mentioned above, It has become a de facto standard for referring to the color capability of a display. The set of additive, block dye primaries are the set of the three theoretical colorants that either absorb or reflect all incident light. They can be simply described by two parameters, λ1 and λ2, the cutoff wavelengths between each of the colorants. Because each their regions correspond to the red, green, and blue region of the spectra, they are denoted as RGB additive primaries. When added together, they produce white. The spectral transmittance of a mixture of such colorants is the sum of spectral transmittance of each of the colorants times their respective concentrations. Equation A: T,mix c R T,R c G T,G c B T,B The Tristimulus values of the mixture are the sum of the product of the spectral power of the illumination, the transmittance of the mixture and the visual color matching functions for XYZ for the degree observer. This is given by, Equation B: X Y Z S,D65 T,mix x S,D65 T,mix y S,D65 T,mix z The first power savings method considered in this analysis was to find the rgb primary chromaticities 7

8 that provide the optimal color volume in CieLuv color space with a D65 white Point and to do that in the simplest fashion using block dye filters to determine the key wavelength regions for these color primaries. There were 3 block dyes, red, green, and blue that are perfectly transmissive in their pass regions and not transmissive in their block regions and the optimization was done on two parameters and 2 which are the cross-over wavelength values between the red-green and green-blue block dyes. For more sophisticated dye shapes like guassian or similar, these can be considered the optimal 50% transmission cross-over points. Figure A shows the optimal block dye regions and crossover points of 470 and 580 nanometers to illustrate the block dye concept. Figure B shows the optimization plots with very specific peak values for the optimal block dye cross-over points. Figure C shows the chromaticity values for these optimal block dyes. These points are very interesting from a visual response perspective. The green block dye has a similar width to the luminance response of human vision. This is not a surprise considering that if the green bandwidth where wider than the luminance response the green region in CieLuv color space would become desaturated and lose volume, and if it were more narrow than the visual luminance response there would be increased green color saturation but a loss in having a single luminance vector to drive the CieLuv color volume and a need to further expand the red and blue to recover the loss luminance which would desaturate those color regions. So the optimal cross-over wavelength results make total visual sense. As Figure F shows they are not dramatically different than srgb which was empirically derived so that is also a reasonable result. Since in this total analysis, we will be adding white that desaturates the color primaries further, it was chosen to start from this optimal block dye filter results rather than srgb. There is also a brightness increase and power savings with these optimal block dyes as shown in table A. Table A also shows the colorfulness measure for these new block dyes compared to srgb primaries illustrating the increase in average colorfulness for a weighted range of image type statistics. Figure D shows the image simulation for these new block dyes compared to srgb with an without eecolor processing showing that these new block dyes with eecolor processing produce a more colorful image result. Although only one image Is shown, the same increase in colorfulness occurs for a broad set of images as demonstrated by the colorfulness measure In table B. More realistic dyes that are not totally block dyes will have a positive effect on these results and will be considered in future work. The positive affect comes from a non constant dye transmittance such as with block dyes will increase the color saturation of the primary thereby allowing more white to be added, and the total light transmittance can be maintained in these shaped dye transmittance responses to maintain the brightness increase. 8

9 Optimized Block Dyes 1 T λ,r 0 λ λ 2 T λ,g T λ,b λ λ 1 0 λ <λ<λ nm λ λ 780 nm 1 2 wavelength Figure A: Optimal Block Dye responses 9

10 Optimizing Gamut Volume - The Base Case - Figure B: Optimization Curves for block dye red-green and green-blue cross-over wavelengths. 10

11 Optimizing Gamut Volume - The Base Case vs srgb Figure C: Chromaticity Diagram for Optimal Block Dyes versus srgb standard Display Chromaticity s Color Primaries Brightness % Brightness Increase Power Savings Colorfulness Measure w/eecolor srgb Optimal Block Dyes % 20% 45 Table B: Comparison of Brightness, Power Savings and Colorfulness for srgb color primaries and the Derive optimal Block Dyes Image results for the standard srgb color primaries and these new block dye primaries are shown in Figure G with the srgb original on the top-left, optimal block dye image without eecolor on the bottomleft and optimal block dye image with eecolor on the middle-right. As the reader can see the color gamut for the optimal block dyes is larger than srgb and the image is more colorful particularly with eecolor. There is a noticeable shift in blue toward a reddish-blue because the IPT space was not used in 11

12 this section for eecolor processing as yet. The use of IPT color processing for eecolor tables will be added in sections 9 and 10 and this reddish blue hue will be eliminated. That is our starting point for all the analyses that follow where brightness is increased by adding white to each pixel to save power and eecolor is use to restore as much as possible of the color saturation loss when adding white. Figure D: Comparison of srgb color gamut, Optimal Block Dye Color Gamut without and with eecolor processing 8. STATIC WHITE SUBPIXEL WITH FORCED D65 WHITE POINT Adding a static white subpixel with forced D65 White point was analyzed with different amounts of added white. The equations that governed the added white were Equation C: XYZ = K Colorant + U(1-k) White Where U was a variable that specified the fraction of additional white added to the rgb primaries and k was a calculated value for red and blue that represents a concentration of the red and blue block dye amounts to generate ad D65 white point for a given U. The combination of U(1-kb) specifies the actual 12

13 amount of added white and is reported for various values of U. In all cases the k value was adjusted to force the amount of white added to green to be 0.0, meaning the blue and red primaries where determined to add to green and create a D65 white. This restriction will be removed in the next section 9 but for this section 8 the blue and red were forced to white amounts that created a D65 white when combined with green, making the analysis highly constrained. The u v chromaticity results for adding static white regions to each pixel with the white forced to D65 are shown in Figure E. Due to the white point being forced to be D65 the decreases in color gamut are significant. Note that the green primary does not change noticeably because it was the reference color that the blue and red concentration were added to for a D65 white, meaning there was no % of open white added to the green primary, u(1-k)=, Table B shows the brightness increases, power savings and colorfulness measures for these static white methods. Figures F-J show the image results with the original srgb on the top-left, image without eecolor processing on the bottom-left and processed eecolor image on the middle-right. The reddish blue tint in the eecolor processed images is because the IPT color processing will not be added until Sections 9 and 10. Much of the lost color is restored particularly given the dramatic change in color gamuts, but more is needed to achieve the goal of power savings near 50% without loss in color. To do that the constraint of a forced D65 white point needs to be removed so that the starting gamuts for eecolor processing are not as small as shown in Figure G. This is presented in the next section 9. Color Primaries Brightness % Brightness Increase Power Savings Colorfulness Measure with eecolor srgb Optimal Block Dyes, % 46 U=0.05 Optimal Block Dyes, % 28% 44 U=0.1 Optimal Block Dyes, % 34% 40 U=0.2 Optimal Block Dyes, % 42% 28 U=0.4 Optimal Block Dyes, U= % 52% 10 Table B: Comparison of Brightness, Power Savings and Colorfulness for srgb and static white sub pixel amounts using Optimal Block Dyes and Forced D65 White Point 13

14 Figure E: Chromaticity Diagrams for Static Added White with force D65 White point and Optimal Dyes for different U values 14

15 Figure F: srgb base Image and static White Added to Optimal Dyes with Force D65 White Point and U=0.05 without and with eecolor 15

16 Figure G: srgb base Image and static White Added to Optimal Dyes with Force D65 White Point and U=0.1 without and with eecolor 16

17 Figure H: srgb base Image and static White Added to Optimal Dyes with Force D65 White Point and U=0.2 without and with eecolor 17

18 Figure I: srgb base Image and static White Added to Optimal Dyes with Force D65 White Point and U=0.40 without and with eecolor 18

19 Figure J: srgb base Image and static White Added to Optimal Dyes with Force D65 White Point and U=0.80 without and with eecolor 9. STATIC WHITE SUBPIXEL WITH VISUAL CHROMATIC ADPATATION TO THE MAXIMUM BRIGHNTESS WHITE POINT The previous section 8 analyzed the increases in brightness and power saving with fixed white added to each pixel primary color, with the fixed amounts for each block dye primary being forced to produce D65 white for all amounts of added static white. This section 9 expands that analysis to remove the D65 white constraint and allow the white point value to move away from D65 for maximum brightness, using visual chromatic adaptation in the eecolor tables to adapt image data to the non-d65 white point. This method results in further brightness increases and power savings over the method in the previous section 8 because it allows the white point to move to a brighter value away from D65. This analysis was done by image simulation, not optimization. This section 9 also includes all processing in the IPT color space to eliminate hue shifts when adding white, gamut mapping or increasing saturation. The amount of visual adaptation allowable requires visual testing with images so this simulation of visual sensitivity to differing white points that are chromatically adapted is the best approach. 19

20 The equation used to determine the amount of white added equally to the red, green and blue pixel prior to chromatic adaptation is shown below. The same U value was added to red, green and blue and a further refinement could be to have different amounts of white added to each pixel color. Figure K shows an example of the white regions in each pixel. Equation D: XYZ = (1-U) Colorant + UWhite Figure K: Example of adding a 30% (U=0.3) open filter white region to each pixel The results for Open Filter white area, brightness, power savings and colorfulness are shown in Table C: which illustrates the increases in brightness and power savings over the the results in Table B. The chromaticity plots for cases analyzed in this section are shown in Figure L for u v and Figure M for xy. The image results are shown in Figures N-Q with the original on top-left, image without eecolor 20

21 processing on the bottom-left and image with eecolor processing on the middle-right. In this section the IPT color space has been used for adding white, doing color gamut mapping and calculating eecolor saturation increases eliminating the reddish blue tints of the previous sections. The color restoration capability of eecolor is shown to be significant in these figures even if there appears to be very little color to begin with as in Figures P-Q. The srgb image and added white images with eecolor processing are similar in colorfulness. A further refinement of this analysis will use different values of U for each RGB color primary to further improve the eecolor restoration in the next Section 10. Color Primaries U % Open White Brightness % Brightness Increase Power Savings Colorfulness Measure without eecolor Colorfulness Measure w/eecolor srgb no eecolor Optimal Block Dyes, % 27% U=0.1 Optimal Block Dyes, % 46% U=0.2 Optimal Block Dyes, % 57% U=0.30 Optimal Block Dyes, % 66% U=0.4 Optimal Block Dyes, U= % 73% 8 12 Table C: Comparison of Brightness, Power Savings and Colorfulness for srgb and static white sub pixel amounts using Optimal Block Dyes and chromatic adaptation to the new White Point 21

22 Figure L: U V Chromaticity Plots for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation 22

23 Figure M: xy Chromaticity Plots for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation 23

24 Figure N: Image examples for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation U=

25 Figure O: Image examples for static white subpixel using Optimal Block Dyes and white point chromatic adaptation U=

26 Figure P: Image examples for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation U=

27 Figure Q: Image examples for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation U=0.40 An interesting plot that shows the strength of eecolor to restore the colorfulness of smaller initial color gamuts is show in Figure R. The Figure R shows that with eecolor the CieCamO2 colorfulness volume can be held fairly constant for CieLuv gamut volumes from 1.22 for an added white or 5% to.53 for an added white of 20%, meaning eecolor can restore the colorfulness for gamut volumes that are more than 50% smaller. 27

28 CieCamo2 Colofulness CieLuv gamut volume without eecolor with eecolor Figure R: Colorfulness/gamut volume ratio with and without eecolor 10. STATIC WHITE SUBPIXEL WITH VISUAL CHROMATIC ADPATATION TO THE MAXIMUM BRIGHTNESS WHITE POINT and different U values for each RGB color The previous two sections used the same value of U for each color primary. That is never going to be optimal for quality and brightness/power savings. Red, Green and blue all play different roles in visual brightness or luminance and they also have different statistical value in images. This section includes a sensitivity analysis of having different U values for each color primary. The results are shown in Figure S using a range of U values from 0.2 to 0.5 for both power savings and the starting gamut volume before eecolor processing in CieLUV color space. The larger the starting color gamut, the less amount of color saturation is needed from eecolor and the higher colorfulness and quality of the images. As the Figure shows as you increase the U value for all colors the power savings increase but the starting color gamut decreases as it should. The Green primary has the most range in both power savings and gamut volume with the power savings ranging from 46% to 64% for U values of The Red and Blue primaries have less range. This is because the Green primary has the most effect on visual luminance. Since there is very little loss in power savings as the Blue and Red U values are decreased and more color gamut, i.e., color quality to begin with for eecolor processing it is advantageous to lower the Blue and Red U values. Figure U shows the resultant image for U=.1,.2,.3 for Blue, red and Green respectively with the original srgb on top-left, unprocessed eecolor bottom-left and eecolor processed middle-right. The restoration of color with eecolor processing is significant with the processed image with eecolor processed image in Figure T being as colorful as the original. Table D shows the brightness, power savings and colorfulness. As Figure T and table D show the color is significantly restored using eecolor with a power saving of 51%. As reference in the Executive Summary section 2, this is the final result of this analysis. 28

29 Figure S: Power Savings and CieLuv Color gamut volume changes for differing U values for each RGB color Primary Color Primaries U % Open White Lightness (lux) % Brightness Increase Power Savings Colorfulness Measure without eecolor Colorfulness Measure with eecolor srgb Optimal Block Dyes 10% Blue 20% Red 30% Green % 51% Table D: Comparison of Brightness, Power Savings and Colorfulness for srgb and static white sub pixel amounts using Optimal Block Dyes and chromatic adaptation to the new White Point with differing amounts of added White for each primary 29

30 Figure T: Image examples for static white sub pixel using Optimal Block Dyes and white point chromatic adaptation for U=.1 Blue, U=.2 Red and U=.3 green 11. FUTURE ANALYSES: DYNAMIC WHITE This report analyzed the addition of static white to each rgb pixel to increase brightness and save power with a fixed amount of white being added for each pixel and primary color. That puts white in all regions of color space equally for each primary including the most saturated color regions that have little initial white or neutral component. As such, it introduces an increased amount of color desaturation compared to using an algorithm that calculates the amount of white to add to a pixel based on the minimum neutral or equal amount of red, green and blue in each pixel. In this dynamic addition of white, the added amount of white is reduced for highly color saturated pixel and selectively used in more neutral, grey-white, color regions. This allows for more white to be added particularly in the white, maximum brightness region of color space to further increase the brightness and power savings with eecolor restoring the color saturation loss. A future analysis using eecolor will consider the dynamic addition of white which is expected to have further significant advantages in power savings and 30

31 image colorfulness 12. SUMMARY AND APPLICATIONS These analyses show that modifications for color primaries can be used to increase brightness and reduce power usage using eecolor even if those modifications include desaturing the colors by adding white. Approximately 50% power savings was achieved with no appreciable loss in color by adding a 20%-30% white or open/clear region equally to the red, green and blue regions of each pixel. This type of power savings approach has never been used before eecolor technology because the losses in color saturation were too significant. This analysis also introduced the colorfulness measure to assess how an average set of images would be ranked by users for colorfulness with a given display capability. That measure rather than display color gamut is recommended as a much better metric for color display quality and it provides results that are consistent with the image examples included in these analyses. Throughout this analysis the colorfulness values for reduced color gamuts were calculated with and without eecolor and the overall results show that with eecolor the same overall image colorfulness can be produced for a color gamut that is 50% reduced. This is significant and allows for the power savings and can also be used to reduce panel design complexity cost and lessen acceptable display variability criteria with individual display calibration. Although the specific display technology considered in these analyses was backlight LCD with rgb filters, the results can be applied to any display technology such as Oled, laser, rgb LED and any display color gamut. In particular, the larger the initial display color gamut such as with gamut expanding technologies such as nanotechnology, Oled, laser or rgb LED, the larger the power savings can become to achieve srgb color quality. 13. ACKNOWLEDMENTS This work was done at the Rochester Institute of Technology, Munsell Color Science Laboratory with all application definitions, sponsorship and funding by Entertainment Experience. LLC. who owns the intellectual property rights to eecolor technology and 3d color tables. 31