Bachelor Graduation Project SOLVING JIGSAW PUZZLES USING COMPUTER VISION


 Justin Lewis
 1 years ago
 Views:
Transcription
1 SOLVING JIGSAW PUZZLES USING COMPUTER VISION AUTHOR : AREEJ MAHDI SUPERVISOR : REIN VAN DEN BOOMGAARD DATE : JUNE 22, 2005 SIGNED BY : Bachelor Graduation Project
2 Solving Jigsaw Puzzles Using Computer Vision Areej Mahdi University of Amsterdam Abstract This paper presents a method to solve jigsaw puzzles by using shape alone. The method uses scans of a real jigsaw puzzle. The pieces of the puzzle are extracted and segmented from the image by using computer vision techniques. The pieces are then each split into four curves, each curve representing a side of the puzzle. The four corners of the pieces are detected using the curvature scale space technique. This technique has not been used in previous work on solving jigsaws by computer. For the local matching the Procrustes Analysis is used. This method is tested on a 24piece puzzle. 1. Introduction Solving a jigsaw puzzle is a very interesting problem both theoretically and practically. By using a computer to solve jigsaw puzzles we come across various known problems in the computer vision field and in matching 2D subcurves. The solution for this problem can be used in solving various practical problems like reconstructing archaeological artefacts or fitting a protein with a known amino acid sequence to a 3D electron density map [1,2]. Yet simply solving a jigsaw puzzle is by its nature a very challenging problem. The very first paper published on solving jigsaw puzzles by shape alone goes back to the year 1964 [3]. Since then many articles have been published on this subject using only the shape of the pieces. Although there are quite a few algorithms introduced to solve jigsaw puzzles, still there is no algorithm known that can solve large jigsaw puzzles in a reliable and efficient way. The purpose of my research is to build a system that solves apictorial jigsaw puzzles from an image containing all the pieces of the puzzle. Computer vision is necessary for solving jigsaw puzzles by computer especially since it mainly depends on the shape of the pieces. Thus an accurate extraction of the information from the image is essential for a good representation of the puzzle pieces and eventually a correct solution of the puzzle. In paragraph 2 related work done on jigsaw puzzle solving will be described as well as the chosen approach. In the paragraphs 3 and 4 we will describe in more detail the setup of the jigsaw puzzle solver and the implementation. The results and experimentations will be presented in paragraph 5. Paragraph 6 contains the conclusions and finally paragraph 7 presents recommendations for future work. 2. Related work In this chapter related work done previously will be described as well as the chosen approach. There are a number of approaches used in the different papers to solve jigsaw puzzles. Generally they consist of the same stages: preprocessing the images of the pieces, matching the sides of the pieces (local matching) and finally matching the jigsaw pieces (global matching). To represent the pieces, most of the published work use polygon approximations of the borders of the pieces. However the way they deal with the puzzle pieces differ, as does their implementation. To represent the boundary of the pieces the following two methods are widely used: the polygon approximation and the Freeman chain code [3]. The advantage of using chain code is that it is more compact to represent. The representation of each point on the boundary costs about 2 to 3 bits per point. Chain code is also invariant to shifting and translation; it is also possible to scale and rotate. Besides the length of the contour is easy to calculate using chain code. As well as calculating the area of a closed loop and checking the contour for multiple loops and closure. In some papers the piece is treated as a sequence of strings. The local shape analysis is based on an approximate string matching [4]. Others transform a piece into a skeleton and apply most of the processing on the skeleton [5]. The rest of the methods use the curves of the pieces themselves to solve the puzzle. A method that worked very well was the algorithm given by Wolfson. It was able to
3 solve a 104piece puzzle [1]. A later work that was based on Wolfson approach was able to solve a 204 piece puzzle. This method differs in how the interior pieces are placed. However the algorithm was not widely tested. It also required many handtuned parameters. In this paper the same overall approach as that of Wolfson will be followed, but some new techniques will be used. The assumptions that have been made in this paper are: the jigsaw puzzle must have a rectangular outside border, each jigsaw puzzle must have well defined neighbours and finally, the pieces must interlock with their neighbours by tabs. The solution of the jigsaw puzzle is done in a number of phases. First the preprocessing step in which the pieces are segmented one by one from the image. Then the border of each piece is extracted and represented. After splitting each piece into four curves each curve is then matched with every other curve. This is done using a local matching algorithm. This will result in a score of how well the curves match each other. A (global) algorithm can then be used to assemble the jigsaw puzzle. This is the general setup of the research. Each of the different phases will each be described in detail. This method will be tested on a 24piece puzzle. 3. Preprocessing Obtaining the pieces The pieces of the puzzle are obtained by scanning the pieces of the jigsaw puzzle. This produces a black and white image containing all the jigsaw pieces (see figure 1). All subsequent processing will be done on a binary image. The pieces are placed in random orientation and position on the scanner. The image contains noise which must first be removed so that all the pieces in the image can be recognized and segmented properly. To remove the noise from the image, morphological operations are applied. The morphological operations openingbyreconstruction followed by a closingbyreconstruction are used to remove the extra noise and fill up the holes in the pieces. The next step is to segment each of the pieces from the image. The pieces of the puzzle in the image lie rather close to each other and in some positions even slightly touch each other. By applying the Watershed Transformation, we are able to segment the image into the areas. Each area contains one of the pieces. The pieces are then stored separately. The pieces have been numbered manually to be able to check if the solution is correct. figure 1: The image obtained from scaning the pieces of the puzzle. Extracting the borders The next step will be to extract and represent the border of the pieces. To extract the border of the puzzle pieces first the inner pixels of the pieces in the image are removed so that only the border is left. Then a contour algorithm is applied to obtain the x,y coordinates of the boundary of each piece. The coordinates are needed to represent the border of the pieces. Other methods to extract the border are available but this approach is both simple and effective. For correctly matching puzzle pieces it is necessary to use a good boundary description method. We will therefore represent the border using the 8connectivity version of the Freeman chain code. Because this also includes diagonal connections. Detecting the corner points The next step is to split each piece into four curves so that they can be matched later on. It is therefore necessary to extract the four corners from each piece. For the detection of the corners we use the curvature scale space. The curvature scale space is a multi scale representation of planar curves. It describes curves at varying levels of details, which makes it possible to detect relevant information at an appropriate scale. The curve is defined in terms of the first and second derivatives of the boundary coordinates x(u), y(u), where u is the length of the path along the
4 curvature. To express the different levels of details, the boundary coordinate functions are convolved with the Gaussian function g(u,σ) producing Gaussian smoothed coordinates X(u,σ) and Y(u,σ) [6]. where The curvature is then calculated according to the following formula: where filtered out. But since some pieces contain rather large corner angles, these angles are lost at high scales. Therefore a method must be used to select the real corner points. The method used in this paper is the first to combine possible corner points found at two different scales. The chosen scales both have to contain all four true corner points. By comparing the points obtained at the two scales, some false corner points can be eliminated. Corner points that appear at one scale and disappear at the other are obviously not true corner points, so they can be deleted. Possible corner points that almost have the same position on the curve on both scales are considered to be the same point. It is not easy to know which point is a better candidate and one does not want to accidentally delete the best corner point. So the two points are replaced with a point that lies between them. The real corner points do not lie close to each other, so possible corner points that lie close to one another are examined. The point with the lowest curvature is deleted. Although this eliminates many false corners, the number of possible corners is still greater than four. To choose the corner points we take the three points with the highest curvature. These points are used to construct a parallelogram, so that the fourth corner point lies at nearly the same distance regarding the other three points. 4. Solving the jigsaw puzzle The detected corner points are used to split the chain code for each piece into four subchain codes. The subchain codes will be matched to each other in the local matching algorithm. Corners of the pieces can be detected due to the high curvature at the corner points. However other interesting points in a piece can also be detected using the curvature scale space. For example points that correspond to the inflection points of the tabs in a piece because these points also contain high curvature. Unfortunately the pieces of the puzzle contain very ridged boundaries and curves created due to the noisy image. This results in lots of high curvature points which are neither corner points nor points of interest in the piece. It is therefore important to choose an appropriate scale which results in the least number of found points with high curvature yet still containing all the actual corner points of a piece. If the scale of the curvature scale space is higher, then it produces fewer points. By smoothing the curve more noisy points will be Local Matching Algorithm To match the pieces we have used the Procrustes Analysis. This is a method used to compare the shape of objects. The shape of the objects must be described by a finite set of points. The Procrustes Analysis determines the linear transformation: translation, reflection, orthogonal rotation and scaling of the points in matrix X of an object to best fit the points in matrix Y of another object so that [8]: Y = s*x*r + 1 t Where s is the scale factor, R is the rotation and reflection matrix, t the translation vector and 1 is a matrix of ones.
5 The matrices X and Y must have the same number of points. Therefore the curves that we are trying to match must be resampled so that their lengths correspond to each other. If the curves have different curve lengths then the curve with a shorter length is repeatedly matched with the longer curve; each time the smallest curve is matched with a part of the long curve. The goodnessoffit criterion is the sum of squared errors. By applying the Procrustes Analysis a matrix is obtained that minimizes the value of the dissimilarity measure. It is a 4Nx4N matrix where N is the number of pieces. If the dissimilarity value is smaller than a specified threshold then the match is good. To prevent performing matches with straight sides, the straight sides are recognized before they are matched with the rest of the other curves. Global matching Based on the scores of the local matching a global algorithm can be implemented to assemble the pieces. It is therefore essential that the local matching algorithm provides good matches. For the assembly of the pieces, we can divide the pieces into two types: frame pieces and interior pieces. A frame piece is a piece that contains one or more straight sides. The frame pieces are assembled first so that they can be used to fit the interior pieces into the frame. The assembly of the frame pieces is done by matching the frame pieces with each other so that they form a cycle. The cycle should contain each piece only once and has the smallest dissimilarity matches between the pieces. This is the wellknown Travelling Salesman problem. Once a cycle is found, we can match the remaining pieces in this frame. One can use the same approach as Wolfson [6]. 5. Experiments and results All the code was implemented in Matlab. The algorithm was tested on a 24piece jigsaw puzzle and it meets the requirements set out by the assumptions made in paragraph 2. The image of the jigsaw puzzle was obtained by laying the pieces upside down on their backside on a flatbedsurface scanner at a resolution of 300 dpi. The pieces were placed at random positions and orientations with no overlap. After the segmentation and representation of each piece, the corners of each piece are determined. The results for the experiments are show in table 1. This table contains the number of real corner points that are detected at each step of the experiment. The pieces are all given an id number according to their location in the puzzle. The puzzle pieces are numbered from left to right, top to bottom. This makes it easy to know with which piece we are dealing with. To detect the corners the curvature scale space is used at scale 18 and 30. Scale 18 is the highest scale that can be chosen where all four real corner points can be found for all the pieces. At scale 30 however some real corner points are lost for some of the pieces. This is the case for pieces number 4 and 5. These pieces contain corner points with a rather large angle, which are hard to recognize because of the low curvature. By using a high scale of 30 the detection becomes even more difficult. Nevertheless this scale is chosen because it provides a sufficient level of detail so that the corner points can be easily found and it works for nearly all the pieces. The results of scale 18 and 30 are combined to obtain one set of points. This set will be narrowed down throughout the experiments. After applying the heuristics, described in paragraph 3, the four corners of 19 of the pieces are still among the points found with high curvature (see table 1). However when the ten points with the highest curvature for each piece are selected, all the corners of only fourteen of the pieces are among these points. For four of the pieces the number of real corner points in the top ten is two and for the remaining pieces only three real corners. This is because some of these pieces contain corners with rather big angles, so the real corner points do not have the highest curvature when compared to other points in the same piece. When the four points with the highest curvature are taken as the possible four corners of a piece, then it only successfully identifies one piece. While by taking the top 3 highest curvature points, the corners of eight of the pieces are correctly detected. This produces better results than the top 4 because the method used to select the fourth corner chooses the closest point that forms a parallelogram together with the other three found corners. The expected position of the fourth point is calculated and displayed as a circle in the figures. The distance between the remaining candidate points and the expected position is calculated. The point with the least distance is selected as the fourth. For one of the eight pieces this happened to be a point other than the real corner point, see figure 2. As can be seen in the figure the true corner point lies further from the expected position than the point chosen as
6 the fourth corner. For pieces that have less than three true corner points as the points with the highest curvature, this results as expected in a wrong estimation of the fourth corner point, see figure 3. The detection of the corners works correctly for seven pieces, see figure 4. These pieces are emphasized in the table. From these seven pieces, six are connected to each other in the puzzle at five different places. figure 4: presents piece with succesfully detected corners The red crosses are the points found in the top 3. The circle marks the estimated location of the fourth corner and the green cross marsk the selected point. figure 2: contains a jigsaw piece that has three succesfully detected corner. The selection of the fourth corner however is incorrect. The red crosses are the points found in the top 3. The circle is the estimated location of the fourth corner and the green cross marks the selected point. The reason that the corner points of some pieces get eliminated so easily at the first rounds is because in these pieces the starting point of the curve is very close to the corner. A wrap around of 5 points is used. But it is difficult to say how long the wrap must be. If the wrap is too large, then incorrect calculations will be made, for example the calculations of the chain code and the curvature of the piece. For points of which the starting point is far from the corner points the real corners can be detected, this is the case for the eight pieces which were successfully found in the top 3 experiment. The starting point used is always on the topside of the piece. Because the pieces were randomly placed on the scan platter, this starting point sometimes just happens to lie rather close to a corner point. So it is not always possible to know what the best wrap around value is because each piece has a random orientation. From experimenting with the data, it became clear that if another starting point is chosen, the four corners are successfully detected for pieces other than the previous eight. figure 3: contains a jigsaw piece that has only two succesfully detected corner points. The red crosses are the points found in the top 3. The circle marks the location of the estimated fourth corner the green cross marks the selected point. Piece ID Top 10 Top 4 Top 3 4 th point estimation
7 Table 1: contains the results of all the experiments performed to determine the corner points of each piece, the numbers in the table refer to number of real corner points that are detected at each experiment. Cornerpoints that are succesfully detected for a piece are emphasized. The matching algorithm produces a dissimilarity matrix. A match is considered to be successful if the dissimilarity value is above a certain threshold value. The optimal threshold value 0.03 is used and has been determined by trial and error. Table 2 contains the results of the matching algorithm. The columns contain the piece id and the side of the piece, which is successfully matched with the side of the piece in the second column. The correct matches are emphasized in the table. The local matching algorithm produces nine matches of which three matches are correct, see table 2. As is obvious the results of the local matching contained a number of false positives matches. Three of these matches are impossible matches, between two indents or two outdents. Specifying whether a curve is an indent or outdent can probably lead to reducing the false positive matches. Matched Pieces Table 2: contains the results of the local matching algorithm. The piece and curve numbere in the first column are considered as a successful match with the piece and curve in the second column. The correct matches are emphasized. To prevent the algorithm from performing impossible matches such as matching a straight side of a piece with other sides, the straight sides in the puzzle must be recognized. A straight side of one of the frame pieces is matched with every side of the other pieces. If the match is below a certain threshold value then the side is considered a straight side and is not matched with sides of other pieces. The threshold value is set on This method successfully recognized all frame sides of the pieces. This can be later used in the global matching algorithm so that only frame pieces are matched at first. The matches produced by the local matching are in fact good, considering the fact that the images are at low resolution and the curves are very noisy. Unfortunately due to time limitations, the global matching has not been implemented. 6. Conclusion A method to solve jigsaw puzzles has been presented which used some new techniques not used by previous work on solving jigsaw puzzles such as the curvature scale space. The curvature scale space is applied to detect the corners of each piece. It succeeded to detect all the corners for some of the pieces. However some improvements must be made to fully make use of potentials of these methods. Better results are probably possible if more scales are combined together to detect the corner points. Also applying more heuristics may help. Furthermore the pieces on which the curvature scale space was used contained very noisy curves. The implemented local matching algorithm was successful in finding three out of the five matches correct matches. However the algorithm also returned a number of false positive matches. A number of these matches are impossible ones, so a strong suggestion is to prevent such matches from happening. This will surely enhance the performance of the matching algorithm. 7. Future work It is possible to obtain better results by applying some changes. Finding the correct corner points of the jigsaw pieces can be improved. In the paper we used the curvature scale space at two different scales and the results were combined. By using
8 more than two scales it is possible to obtain more accurate findings, which lead to a better detection of the corner points. It became clear from the experiments that choosing another starting point, which is not very close to a corner, will result in better corner findings. Furthermore the method in which the final corner points are chosen can be enhanced. This method first chooses three of the possible corner points with the highest value as three of the final corners. Applying more heuristics before this method is used, may lead to better results. So that the remaining points with a highest curvature are for more pieces indeed the corner points. The way the fourth corner point is determined can also be improved. The matching of the curves can produce a better score if we specify whether a curve that we are matching is an indent or an outdent. This will eliminate many impossible matches such as matching two indents together. This will lead to a more accurate matching algorithm and a better score. Because the curves are so noisy, smoothing them a bit before applying the matching algorithm may also help improve the matching results. The images used in this paper are of relatively low resolution. This resulted in lots of noise in the image and curves. Another possible reason for the noisy curves is that the jigsaw used in the development of the method described in this paper is handmade from wood. Therefore the boundaries of the pieces are not entirely smooth. Jigsaw Puzzles in Computer Vision. IEEE Transactions on Systems, Man, and Cybernetics, vol. 21, no. 5, [6] H. Wolfson, E. Schonberg, A. Kalvin and Y. Lamdan. Solving jigsaw puzzles by computer. Annals of Operations Research 12 (1988), [7] F. Mokhtarian, K. Mackworth. A Theory of Multiscale, CurvatureBased Shape Representation for Planar Curves. IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 14, no. 8, [8] I. Borg and P. Groenen. Modern Multidimensional Scaling: Theory and Applications. Springer Series in Statistics (1997) References [1] D. Goldberg, C. Malon and M. Bern. A Global Approach to Automatic Solution of Jigsaw Puzzles. SoCG 02, June 57, [2] W. Kong and B.B. Kimia. On solving 2D and 3D puzzles using curve matching. Proc. IEEE Computer Vision and Pattern Recognition, [3] H. Freeman and L. Gardner. Apictorial jigsaw puzzles: The computer solution of a problem in pattern recognition. IEEE trans. on Electronic Computers 13 (1964) [4] H. Bunke and G. Kaufmann. Jigsaw puzzle solving using approximate string matching and bestfirst search. In Computer Analysis of Images and Patterns, Chetverikov and Kropatchs. ed., Springer (1993) [5] R.W. Webster, P.S. LaFolette and R. L. Stafford. Isthmus Critical Points for Solving
9 Solving Jigsaw Puzzles using Computer Vision Name: Areej Mahdi Supervisor: Rein van den Boomgaard Date: July 1, 2005
10 Overview Introduction Related work Preprocessing Local Matching Global Matching Discussion Conclusions Questions
11 Introduction The goal of this project: Building a system that solves jigsaw puzzles by shape alone from an image containing all the pieces of the puzzle
12 Interesting problem.. Reconstruction of archeological artifacts Art restoration Fitting a protein with known amino acid sequence to a 3D density map
13 Related Work (1) First article written on this subject goes back to 1964 Different approaches generally consists of the same phases Wolfson s approach
14 Related Work (2) Succeeded in solving 104piece puzzle Divided border into four sub curves corresponding to the four sides Local SchwartzSharir matching algorithm Global matching Overall approach of Wolfson but applying new techniques
15 Jigsaw Puzzle Assumptions: Rectangular outside border Four well defined neighbors Pieces interlock with their neighbors by tabs
16 Jigsaw Puzzle Assumptions: Rectangular outside border Four well defined neighbors Pieces interlock with their neighbors by tabs
17 Preprocessing: Obtaining the pieces Scanning the pieces Removing the noise Segmenting the image
18 Preprocessing: Extracting the pieces Scanning the pieces Removing the noise Segmenting the image
19 Preprocessing: Extracting the pieces Scanning the pieces Removing the noise Segmenting the image
20 Preprocessing: Extracting the border Scanning the pieces Removing the noise Segmenting the image
21 Preprocessing: Extracting the border Scanning the pieces Removing the noise Segmenting the image
22 Preprocessing: Representing the border Borders are represented by Chain Code Chain code describes relative step from previous pixel 8connectivity Chain Code: 2,1,0,7,7,0,1,1
23 Preprocessing: Dividing the border Finding the corners of the pieces: divide the border into 4 subcurves Use Curvature Scale Space
24 Curvature Scale Space (1) Mokhtarian Describes curves at varying levels of details Detection of relevant information at a appropriate scale It is invariant under rotation, uniformed scaling and translation Very suitable for noisy curves
25 Curvature Scale Space (2) Curve defined in terms of derivative of the boundary coordinates x(u), y(u), where u is the length of the path along the curve To express the different levels of details the boundary coordinate functions are convolved with the Gaussian function g(u,σ) producing Gaussian smoothed coordinates X(u,σ) and Y(u,σ)
26 Curvature Scale Space (3) The curvature of a curve is calculated as follows: where
27 Piece nr. 11
28 Piece nr. 11 Scale: 18
29 Piece nr. 19
30 Piece nr. 19 Scale: 18
31 Piece nr. 10 Scale: 18
32 Detecting the corners (1) More than four points are detected with high curvature even at high scales
33 Detecting the corners (2) Combining different scales to eliminate false corner points
34 Detecting the corners (3) Choosing the top 3 highest candidate corners and constructing a parallelogram from these points
35 Detecting the corners (3) Choosing the top 3 highest candidate corners and constructing a parallelogram from these points
36 Piece ID Top 4 Top 3 4 th point estimation
37 Detecting the corners (5)
38 Detecting the corners (5)
39 Detecting the corners (5)
40 Local Matching The corner of seven pieces are successfully detected Local matching is applied only to these pieces
41 Local Matching (1) Procrustes Analysis: Compares shapes of objects Determines linear transformation of matrix X to best fit into Y: Y = sxr + 1 T t where s is scale factor, R the rotation and reflection matrix, t the translation vector
42 Local Matching (2) Matched curves must have same lengths The minimized value of the dissimilarity measure between two curves is calculated Each curve of a piece is matched with all curves of the remaining pieces This produces a 4N x 4N matrix Straight sides of pieces are detected to prevent them from being matched with other pieces
43
44 Global Matching Same approach as Wolfson: Frame assembly Traveling Salesman Problem Interior assembly
45 Discussion Detection of the corner points: Higher resolution of jigsaw image Combining more Curvature scales Applying better heuristics to select the four corner points Local Matching: Eliminating false positive matches: indents and outdents
46 Conclusion New combination of techniques works, but lots of room for improvement Better results possible after applying suggestions to current approach
47 Questions??
48
49
50 Local Matching (3) Result table of found matches Matched Pieces
51
52 Detecting the corners (1) More than four points are detected with high curvature even at high scales
53 1 <image of curvature and X & Y in a 3D plot>
Assembling Jigsaw Puzzles
Assembling Jigsaw Puzzles Aric Bartle Tianye Lu Abstract We propose an improved method for assembling real world jigsaw puzzles. Our approach uses both shape and color measures to determine piece compatibility.
More informationA Global Approach to Automatic Solution of Jigsaw Puzzles
A Global Approach to Automatic Solution of Jigsaw Puzzles David Goldberg 1, Christopher Malon 2, and Marshall Bern 1 1 Xerox Palo Alto Research Center 3333 Coyote Hill Rd., Palo Alto, CA 94304 USA goldberg@parc.xerox.com
More informationBinary Image Analysis
Binary Image Analysis Segmentation produces homogenous regions each region has uniform graylevel each region is a binary image (0: background, 1: object or the reverse) more intensity values for overlapping
More informationDepartment Of Computer Science, Mewar University, Rajasthan, India
Volume 5, Issue 5, May 2015 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Standard Jigsaw
More informationA Chain Code Approach for Recognizing Basic Shapes
A Chain Code Approach for Recognizing Basic Shapes Dr. Azzam Talal Sleit (Previously, Azzam Ibrahim) azzam_sleit@yahoo.com Rahmeh Omar Jabay King Abdullah II for Information Technology College University
More informationMETHODS TO ESTIMATE AREAS AND PERIMETERS OF BLOBLIKE OBJECTS: A COMPARISON
MVA'94 IAPR Workshop on Machine Vision Applications Dec. 131 5, 1994, Kawasaki METHODS TO ESTIMATE AREAS AND PERIMETERS OF BLOBLIKE OBJECTS: A COMPARISON Luren Yang, Fritz Albregtsen, Tor Lgnnestad and
More informationFourier Descriptors For Shape Recognition. Applied to Tree Leaf Identification By Tyler Karrels
Fourier Descriptors For Shape Recognition Applied to Tree Leaf Identification By Tyler Karrels Why investigate shape description? Hard drives keep getting bigger. Digital cameras allow us to capture, store,
More informationReconstructing shredded documents through feature matching
Forensic Science International 160 (2006) 140 147 www.elsevier.com/locate/forsciint Reconstructing shredded documents through feature matching Edson Justino a, Luiz S. Oliveira a,b, *, Cinthia Freitas
More informationTracking Moving Objects In Video Sequences Yiwei Wang, Robert E. Van Dyck, and John F. Doherty Department of Electrical Engineering The Pennsylvania State University University Park, PA16802 Abstract{Object
More informationConstructing Problems of Geometric Combinatorics
Constructing Problems of Geometric Combinatorics Daniel S. Hirschberg Dept. of Computer Science Univ. of California, Irvine Irvine, CA 997 USA dan@ics.uci.edu Abstract Global information is used to simplify
More informationAutomated Sudoku Solver. Marty Otzenberger EGGN 510 December 4, 2012
Automated Sudoku Solver Marty Otzenberger EGGN 510 December 4, 2012 Outline Goal Problem Elements Initial Testing Test Images Approaches Final Algorithm Results/Statistics Conclusions Problems/Limits Future
More informationComponent Ordering in Independent Component Analysis Based on Data Power
Component Ordering in Independent Component Analysis Based on Data Power Anne Hendrikse Raymond Veldhuis University of Twente University of Twente Fac. EEMCS, Signals and Systems Group Fac. EEMCS, Signals
More informationColorCrack: Identifying Cracks in Glass
ColorCrack: Identifying Cracks in Glass James Max Kanter Massachusetts Institute of Technology 77 Massachusetts Ave Cambridge, MA 02139 kanter@mit.edu Figure 1: ColorCrack automatically identifies cracks
More informationAn Optical Sudoku Solver
An Optical Sudoku Solver Martin Byröd February 12, 07 Abstract In this report, a visionbased sudoku solver is described. The solver is capable of solving a sudoku directly from a photograph taken with
More informationModelbased Chart Image Recognition
Modelbased Chart Image Recognition Weihua Huang, Chew Lim Tan and Wee Kheng Leow SOC, National University of Singapore, 3 Science Drive 2, Singapore 117543 Email: {huangwh,tancl, leowwk@comp.nus.edu.sg}
More informationSubspace Analysis and Optimization for AAM Based Face Alignment
Subspace Analysis and Optimization for AAM Based Face Alignment Ming Zhao Chun Chen College of Computer Science Zhejiang University Hangzhou, 310027, P.R.China zhaoming1999@zju.edu.cn Stan Z. Li Microsoft
More informationMarkov chains and Markov Random Fields (MRFs)
Markov chains and Markov Random Fields (MRFs) 1 Why Markov Models We discuss Markov models now. This is the simplest statistical model in which we don t assume that all variables are independent; we assume
More informationSignature Region of Interest using Auto cropping
ISSN (Online): 16940784 ISSN (Print): 16940814 1 Signature Region of Interest using Auto cropping Bassam AlMahadeen 1, Mokhled S. AlTarawneh 2 and Islam H. AlTarawneh 2 1 Math. And Computer Department,
More informationSbLRS: Shape based Leaf Retrieval System
SbLRS: Shape based Leaf Retrieval System Komal Asrani Department of Information Technology B.B.D.E.C., Lucknow, India Renu Jain Deptt. of C.S.E University Institute of Engineering and Technology, Kanpur,
More informationShape Descriptors for Nonrigid Shapes with a Single Closed Contour
IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 424429, 2000. Shape Descriptors for Nonrigid Shapes with a Single Closed Contour Longin Jan Latecki and Rolf Lakämper and Ulrich Eckhardt
More information3D Pipeline Segmentation (planar)
3D Pipeline Segmentation (planar) Registration (using lines) 3D PHOTOGRAPHY EXAMPLE Ten scans were acquired of façade of Thomas Hunter Building (NYC) Automatic registration. Each scan has a different color.
More informationConvolution. 1D Formula: 2D Formula: Example on the web: http://www.jhu.edu/~signals/convolve/
Basic Filters (7) Convolution/correlation/Linear filtering Gaussian filters Smoothing and noise reduction First derivatives of Gaussian Second derivative of Gaussian: Laplacian Oriented Gaussian filters
More informationClassification of Fingerprints. Sarat C. Dass Department of Statistics & Probability
Classification of Fingerprints Sarat C. Dass Department of Statistics & Probability Fingerprint Classification Fingerprint classification is a coarse level partitioning of a fingerprint database into smaller
More informationCircle Object Recognition Based on Monocular Vision for Home Security Robot
Journal of Applied Science and Engineering, Vol. 16, No. 3, pp. 261 268 (2013) DOI: 10.6180/jase.2013.16.3.05 Circle Object Recognition Based on Monocular Vision for Home Security Robot ShihAn Li, ChingChang
More informationAPPLYING COMPUTER VISION TECHNIQUES TO TOPOGRAPHIC OBJECTS
APPLYING COMPUTER VISION TECHNIQUES TO TOPOGRAPHIC OBJECTS Laura Keyes, Adam Winstanley Department of Computer Science National University of Ireland Maynooth Co. Kildare, Ireland lkeyes@cs.may.ie, Adam.Winstanley@may.ie
More informationCentroid Distance Function and the Fourier Descriptor with Applications to Cancer Cell Clustering
Centroid Distance Function and the Fourier Descriptor with Applications to Cancer Cell Clustering By, Swati Bhonsle Alissa Klinzmann Mentors Fred Park Department of Mathematics Ernie Esser Department of
More informationCalibrating a Camera and Rebuilding a Scene by Detecting a Fixed Size Common Object in an Image
Calibrating a Camera and Rebuilding a Scene by Detecting a Fixed Size Common Object in an Image Levi Franklin Section 1: Introduction One of the difficulties of trying to determine information about a
More information2D GEOMETRIC SHAPE AND COLOR RECOGNITION USING DIGITAL IMAGE PROCESSING
2D GEOMETRIC SHAPE AND COLOR RECOGNITION USING DIGITAL IMAGE PROCESSING Sanket Rege 1, Rajendra Memane 2, Mihir Phatak 3, Parag Agarwal 4 UG Student, Dept. of E&TC Engineering, PVG s COET, Pune, Maharashtra,
More informationReducing the time complexity of Minkowski sum based similarity calculations by using geometric inequalities
Reducing the time complexity of Minkowski sum based similarity calculations by using geometric inequalities Henk Bekker 1, Axel Brink 1 1 Institute for Mathematics and Computing Science, University of
More informationPoker Vision: Playing Cards and Chips Identification based on Image Processing
Poker Vision: Playing Cards and Chips Identification based on Image Processing Paulo Martins 1, Luís Paulo Reis 2, and Luís Teófilo 2 1 DEEC Electrical Engineering Department 2 LIACC Artificial Intelligence
More informationExistence of Simple Tours of Imprecise Points
Existence of Simple Tours of Imprecise Points Maarten Löffler Department of Information and Computing Sciences, Utrecht University Technical Report UUCS00700 www.cs.uu.nl ISSN: 097 Existence of Simple
More informationRelational and Spatial 3D Visualisation with Dental Cone Beam CT Images. Ma Tengfei 1, Ong Sim Heng 2
Relational and Spatial 3D Visualisation with Dental Cone Beam CT Images Ma Tengfei 1, Ong Sim Heng 2 Department of Electrical & Computer Engineering National University of Singapore, 10 Kent Ridge Crescent,
More informationSegmentation of Ureteric and Bladder Calculi in Ultrasound Images
Journal of Computer Science 8 (5): 716720, 2012 ISSN 15493636 2012 Science Publications Segmentation of Ureteric and Bladder Calculi in Ultrasound Images Sridhar, S. Department of Information Science
More informationBIL Computer Vision Mar 5, 2014
BIL 719  Computer Vision Mar 5, 2014 Image pyramids Aykut Erdem Dept. of Computer Engineering Hacettepe University Image Scaling This image is too big to fit on the screen. How can we reduce it? How to
More informationPalmprint Recognition. By Sree Rama Murthy kora Praveen Verma Yashwant Kashyap
Palmprint Recognition By Sree Rama Murthy kora Praveen Verma Yashwant Kashyap Palm print Palm Patterns are utilized in many applications: 1. To correlate palm patterns with medical disorders, e.g. genetic
More informationGalaxy Morphological Classification
Galaxy Morphological Classification Jordan Duprey and James Kolano Abstract To solve the issue of galaxy morphological classification according to a classification scheme modelled off of the Hubble Sequence,
More informationC A R I B B E A N E X A M I N A T I O N S C O U N C I L REPORT ON CANDIDATES WORK IN THE SECONDARY EDUCATION CERTIFICATE EXAMINATION MAY/JUNE 2011
C A R I B B E A N E X A M I N A T I O N S C O U N C I L REPORT ON CANDIDATES WORK IN THE SECONDARY EDUCATION CERTIFICATE EXAMINATION MAY/JUNE 2011 MATHEMATICS GENERAL PROFICIENCY EXAMINATION Copyright
More informationMore Local Structure Information for MakeModel Recognition
More Local Structure Information for MakeModel Recognition David Anthony Torres Dept. of Computer Science The University of California at San Diego La Jolla, CA 9093 Abstract An object classification
More informationShapes Puzzle 1. Shapes Puzzle 2. Shapes Puzzle 3
Shapes Puzzle The twelve pentominoes are shown on the left. On the right, they have been placed together in pairs. Can you show which two pentominoes have been used to make each shape? (Each pentomino
More informationLine Separation for Complex Document Images Using Fuzzy Runlength
Line Separation for Complex Document Images Using Fuzzy Runlength Zhixin Shi and Venu Govindaraju Center of Excellence for Document Analysis and Recognition(CEDAR) State University of New York at Buffalo,
More informationBlind Deconvolution of Corrupted Barcode Signals
Blind Deconvolution of Corrupted Barcode Signals Everardo Uribe and Yifan Zhang Advisors: Ernie Esser and Yifei Lou Interdisciplinary Computational and Applied Mathematics Program University of California,
More informationThe Role of Size Normalization on the Recognition Rate of Handwritten Numerals
The Role of Size Normalization on the Recognition Rate of Handwritten Numerals Chun Lei He, Ping Zhang, Jianxiong Dong, Ching Y. Suen, Tien D. Bui Centre for Pattern Recognition and Machine Intelligence,
More informationSheet Music Reader 1 INTRODUCTION 2 IMPLEMENTATION
1 Sheet Music Reader Sevy Harris, Prateek Verma Department of Electrical Engineering Stanford University Stanford, CA sharris5@stanford.edu, prateekv@stanford.edu Abstract The goal of this project was
More informationDesigning and Testing an Anonymous Face Recognition System
Designing and Testing an Anonymous Face Recognition System Joris Diesvelt University of Twente P.O. Box 217, 7500AE Enschede The Netherlands j.j.diesvelt@student.utwente.nl ABSTRACT This paper contains
More informationMorphological segmentation of histology cell images
Morphological segmentation of histology cell images A.Nedzved, S.Ablameyko, I.Pitas Institute of Engineering Cybernetics of the National Academy of Sciences Surganova, 6, 00 Minsk, Belarus Email abl@newman.basnet.by
More informationSolving Square Jigsaw Puzzles Using Dynamic Programming and the Hungarian Procedure
American Journal of Applied Sciences 6 (11): 19411947, 2009 ISSN 15469239 2009 Science Publications Solving Square Jigsaw Puzzles Using Dynamic Programming and the Hungarian Procedure Naif Alajlan Advanced
More informationA recognition system for symbols of electronic components in handwritten circuit diagrams.
A recognition system for symbols of electronic components in handwritten circuit diagrams. Pablo Sala Department of Computer Science University of Toronto Toronto, Canada psala@cs.toronto.edu Abstract
More informationAIMA Chapter 3: Solving problems by searching
AIMA Chapter 3: Solving problems by searching Ana Huaman January 13, 2011 1. Define in your own words the following terms: state, state space, search tree, search node, goal, action, transition model and
More informationComputer Vision: Filtering
Computer Vision: Filtering Raquel Urtasun TTI Chicago Jan 10, 2013 Raquel Urtasun (TTIC) Computer Vision Jan 10, 2013 1 / 82 Today s lecture... Image formation Image Filtering Raquel Urtasun (TTIC) Computer
More informationA New Robust Algorithm for Video Text Extraction
A New Robust Algorithm for Video Text Extraction Pattern Recognition, vol. 36, no. 6, June 2003 Edward K. Wong and Minya Chen School of Electrical Engineering and Computer Science Kyungpook National Univ.
More informationFace detection is a process of localizing and extracting the face region from the
Chapter 4 FACE NORMALIZATION 4.1 INTRODUCTION Face detection is a process of localizing and extracting the face region from the background. The detected face varies in rotation, brightness, size, etc.
More informationDetermining optimal window size for texture feature extraction methods
IX Spanish Symposium on Pattern Recognition and Image Analysis, Castellon, Spain, May 2001, vol.2, 237242, ISBN: 8480213515. Determining optimal window size for texture feature extraction methods Domènec
More informationLecture 9: Shape Description (Regions)
Lecture 9: Shape Description (Regions) c Bryan S. Morse, Brigham Young University, 1998 2000 Last modified on February 16, 2000 at 4:00 PM Contents 9.1 What Are Descriptors?.........................................
More informationMap Scanning and Automated Conversion
Objectives (Entry) Map Scanning and Automated Conversion This unit will introduce quick and less costly method of data capture  map scanning and automated conversion. This unit will briefly discuss about
More informationContrast enhancement of soft tissues in Computed Tomography images
Contrast enhancement of soft tissues in Computed Tomography images Roman Lerman, Daniela S. Raicu, Jacob D. Furst Intelligent Multimedia Processing Laboratory School of Computer Science, Telecommunications,
More informationAN INTRODUCTION TO ERROR CORRECTING CODES Part 1
AN INTRODUCTION TO ERROR CORRECTING CODES Part 1 Jack Keil Wolf ECE 154C Spring 2008 Noisy Communications Noise in a communications channel can cause errors in the transmission of binary digits. Transmit:
More informationCS 534: Computer Vision 3D Modelbased recognition
CS 534: Computer Vision 3D Modelbased recognition Ahmed Elgammal Dept of Computer Science CS 534 3D Modelbased Vision  1 High Level Vision Object Recognition: What it means? Two main recognition tasks:!
More informationA Guide to Creating MachineReadable Forms
A Guide to Creating MachineReadable Forms 2009 ABBYY. All rights reserved. Table of Contents What is a Form?... 3 Machine Readable Forms... 4 Form Completion Methods...4 Elements of Machine Readable Forms...4
More informationDocument Image Retrieval using Signatures as Queries
Document Image Retrieval using Signatures as Queries Sargur N. Srihari, Shravya Shetty, Siyuan Chen, Harish Srinivasan, Chen Huang CEDAR, University at Buffalo(SUNY) Amherst, New York 14228 Gady Agam and
More informationBattleship. Big bands
Ball Fill in the grid so that every row, column (six smaller cells and three bigger circles or stars), outlined figures (eight smaller cells and a bigger circle), nine bigger circles and nine bigger stars
More informationGauss's Law. Gauss's Law in 3, 2, and 1 Dimension
[ Assignment View ] [ Eðlisfræði 2, vor 2007 22. Gauss' Law Assignment is due at 2:00am on Wednesday, January 31, 2007 Credit for problems submitted late will decrease to 0% after the deadline has passed.
More information(Refer Slide Time: 00:01:23 min)
Computer Aided Design Prof. Anoop Chalwa Department of Mechanical Engineering Indian Institute of Technology, Delhi Lecture No. # 02 Input Output Devices, Raster Graphics Today we will be talking about
More informationPuzzleSolver. An application of computer vision. Joe Zeimen Computer Vision Colorado School of Mines Spring 2013
PuzzleSolver An application of computer vision. Joe Zeimen Computer Vision Colorado School of Mines Spring 2013 Table of Contents Introduction... 3 Previous Work... 3 Program Implementation... 4 Assumptions...
More information3D Human Face Recognition Using Point Signature
3D Human Face Recognition Using Point Signature ChinSeng Chua, Feng Han, YeongKhing Ho School of Electrical and Electronic Engineering Nanyang Technological University, Singapore 639798 ECSChua@ntu.edu.sg
More informationElectronic Schematic Recognition
Electronic Schematic Recognition Donald Bailey 1, Andrew Norman 2, and Giovanni Moretti 2 1 Physics Department and 2 Computer Science Department Massey University Palmerston North Email: D.G.Bailey@massey.ac.nz,
More informationEyeglass Localization for Low Resolution Images
Eyeglass Localization for Low Resolution Images Earl Arvin Calapatia 1 1 De La Salle University 1 earl_calapatia@dlsu.ph Abstract: Facial data is a necessity in facial image processing technologies. In
More informationA Simple Feature Extraction Technique of a Pattern By Hopfield Network
A Simple Feature Extraction Technique of a Pattern By Hopfield Network A.Nag!, S. Biswas *, D. Sarkar *, P.P. Sarkar *, B. Gupta **! Academy of Technology, Hoogly  722 *USIC, University of Kalyani, Kalyani
More informationFingerprint Identification Using Delaunay Triangulation
Fingerprint Identification Using Delaunay Triangulation George Bebis, Taisa Deaconu and Michael Georgiopoulos Department of Computer Science, University of Nevada, Reno Department of Electrical and Computer
More information3194 Mersenne Composite Numbers
3194 Mersenne Composite Numbers One of the world wide cooperative computing tasks is the ``Grand Internet Mersenne Prime Search" GIMPS striving to find ever larger prime numbers by examining a particular
More informationRobust Panoramic Image Stitching
Robust Panoramic Image Stitching CS231A Final Report Harrison Chau Department of Aeronautics and Astronautics Stanford University Stanford, CA, USA hwchau@stanford.edu Robert Karol Department of Aeronautics
More informationSolving Multiple Square Jigsaw Puzzles with Missing Pieces
Solving Multiple Square Jigsaw Puzzles with Missing Pieces Genady Paikin Technion Ayellet Tal Technion genadyp28@gmail.com ayellet@ee.technion.ac.il Abstract Jigsawpuzzle solving is necessary in many
More informationSegmentation of building models from dense 3D pointclouds
Segmentation of building models from dense 3D pointclouds Joachim Bauer, Konrad Karner, Konrad Schindler, Andreas Klaus, Christopher Zach VRVis Research Center for Virtual Reality and Visualization, Institute
More informationAutomatic Liver Segmentation using the Random Walker Algorithm
Automatic Liver Segmentation using the Random Walker Algorithm F. Maier 1,2, A. Wimmer 2,3, G. Soza 2, J. N. Kaftan 2,4, D. Fritz 1,2, R. Dillmann 1 1 Universität Karlsruhe (TH), 2 Siemens Medical Solutions,
More informationImage ContentBased Email Spam Image Filtering
Image ContentBased Email Spam Image Filtering Jianyi Wang and Kazuki Katagishi Abstract With the population of Internet around the world, email has become one of the main methods of communication among
More informationSudoku Solver. Yixin Wang
Sudoku Solver Yixin Wang wangyix@stanford.edu Abstract An Android app was developed that allows a Sudoku puzzle to be extracted and solved in real time using captured images from the device s camera.
More informationhttp://www.elsevier.com/copyright
This article was published in an Elsevier journal. The attached copy is furnished to the author for noncommercial research and education use, including for instruction at the author s institution, sharing
More informationImage Estimation Algorithm for Out of Focus and Blur Images to Retrieve the Barcode Value
IJSTE  International Journal of Science Technology & Engineering Volume 1 Issue 10 April 2015 ISSN (online): 2349784X Image Estimation Algorithm for Out of Focus and Blur Images to Retrieve the Barcode
More informationScalar Visualization
Scalar Visualization 41 Motivation Visualizing scalar data is frequently encountered in science, engineering, and medicine, but also in daily life. Recalling from earlier, scalar datasets, or scalar fields,
More informationA Learning Based Method for SuperResolution of Low Resolution Images
A Learning Based Method for SuperResolution of Low Resolution Images Emre Ugur June 1, 2004 emre.ugur@ceng.metu.edu.tr Abstract The main objective of this project is the study of a learning based method
More informationComputer Vision  part II
Computer Vision  part II Review of main parts of Section B of the course School of Computer Science & Statistics Trinity College Dublin Dublin 2 Ireland www.scss.tcd.ie Lecture Name Course Name 1 1 2
More informationDigital Image Processing
Digital Image Processing Using MATLAB Second Edition Rafael C. Gonzalez University of Tennessee Richard E. Woods MedData Interactive Steven L. Eddins The MathWorks, Inc. Gatesmark Publishing A Division
More informationChess Vision. Chua Huiyan Le Vinh Wong Lai Kuan
Chess Vision Chua Huiyan Le Vinh Wong Lai Kuan Outline Introduction Background Studies 2D Chess Vision Realtime Board Detection Extraction and Undistortion of Board Board Configuration Recognition 3D
More informationChapter 2  Graphical Summaries of Data
Chapter 2  Graphical Summaries of Data Data recorded in the sequence in which they are collected and before they are processed or ranked are called raw data. Raw data is often difficult to make sense
More informationLocating and Decoding EAN13 Barcodes from Images Captured by Digital Cameras
Locating and Decoding EAN13 Barcodes from Images Captured by Digital Cameras W3A.5 Douglas Chai and Florian Hock Visual Information Processing Research Group School of Engineering and Mathematics Edith
More informationCALCULATIONS & STATISTICS
CALCULATIONS & STATISTICS CALCULATION OF SCORES Conversion of 15 scale to 0100 scores When you look at your report, you will notice that the scores are reported on a 0100 scale, even though respondents
More informationLINE INTEGRALS OF VECTOR FUNCTIONS: GREEN S THEOREM. Contents. 2. Green s Theorem 3
LINE INTEGRALS OF VETOR FUNTIONS: GREEN S THEOREM ontents 1. A differential criterion for conservative vector fields 1 2. Green s Theorem 3 1. A differential criterion for conservative vector fields We
More informationSpeaker Change Detection using Support Vector Machines
Speaker Change Detection using Support Vector Machines V. Kartik and D. Srikrishna Satish and C. Chandra Sekhar Speech and Vision Laboratory Department of Computer Science and Engineering Indian Institute
More informationWinter 2016 Math 213 Final Exam. Points Possible. Subtotal 100. Total 100
Winter 2016 Math 213 Final Exam Name Instructions: Show ALL work. Simplify wherever possible. Clearly indicate your final answer. Problem Number Points Possible Score 1 25 2 25 3 25 4 25 Subtotal 100 Extra
More informationRecognition of Handwritten Digits using Structural Information
Recognition of Handwritten Digits using Structural Information Sven Behnke MartinLuther University, HalleWittenberg' Institute of Computer Science 06099 Halle, Germany { behnke Irojas} @ informatik.unihalle.de
More informationFace Recognition using SIFT Features
Face Recognition using SIFT Features Mohamed Aly CNS186 Term Project Winter 2006 Abstract Face recognition has many important practical applications, like surveillance and access control.
More informationVisual Structure Analysis of Flow Charts in Patent Images
Visual Structure Analysis of Flow Charts in Patent Images Roland Mörzinger, René Schuster, András Horti, and Georg Thallinger JOANNEUM RESEARCH Forschungsgesellschaft mbh DIGITAL  Institute for Information
More informationImplementation of Canny Edge Detector of color images on CELL/B.E. Architecture.
Implementation of Canny Edge Detector of color images on CELL/B.E. Architecture. Chirag Gupta,Sumod Mohan K cgupta@clemson.edu, sumodm@clemson.edu Abstract In this project we propose a method to improve
More informationA network flow algorithm for reconstructing. binary images from discrete Xrays
A network flow algorithm for reconstructing binary images from discrete Xrays Kees Joost Batenburg Leiden University and CWI, The Netherlands kbatenbu@math.leidenuniv.nl Abstract We present a new algorithm
More informationBiometric Authentication using Online Signatures
Biometric Authentication using Online Signatures Alisher Kholmatov and Berrin Yanikoglu alisher@su.sabanciuniv.edu, berrin@sabanciuniv.edu http://fens.sabanciuniv.edu Sabanci University, Tuzla, Istanbul,
More informationTracking of Small Unmanned Aerial Vehicles
Tracking of Small Unmanned Aerial Vehicles Steven Krukowski Adrien Perkins Aeronautics and Astronautics Stanford University Stanford, CA 94305 Email: spk170@stanford.edu Aeronautics and Astronautics Stanford
More informationStatic Environment Recognition Using Omnicamera from a Moving Vehicle
Static Environment Recognition Using Omnicamera from a Moving Vehicle Teruko Yata, Chuck Thorpe Frank Dellaert The Robotics Institute Carnegie Mellon University Pittsburgh, PA 15213 USA College of Computing
More informationG73 Measuring and Drawing Angles and Triangles Pages
G73 Measuring and Drawing Angles and Triangles Pages 102 104 Curriculum Expectations Ontario: 5m51, 5m52, 5m54, 6m48, 6m49, 7m3, 7m4, 7m46 WNCP: 6SS1, review, [T, R, V] Vocabulary angle vertex arms acute
More informationAutomatic Extraction of Signatures from Bank Cheques and other Documents
Automatic Extraction of Signatures from Bank Cheques and other Documents Vamsi Krishna Madasu *, Mohd. Hafizuddin Mohd. Yusof, M. Hanmandlu ß, Kurt Kubik * *Intelligent RealTime Imaging and Sensing group,
More informationPrediction of OCR Accuracy
Prediction of OCR Accuracy Luis R. Blando, 1 Junichi Kanai, Thomas A. Nartker, and Juan Gonzalez 1 Introduction The accuracy of all contemporary OCR technologies varies drastically as a function of input
More informationOneDimensional Processing for Edge Detection using Hilbert Transform P. Kiran Kumar, Sukhendu Das and B. Yegnanarayana Department of Computer Science and Engineering Indian Institute of Technology, Madras
More information