Robust Computer Vision

by Sebe, Nicu; Lew, Michael S.

• ISBN13: 9781402012938
• ISBN10: 1402012934
• Format: Hardcover
• Copyright: 2003-04-01
• Publisher: Kluwer Academic Pub

Summary

From the foreword by Thomas Huang: "During the past decade, researchers in computer vision have found that probabilistic machine learning methods are extremely powerful. This book describes some of these methods. In addition to the Maximum Likelihood framework, Bayesian Networks, and Hidden Markov models are also used. Three aspects are stressed: features, similarity metric, and models. Many interesting and important new results, based on research by the authors and their collaborators, are presented. Although this book contains many new results, it is written in a style that suits both experts and novices in computer vision."

Author Biography

Nicu Sebe received his PhD degree from Leiden University in 2001. Currently, he is an Assistant Professor at Leiden University in the Netherlands. His main interest is in the fields of computer vision and pattern recognition, in particular content-based retrieval and robust techniques in computer vision. He was co-editing the proceedings of the International Conference on Image and Video Retrieval 2002. He is also acting as the technical program co-chair for the International Conference on Image and Video Retrieval 2003.Michael S. Lew received his PhD degree in Electrical Engineering from the University of Illinois at Urbana-Champaign. He is currently an Associate Professor at Leiden University in the Netherlands. He has published over 100 scientific papers and helped organize several large conferences including IEEE Multimedia, ACM Multimedia, and the International Conference on Image and Video Retrieval.

Table of Contents

Foreword	xi
Preface	xiii
1. INTRODUCTION	1	(24)
1 Visual Similarity	2	(14)
1.1 Color	4	(3)
1.2 Texture	7	(2)
1.3 Shape	9	(2)
1.4 Stereo	11	(2)
1.5 Motion	13	(1)
1.6 Facial expression	13	(2)
1.7 Summary	15	(1)
2 Evaluation of Computer Vision Algorithms	16	(3)
3 Overview of the Book	19	(6)
2. MAXIMUM LIKELIHOOD FRAMEWORK	25	(36)
1 Introduction	25	(1)
2 Statistical Distributions	26	(17)
2.1 Gaussian Distribution	27	(11)
2.2 Exponential Distribution	38	(3)
2.3 Cauchy Distribution	41	(2)
3 Robust Statistics	43	(2)
3.1 Outliers	44	(1)
4 Maximum Likelihood Estimators	45	(2)
5 Maximum Likelihood in Relation to Other Approaches	47	(3)
6 Our Maximum Likelihood Approach	50	(7)
6.1 Scale Parameter Estimation in a Cauchy Distribution	54	(3)
7 Experimental Setup	57	(2)
8 Concluding Remarks	59	(2)
3. COLOR BASED RETRIEVAL	61	(22)
1 Introduction	61	(3)
2 Colorimetry	64	(1)
3 Color Models	64	(4)
3.1 RGB Color system	65	(1)
3.2 HSV Color System	66	(1)
3.3 l1 l2 l3 Color System	67	(1)
4 Color Based Retrieval	68	(5)
4.1 Color indexing	69	(4)
5 Experiments with the Corel Database	73	(6)
5.1 Early Experiments	73	(1)
5.2 Usability Issues	74	(1)
5.3 Printer-Scanner Noise Experiments	75	(1)
5.4 Color Model	76	(1)
5.5 Quantization	76	(1)
5.6 Distribution Analysis	77	(2)
6 Experiments with the Objects Database	79	(2)
7 Concluding Remarks	81	(2)
4. ROBUST TEXTURE ANALYSIS	83	(28)
1 Introduction	83	(3)
2 Human Perception of Texture	86	(1)
3 Texture Features	87	(8)
3.1 Texture Distribution Models	88	(4)
3.1.1 Gray-level differences	89	(1)
3.1.2 Laws' texture energy measures	89	(1)
3.1.3 Center-symmetric covariance measures	89	(2)
3.1.4 Local binary patterns and trigrams	91	(1)
3.1.5 Complementary feature pairs	91	(1)
3.2 Gabor and Wavelet Models	92	(3)
4 Texture Classification Experiments	95	(9)
4.2 Distribution Analysis	97	(2)
4.3 Misdetection Rates	99	(5)
4.3. 1 Summary	104	(1)
5 Texture Retrieval Experiments	104	(5)
5.1 Texture Features	105	(1)
5.2 Experiments Setup	106	(1)
5.3 Similarity Noise for QMF-Wavelet Transform	106	(2)
5.4 Similarity Noise for Gabor Wavelet Transform	108	(1)
6 Concluding Remarks	109	(2)
5. SHAPE BASED RETRIEVAL	111	(24)
1 Introduction	111	(2)
2 Human Perception of Visual Form	113	(5)
3 Active Contours	118	(12)
3.1 Behavior of Traditional Active Contours	120	(4)
3.2 Generalized Force Balance Equations	124	(1)
3.3 Gradient Vector Flow	125	(5)
4 Invariant Moments	130	(1)
5 Experiments	131	(3)
6 Conclusions	134	(1)
6. ROBUST STEREO MATCHING AND MOTION TRACKING	135	(28)
1 Introduction	135	(3)
1.1 Stereoscopic Vision	137	(1)
2 Stereo Matching	138	(6)
2.1 Related Work	142	(2)
3 Stereo Matching Algorithms	144	(6)
3.1 Template Based Algorithm	144	(2)
3.2 Multiple Windows Algorithm	146	(1)
3.3 Cox' Maximum Likelihood Algorithm	147	(3)
4 Stereo Matching Experiments	150	(7)
4.1 Stereo sets	151	(1)
4.2 Stereo Matching Results	151	(6)
4.3 Summary	157	(1)
5 Motion Tracking Experiments	157	(3)
6 Concluding Remarks	160	(3)
7. FACIAL EXPRESSION RECOGNITION	163	(36)
1 Introduction	164	(2)
2 Emotion Recognition	166	(5)
2.1 Judgment Studies	167	(1)
2.2 Review of Facial Expression Recognition	167	(4)
3 Face Tracking and Feature Extraction	171	(2)
4 The static Approach: Bayesian Network Classifiers	173	(6)
4.1 Continuous Naive-Bayes: Gaussian and Cauchy Naive Bayes Classifiers	175	(1)
4.2 Beyond the Naive-Bayes Assumption: Finding Dependencies among Features Using a Gaussian TAN Classifier	176	(3)
5 The Dynamic Approach: Expression Recognition Using Multi-level HMMs	179	(8)
5.1 Hidden Markov Models	182	(1)
5.2 Expression Recognition Using Emotion-Specific HMMs	183	(1)
5.3 Automatic Segmentation and Recognition of Emotions Using Multi-level HMMs	184	(3)
6 Experiments	187	(8)
6.1 Results Using the Chen Database	191	(3)
6.1.1 Person-Dependent Tests	191	(2)
6.1.2 Person-Independent Tests	193	(1)
6.2 Results Using the Cohn-Kanade Database	194	(1)
7 Summary and Discussion	195	(4)
References	199	(11)
Index	210