yap, guan, perry, wong - adaptive image processing. a computational intelligence perspective 2nd

November 18, 2009 17:7 84356 84356˙C000 November 18, 2009 17:7 84356 84356˙C000 November 18, 2009 17:7 84356 84356˙C000 November 18, 2009 17:7 84356 84356˙C000 November 18, 2009 17:7 84356 84356˙C000 November 18, 2009 17:7 84356 84356˙C000 CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2010 by Taylor and Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original U.S Government works Printed in the United States of America on acid-free paper 10 International Standard Book Number: 978-1-4200-8435-1 (Hardback) This book contains information obtained from authentic and highly regarded sources Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint Except as permitted under U.S Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400 CCC is a not-for-profit organization that provides licenses and registration for a variety of users For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com November 18, 2009 17:7 84356 84356˙C000 Contents Preface xiii Introduction 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 Importance of Vision Adaptive Image Processing Three Main Image Feature Classes 1.3.1 Smooth Regions 1.3.2 Edges 1.3.3 Textures .4 Difficulties in Adaptive Image-Processing System Design 1.4.1 Segmentation 1.4.2 Characterization 1.4.3 Optimization .7 Computational Intelligence Techniques .8 1.5.1 Neural Networks 10 1.5.2 Fuzzy Logic 11 1.5.3 Evolutionary Computation 12 Scope of the Book 13 1.6.1 Image Restoration 13 1.6.2 Edge Characterization and Detection 15 1.6.3 Self-Organizing Tree Map for Knowledge Discovery 16 1.6.4 Content-Based Image Categorization and Retrieval 18 Contributions of the Current Work 19 1.7.1 Application of Neural Networks for Image Restoration 19 1.7.2 Application of Neural Networks to Edge Characterization 20 1.7.3 Application of Fuzzy Set Theory to Adaptive Regularization .20 1.7.4 Application of Evolutionary Programming to Adaptive Regularization and Blind Deconvolution 21 1.7.5 Application of Self-Organization to Image Analysis and Retrieval 21 1.7.6 Application of Evolutionary Computation to Image Categorization 22 1.7.7 Application of Computational Intelligence to Content-Based Image Retrieval 22 Overview of This Book .23 vii November 18, 2009 viii 17:7 84356 84356˙C000 Contents Fundamentals of CI-Inspired Adaptive Image Restoration 25 2.1 Neural Networks as a CI Architecture 25 2.2 Image Distortions 25 2.3 Image Restoration 29 2.4 Constrained Least Square Error 29 2.4.1 A Bayesian Perspective 30 2.4.2 A Lagrangian Perspective 32 2.5 Neural Network Restoration 35 2.6 Neural Network Restoration Algorithms in the Literature 37 2.7 An Improved Algorithm 40 2.8 Analysis 43 2.9 Implementation Considerations 45 2.10 Numerical Study of the Algorithms 45 2.10.1 Setup 45 2.10.2 Efficiency 46 2.11 Summary 46 Spatially Adaptive Image Restoration 49 3.1 Introduction 49 3.2 Dealing with Spatially Variant Distortion 51 3.3 Adaptive Constraint Extension of the Penalty Function Model 53 3.3.1 Motivation 54 3.3.2 Gradient-Based Method 56 3.3.3 Local Statistics Analysis 64 3.4 Correcting Spatially Variant Distortion Using Adaptive Constraints 69 3.5 Semiblind Restoration Using Adaptive Constraints 74 3.6 Implementation Considerations 78 3.7 More Numerical Examples 79 3.7.1 Efficiency 79 3.7.2 Application Example .80 3.8 Adaptive Constraint Extension of the Lagrange Model 80 3.8.1 Problem Formulation 80 3.8.2 Problem Solution 83 3.8.3 Conditions for KKT Theory to Hold 85 3.8.4 Discussion 87 3.9 Summary 88 Perceptually Motivated Image Restoration 89 4.1 Introduction 89 4.2 Motivation 90 4.3 LVMSE-Based Cost Function 91 4.3.1 Extended Algorithm for the LVMSE-Modified Cost Function 92 4.3.2 Analysis 96 November 18, 2009 17:7 Contents 84356 84356˙C000 ix Log LVMSE-Based Cost Function 100 4.4.1 Extended Algorithm for the Log LVR-Modified Cost Function 101 4.4.2 Analysis 103 4.5 Implementation Considerations 105 4.6 Numerical Examples 106 4.6.1 Color Image Restoration 106 4.6.2 Grayscale Image Restoration 109 4.6.3 LSMSE of Different Algorithms 109 4.6.4 Robustness Evaluation 111 4.6.5 Subjective Survey 113 4.7 Local Variance Extension of the Lagrange Model 114 4.7.1 Problem Formulation 114 4.7.2 Computing Local Variance 116 4.7.3 Problem Solution 117 4.7.4 Conditions for KKT Theory to Hold 118 4.7.5 Implementation Considerations for the Lagrangian Approach 120 4.7.6 Numerical Experiment 121 4.8 Summary 122 Acknowledgments 122 4.4 Model-Based Adaptive Image Restoration 123 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 5.15 Model-Based Neural Network 123 5.1.1 Weight-Parameterized Model-Based Neuron 124 Hierarchical Neural Network Architecture 125 Model-Based Neural Network with Hierarchical Architecture 125 HMBNN for Adaptive Image Processing 126 Hopfield Neural Network Model for Image Restoration 127 Adaptive Regularization: An Alternative Formulation .128 5.6.1 Correspondence with the General HMBNN Architecture 130 Regional Training Set Definition 134 Determination of the Image Partition 137 Edge-Texture Characterization Measure 139 ETC Fuzzy HMBNN for Adaptive Regularization 142 Theory of Fuzzy Sets 143 Edge-Texture Fuzzy Model Based on ETC Measure 145 Architecture of the Fuzzy HMBNN 147 5.13.1 Correspondence with the General HMBNN Architecture 148 Estimation of the Desired Network Output .149 Fuzzy Prediction of Desired Gray-Level Value 151 5.15.1 Definition of the Fuzzy Estimator Membership Function 151 References 349 219 ISO/IEC JTC 1/SC 29/WG N 751, Coding of Still Pictures International Organization for Standardization, Geneva, Switzerland, 1998 220 B Furht, 1995, A survey of multimedia compression techniques and standards, Real-Time Imaging 1: 49–67 221 A W M Smeulders, M Worring, S Santini, A Gupta, and R Jain, 2000, Contentbased image retrieval at the end of the early years, IEEE Trans Pattern Analysis and Machine Intelligence 22(12): 1349–1380 222 Y Rui, T S Huang, and S F Chang, 1999, Image retrieval: Current techniques, promising directions and open issues, J Visual Comm and Image Representation 10(1): 39–62 223 A Gupta and R Jain, 1997, Visual information retrieval, Comm ACM 40(5): 71–79 224 C C Hsu, W W Chu, and R K Taira, 1996, A knowledge-based approach for retrieving images by content, IEEE Trans on Knowledge and Data Engineering 8(4): 522–532 225 M K Mandal, F M Idris, and S Panchanathan, 1999, A critical evaluation of image and video indexing techniques in the compressed domain, Image Vision Computing 17(7): 513–529 226 E Regentova, S Latifi, and S Deng, 2001, Images similarity estimation by processing compressed data in the compressed domain, Image and Vision Computing, 19(7): 485–500 227 R Chang, W Kuo, and H Tsai, 2000, Image retrieval on uncompressed and compressed domains DCT coefficients, in Proc Int Conf on Image Processing 2: 546–549 228 H H Yu, 1999, Visual image retrieval on compressed domain with Q-distance, in Proc ICCIMA ’99, Proc 3rd International Conference on Computational Intelligence and Multimedia Applications: 285–289 229 S Climer and S K Bhatia, 2002, Image database indexing using JPEG coefficients, Patt Recog 35(11): 2479–2488 230 N Ahmed, T Natarajan, and K Rao, 1974, Discrete cosine transform, IEEE Trans Comput 23: 90–93 231 M Shneier and M A Mottaleb, 1996, Exploiting the JPEG compression scheme for image retrieval, IEEE Trans Pattern Analysis and Machine Intelligence 18(8): 849–853 232 J Jian, A J Armstrong, and G C Feng, 2001, Direct content access and extraction from JPEG compressed images, Patt Recog 35(11): 2511–2519 233 A Vailaya, M Figueiredo, A K Jain, and H Zhang, 2001, Image classification for content-based indexing, IEEE Trans on Image Processing 10(1) 234 A Conci and E Castro, 2002, Image mining by content, Expert Systems with Applications 23(4): 377–383 235 J M Corridoni, A D Bimbo, and P Pala, 1999, Image retrieval by color semantics, Multimedia Systems 7(3): 175–183 236 J Hafner, H Sawhney, W Equitz, M Flickner, and W Niblack, 1995, Efficient color histogram indexing for quadratic form distance functions, IEEE Trans Pattern Analysis and Machine Intelligence 17(7): 729–736 237 S Cha and S N Srihari, 2002, On measuring the distance between histograms, Patt Recog 35(6): 1355–1370 238 K Sirlantzis and M Fairhurst, 2001, Optimisation of multiple classifier systems using genetic algorithms, in Proc IEEE Int Conf on Image Processing: 1094–1097 350 Adaptive Image Processing: A Computational Intelligence Perspective 239 K Sirlantzis, M C Fairhurst, and R M Guest, 2002, An evolutionary algorithm for classifier and combination rule selection in multiple classifier systems, in Proc IEEE Int Conf on Pattern Recognition: 771–774 240 L Xu, A Krzyzak, and C Suen, 1992, Methods of combining multiple classifiers and their applications to handwriting recognition, IEEE Trans System, Man and Cybernetics 22(3): 418–435 241 P Smits, 2002, Multiple classifier systems for supervised remote sensing image classification based on dynamic classifier selection, IEEE Trans Geoscience and Remote Sensing 40(4): 801–813 242 J Kittler and F M Alkoot, 2003, Sum versus vote fusion in multiple classifier systems, IEEE Trans Pattern Analysis and Machine Intelligence 25(1): 110–115 243 TelecomsEurope, 2008, Worldwide mobile phone user base hits record 3.25b http://www.telecomseurope.net/article.php?id\_article=4208 244 InfoTrends, 2006, InfoTrends releases mobile imaging study results http://www.infotrends-rgi.com/home/Press/itPress/2006/1.18.2006.html 245 PRNewswire, 2008, Flickr adds video to its popular photo-sharing community http://www.hispanicprwire.com/generarnews.php?l=in&id=11197&cha=0 246 M Flickher, H Sawhney, W Niblack, J Ashley, Q Huang, B Dom, M Gorkani, J Hafner, D Lee, D Petkovic, et al., 1995, Query by image and video content: The QBIC system, IEEE Computer 28(Sept.): 23–32 247 A Pentland, R Picard, and S Sclaroff, 1997, Photobook: Content-based manipulation of image databases, Commun ACM 40(Sept.): 70–79 248 J R Smith and S F Chang, 1996, VisualSEEk: a fully automated content-based image query system, in Proc ACM Multimedia: 87–98 249 T Gevers and A W M Smeulders, 2000, PicToSeek: Combining color and shape invariant features for image retrieval, IEEE Trans Image Processing 9(Jan.): 102– 119 250 I J Cox, M L Miller, T P Minka, T V Papathomas, and P N Yianilos, 2000, The Bayesian image retrieval system, PicHunter: Theory, implementation, and psychophysical experiments, IEEE Trans Image Processing 9(Jan.): 20–37 251 J Huang, S R Kumar, and M Metra, 1997, Combining supervised learning with color correlograms for content-based image retrieval, in Proc ACM Multimedia: 325–334 252 Y Rui, T S Huang, M Ortega, and S Mehrotra, 1998, Relevance feedback: A power tool for interactive content-based image retrieval, IEEE Trans Circuits and Video Technology 8(Sept.): 644–655 253 Y Rui and T S Huang, 2000, Optimizing learning in image retrieval, in Proc IEEE Int Conf Computer Vision and Pattern Recognition 1(Jun.): 236–243 254 N Vasconcelos and A Lippman, 1999, Learning from user feedback in image retrieval systems, in Proc Neural Information Processing Systems: 977–986 255 Z Su, H J Zhang, S Li, and S P Ma, 2003, Relevance feedback in contentbased image retrieval: Bayesian framework, feature subspaces, and progressive learning, IEEE Trans Image Processing 12(Aug.): 924–937 256 H K Lee and S I Yoo, 2001, A neural network-based image retrieval using nonlinear combination of heterogeneous features, Int J Computational Intelligence and Applications 1(2): 137–149 257 J Laaksonen, M Koskela, and E Oja, 2002, PicSom: self-organizing image retrieval with MPEG-7 content descriptions, IEEE Trans Neural Networks 13(Jul.): 841–853 References 351 258 P Muneesawang and L Guan, 2002, Automatic machine interactions for content-based image retrieval using a self-organizing tree map architecture, IEEE Trans Neural Network 13(Jul.): 821–834 259 K H Yap and K Wu, 2005, Fuzzy relevance feedback in content-based image retrieval systems using radial basis function network, in Proc IEEE Int Conf Multimedia and Expo: 177–180 260 K H Yap and K Wu, 2005, A soft relevance framework in content-based image retrieval systems, IEEE Trans Circuits and Systems for Video Technology 15(Dec.): 1557–1568 261 S Tong and E Chang, 2001, Support vector machine active leaning for image retrieval, in Proc ACM Int Conf Multimedia: 107–118 262 Y Chen, X S Zhou, and T S Huang, 2001, One-class SVM for learning in image retrieval, in Proc IEEE Int Conf Image Processing: 815–818 263 G D Guo, A K Jain, W Y Ma, and H J Zhang, 2002, Learning similarity measure for natural image retrieval with relevance feedback, IEEE Trans Neural Networks 13(Jul.): 811–820 264 L Wang and K L Chan, 2003, Bootstrapping SVM active learning by incorporating unlabelled images for image retrieval, in Proc IEEE Int Conf Computer Vision and Pattern Recognition: 629–634 265 L Wang and K L Chan, 2004, Incorporating prior knowledge into SVM for image retrieval, in Proc IEEE Int Conf Pattern Recognition: 981–984 266 Y Wu, Q Tian, and T S Huang, 2000, Discriminant-EM algorithm with application to image retrieval, in Proc IEEE Int Conf Computer Vision and Pattern Recognition, 222–227 267 S Chiu, 1994, Fuzzy model identification based on cluster estimation, J Intelligent and Fuzzy Systems 2(Sept.) 268 X S Zhou and T S Huang, 2003, Relevance feedback in image retrieval: A comprehensive review, ACM Multimedia Systems J 8(6): 536–544 269 J Friedman, Regularized discriminant analysis, 1989, J Am Stat Assoc 84(405): 165–175 270 M Swain and D Ballard, 1991, Color indexing, Int J Comp Vis 7(1): 11–32 271 S Markus and O Markus, 1995, Similarity of color images, in Proc SPIE Storage and Retrieval for Image and Video Databases: 381–392 272 J Huang, S R Kumar, M Mitra, W J Zhu, and R Zabih, 1997, Image indexing using color correlograms, in Proc IEEE Conf Computer Vision and Pattern Recognition Conference: 762–768, San Juan, Puerto Rico 273 J R Smith and S F Chang, 1996, Automated binary texture feature sets for image retrieval, in Proc Int Conf Acoustics, Speech, and Signal Processing: 2239– 2242, Atlanta, GA 274 X F He, O King, W Y Ma, M J Li, and H J Zhang, 2003, Learning a semantic space from user’s relevance feedback for image retrieval, IEEE Trans Circuits and Systems for Video Technology 13(Jan.): 39–48 275 C F Lin and S D Wang, Fuzzy support vector machines, 2002, IEEE Trans Neural Networks 13(Mar.): 464–471 276 V N Vapnik, The Nature of Statistical Learning Theory New York: Springer-Verlag, 1995 Index A Active learning, 334 Adaptive constraint extension of Lagrange model, 80–87 Adaptive constraint restoration, 50 Adaptive image processing, 2–3 main requirements in, system design, difficulties in, 5–8 characterization, optimization, 7–8 segmentation, Adaptive image restoration, see CI-inspired adaptive image restoration; Model-based adaptive image restoration; Spatially adaptive image restoration Adaptive nonlinear filtering, 20 Adaptive-RBF Network (ARBFN), 290 Adaptive regularization, see also Evolutionary computation, adaptive regularization using application of evolutionary programming to, 21 application of fuzzy set theory to, 20–21 in image restoration, 13, 14 parameters, Adaptive resonance theory (ART), 263 Additive noise, 231–232 Additive white Gaussian noise (AWGN), 195 AI, see Artificial intelligence AI-CBR, see Automatic interactive content-based retrieval ARBFN, see Adaptive-RBF Network ARMA, see Autoregressive moving average ART, see Adaptive resonance theory Artificial intelligence (AI), 8, 315 Automatic interaction procedure, 290 Automatic interactive content-based retrieval (AI-CBR), 286 Autoregressive moving average (ARMA), 195 AWGN, see Additive white Gaussian noise B Bias vector, 44 Binary strings, 170 Blind deconvolution, 15, 21, 197, 234, 238 Blind image deconvolution, 195–238 blur identification by recursive soft decision, 198 computational reinforced learning, 197, 198–215 formulation of blind image deconvolution as evolutionary strategy, 198–205 knowledge-based reinforced mutation, 205–209 perception-based image restoration, 210–211 performance evaluation and selection, 213–215 recombination based on niche-space residency, 212–213 simulation examples, 229–238 identification of 2-D Gaussian blur, 230–231 identification of 2-D Gaussian blur from degraded image with additive noise, 231–232 353 354 identification of nonstandard blur by RSD, 235–238 identification of 2-D uniform blur by CRL, 232–234 soft-decision method, 215–228 blur compensation, 226–228 blur identification by conjugate gradient optimization, 223–225 hierarchical neural network for image restoration, 217–222 recursive subspace optimization, 215–217 soft parametric blur estimator, 222–223 Block-Toeplitz matrix, 28 Blur compensation, 226–228 Book overview, 23 Book scope, 13–19 content-based image categorization and retrieval, 18–19 content analysis, 18 relevance feedback, 18–19 edge characterization and detection, 15–16 image restoration, 13–15 adaptive regularization, 14 blind deconvolution, 15 image degradations, 14 perception-based error measure for image restoration, 14–15 self-organizing tree map for knowledge discovery, 16–17 Brodatz database, 292 C CBIR, see Content-based image retrieval CBR, see Content-based retrieval Centroid defuzzification, 145, 154 CI, see Computational intelligence CI-inspired adaptive image restoration, 25–47 analysis, 43–44 constrained least square error, 29–35 Bayesian perspective, 30–32 Lagrangian perspective, 32–35 Index image distortions, 25–28 image restoration, 29 implementation considerations, 45 improved algorithm, 40–42 neural network restoration, 35–37 neural network restoration algorithms in literature, 37–40 neural networks as CI architecture, 25 numerical study of algorithms, 45–46 efficiency, 46 setup, 45–46 summary, 46–47 Class membership-based fusion, 306 CLS algorithm, see Constraint least squares algorithm Compressed domain analysis, 18 descriptors, 287 image categorization, see Feature representation, genetic optimization of for compressed-domain image categorization image classification, 22, 23 image retrieval, 382 processing, 288 AI-CBR System with, 287 CBR system and, 286 relevance identification and, 289 Computational intelligence (CI), 25 Computational intelligence techniques, 8–13, 315, see also Content-based image retrieval using computational intelligence techniques application of to content-based image retrieval, 22–23 evolutionary computation, 12–13 fuzzy logic, 9, 11–12 Computational reinforced learning (CRL), 197, 198–215 formulation of blind image deconvolution as evolutionary strategy, 198–205 Index network weight estimation, 201–205 probability-based stochastic initialization, 200–201 knowledge-based reinforced mutation, 205–209 dynamic mutation attractors, 205–206 reinforced mutation in attractor space, 206–209 perception-based image restoration, 210–211 performance evaluation and selection, 213–215 recombination based on niche-space residency, 212–213 Conditional probability mass function, 301 Conjugate gradient optimization, 198, 223–225 Constrained least square error, 29 Constraint least squares (CLS) algorithm, 122 Content analysis, 18 Content-based image categorization and retrieval, 18–19 content analysis, 18 relevance feedback, 18–19 Content-based image classification, 18, 299, 300 Content-based image retrieval (CBIR), 18, 314 content analysis, 18 relevance feedback, 18–19 Content-based image retrieval using computational intelligence techniques, 313–338 experimental results, 324–329 network training, 320–324 overview and structure of RFRBFN, 317–320 predictive-label fuzzy support vector machine for small sample problem, 329–337 experimental results, 335–336 overview, 330–331 training, 331–335 problem description and formulation, 315–317 355 query refinement, 314 soft relevance feedback, 317–329 Content-based retrieval (CBR), 286 Content-based retrieval, SOTM in, 286–298 architecture of AI-CBR system with compressed domain processing, 287–289 automatic interaction, 289–291 features extraction for retrieval, 291–292 features for relevance classification, 292 retrieval of texture images in compressed domain, 292–298 Cost function image-domain, 217–221 LVMSE-based, 91–99 Crisp set, 9, 143 CRL, see Computational reinforced learning Current work, contributions of, 19–23 computational intelligence, application of to content-based image retrieval, 22–23 evolutionary computation, application of to image categorization, 22 evolutionary programming, application of to adaptive regularization and blind deconvolution, 21 fuzzy set theory, application of to adaptive regularization, 20–21 neural networks, application of to edge characterization, 20 neural networks, application of for image restoration, 19–20 self-organization, application of to image analysis and retrieval, 21–22 D DCT, see Discrete cosine transform Defuzzification, 145, 154 Discrete cosine transform (DCT), 299 356 Discrete nonstationary random process, Discrete wavelet transform (DWT), 291 Distortions, spatially variant, 26 DWT, see Discrete wavelet transform Dynamic mutation attractors, 205–206 Dynamic tracking neuron, 246, 251 E Edge characterization, 15–16, 20 Edge detection, Edge detection using model-based neural networks, 239–260 experimental results, 252–258 MBNN model for edge characterization, 240–244 determination of subnetwork output, 242 edge characterization and detection, 242–244 input-parameterized model-based neuron, 240–241 network architecture, 244–249 binary edge configuration, 247–248 characterization of edge information, 245 correspondence with general HMBNN architecture, 248–249 dynamic tracking neuron, 246–247 neuron in subnetwork, 246 subnetwork, 245 recognition stage, 251–252 identification of primary edge points, 251 identification of secondary edge points, 251–252 summary, 260 training stage, 249–250 acquisition of valid edge configurations, 250 determination of pr for subnetwork, 249–250 determination of wr ∗s∗ for neuron, 250 Index Edge regularization parameter, 148 Edges, Edge-texture characterization (ETC) measure, 139, 145–147 Edge-texture fuzzy model, 145–147 EM, see Expectation maximization Empirical probability mass functions, 303 Energy minimization, 46, 62, 217 EP, see Evolutionary programming ES, see Evolutionary strategy ETC measure, see Edge-texture characterization measure Evolutionary computation, 12–13, 22 Evolutionary computation, adaptive regularization using, 169–193 adaptive regularization using evolutionary programming, 178–185 choice of optimal regularization strategy, 183–185 competition under approximate fitness criterion, 181–182 ETC-pdf image model, 174–178 evolutionary programming, 172–174 evolutionary strategy, 171–172 experimental results, 185–190 genetic algorithm, 170–171 other evolutionary approaches, 190–192 evolutionary strategy optimization, 192 hierarchical cluster model, 192 image segmentation and cluster formation, 192 summary, 193 Evolutionary programming (EP), 21, 170 Evolutionary strategy (ES), 170, 171–172 Expectation maximization (EM), 196 F FCM, see Fuzzy C-means Feature representation, genetic optimization of for compressed-domain image categorization, 299–312 Index compressed-domain representation, 301 experimental results, 307–312 multiple-classifier approach, 305–307 problem formulation, 302–305 Feedback images, 315 Fitness function, 171, 179, 303, 306 FSD, see Fundamental statistic descriptors FSF, see Fundamental statistic feature FSVM, see Fuzzy support vector machine Fundamental statistic descriptors (FSD), 291 Fundamental statistic feature (FSF), 291 Fuzzy C-means (FCM), 12, 319 Fuzzy coefficient value, 153 Fuzzy labeling, 22, 316 Fuzzy logic, 9, 11–12 Fuzzy membership function, 154 Fuzzy reasoning, 315, 331 Fuzzy set, Fuzzy set theory, 11, 20–21 Fuzzy support vector machine (FSVM), 23, 317 Fuzzy user perception, 317, 338 G GA, see Genetic algorithm Gabor wavelet transform, 292 Gaussian blur, 238 blind deconvolution and, 238 flower image degraded by, 74, 100, 111 identification of, 231 image degraded using, 46 PSF modeled by, 64 Gaussian mask, 195, 226, 231, 234 Gauss–Markov random field (GMRF), GCV, see Generalized cross validation GEM, see Greatest energy minimization Generalized cross validation (GCV), 196 Genetic algorithm (GA), 22 candidate classifier, 312 distinguishing feature, 170 357 evolutionary computation, 170 main idea of, 303 optimal transformation on random variable, 22 Genotypes, 170 GMRF, see Gauss–Markov random field Gradient-based method, 56–64 Gradient-descent algorithm, 323 Gray level, fuzzy prediction of, 151–157 definition of fuzzy estimator membership function, 151–152 defuzzification of fuzzy set G, 153–154 fuzzy inference procedure for predicted gray-level value, 152–153 regularization parameter update, 155–156 update of estimator fuzzy set width parameters, 156–157 Gray-level value, 3, Greatest energy minimization (GEM), 57 H Hard-decision approach, 315, 316 HCM, see Hierarchical cluster model Hierarchical architecture, 142, 166, 239, 260 Hierarchical cluster model (HCM), 192, 217 energy minimization, 217–221 structure and properties, 217 Hierarchical neural network for image restoration, 217–222 cluster formation and restoration, 221–222 optimization of image-domain cost function as HCM energy minimization, 217–221 structure and properties of hierarchical cluster model, 217 HMBNN, see Model-based neural network with hierarchical architecture 358 Hopfield neural network model, 127–128 Human vision system (HVS), 2, HVS, see Human vision system I IBD, see Iterative blind deconvolution Image analysis and retrieval, 31, see also Self-organization, image analysis and retrieval via Image categorization, application of evolutionary computation to, 22 Image contents, 314 Image degradations, 14 Image distortions, 25 Image feature classes, 3–5 edges, smooth regions, textures, 4–5 Image filtering, Image model, Image partition, 133, 137–138 Image restoration, 13–15, see also Perceptually motivated image restoration adaptive regularization, 14 application of neural networks for, 19–20 blind deconvolution, 15 image degradations, 14 perception-based error measure for image restoration, 14–15 Image retrieval, application of computational intelligence to content-based, 22–23 Images, important classes of features in, Impulse noise removal, SOTM in, 269–286 experimental results, 279–286 models of impulse noise, 272–274 noise-exclusive adaptive filtering, 274–279 Information hierarchy, 316 Input-parameterized model-based neuron, 240–241 Iterative blind deconvolution (IBD), 196 Index J JPEG 2000, 291 JPEG-compressed images, 301, 307 K Karush–Kuhn–Tucker theorem, 32, 33, 80, 83, 116, 122 k-means algorithm, 12 Knowledge-based reinforced mutation, 205–209 dynamic mutation attractors, 205–206 reinforced mutation in attractor space, 206–209 Knowledge discovery, self-organizing tree map for, 16–17 L Laplacian of Gaussian (LoG) filtering, 4, 240 Lexicographical mapping, 129 Local modeling, 317, 325 Local neuron output, 125, 240 Local standard deviation mean square error (LSMSE), 14, 57, 91 Local statistics analysis, 64–69 Local variance mean square error (LVMSE), 91, 122 LoG filtering, see Laplacian of Gaussian filtering Log-normal adaptation rule, 180 LSMSE, see Local standard deviation mean square error LVMSE, see Local variance mean square error M MA coefficients, see Moving average coefficients MAP estimator, see Maximum a posteriori estimator MARS, see Multimedia analysis and retrieval system Maximum a posteriori (MAP) estimator, 205 Maximum likelihood (ML), 196 Mean square error (MSE), 14 Index Membership function, 143, 146 MHI, see Multiresolution histogram indexing ML, see Maximum likelihood Model-based adaptive image restoration, 123–167 adaptive regularization, 128–133 architecture of fuzzy HMBNN, 147–149 edge-texture characterization measure, 139–142 edge-texture fuzzy model based on ETC measure, 145–147 estimation of desired network output, 149–151 ETC fuzzy HMBNN for adaptive regularization, 142–143 experimental results, 158–166 fuzzy prediction of desired gray-level value, 151–157 definition of fuzzy estimator membership function, 151–152 defuzzification of fuzzy set G, 153–154 fuzzy inference procedure for predicted gray-level value, 152–153 regularization parameter update, 155–156 update of estimator fuzzy set width parameters, 156–157 hierarchical neural network architecture, 125 HMBNN for adaptive image processing, 126–127 Hopfield neural network model for image restoration, 127–128 image partition, determination of, 137–138 model-based neural network, 123–125 model-based neural network with hierarchical architecture, 125–126 regional training set definition, 134–136 summary, 166–167 359 theory of fuzzy sets, 143–145 weight-parameterized modelbased neuron, 124–125 Model-based neural network with hierarchical architecture (HMBNN), 166, 260 Model-based neural networks, see Edge detection using model-based neural networks Motion blur, 26 Moving average (MA) coefficients, 196 MPEG-4, 291 MPEG-7, 292 MSE, see Mean square error Multimedia analysis and retrieval system (MARS), 314 Multiple classifier, 309 Multiresolution histogram indexing (MHI), 292 N Network weights, 10 Neural networks, 10–11, see also Edge detection using model-based neural networks application of to edge characterization, 20 application of for image restoration, 19–20 Hopfield model, 127–128 model-based, 123–125 Niche-space, 211 Noise, 26 Nonadaptive regularization, O Optimal transformation, 22, 307 Optimization, in adaptive image processing, P Parameter vector, 5, PCA, see Principal component analysis Penalty function model, 53 Perception-based error measure, 14 360 Perception-based image restoration, 210–211 Perceptually motivated image restoration, 89–122 implementation considerations, 105–106 local variance extension of Lagrange model, 114–121 computing local variance, 116 conditions for KKT theory to hold, 118–120 implementation considerations for Lagrangian approach, 120 numerical experiment, 121 problem formulation, 114–116 problem solution, 117–118 log LVMSE-based cost function, 100–105 analysis, 103–105 extended algorithm, 101–103 LVMSE-based cost function, 91–99 analysis, 96–99 extended algorithm, 92–96 motivation, 90–91 numerical examples, 106–114 color image restoration, 106–109 grayscale image restoration, 109 LSMSE of different algorithms, 109 robustness evaluation, 111–113 subjective survey, 113–114 summary, 122 PLFSVM, see Predictive-label fuzzy support vector machine Point spread function (PSF), 13, 25 Predictive-label fuzzy support vector machine (PLFSVM), 317, 329–337 experimental results, 335–336 overview, 330–332 training, 331–335 Predictive-labeling, 333 Principal component analysis (PCA), 241 Probability density function, 12, 21, 30, 169 Probability mass function, 301, 302 PSF, see Point spread function Index Q QBE, see Query-by-example Qualitative opinions, 16 Quantitative values, 16 Query-by-example (QBE), 18, 315 R Radial basis function (RBF) networks, 317 Random motion blur, 26 RBF networks, see Radial basis function networks Recombination, 212–213 Recursive fuzzy radial basis function network (RFRBFN), 22, 316, 317 Recursive soft decision, 198 Regularization parameter, 6, 14 strategies, 21, 184 Reinforced mutation in attractor space, 206–209 Relevance feedback (RF), 286 Relevance feedback method (RFM), 290 RF, see Relevance feedback RFM, see Relevance feedback method RFRBFN, see Recursive fuzzy radial basis function network S SDR, see Symmetrical double regularization Segmentation in adaptive image processing, problem, Self-organization, application of to image analysis and retrieval, 21–22 Self-organization, image analysis and retrieval via, 261–298 self-organizing map, 261–263 self-organizing tree map, 263–269 competitive learning algorithm, 264–267 dynamic topology and classification capability of SOTM, 267–268 Index model architecture, 263–264 summary, 268–269 SOTM in content-based retrieval, 286–298 architecture of AI-CBR system with compressed domain processing, 287–289 automatic interaction, 289–291 features extraction for retrieval, 291–292 features for relevance classification, 292 retrieval of texture images in compressed domain, 292–298 SOTM in impulse noise removal, 269–286 experimental results, 279–286 models of impulse noise, 272–274 noise-exclusive adaptive filtering, 274–279 Self-organizing map (SOM), 261 Self-organizing tree map (SOTM), 21, 261, 2633–269 competitive learning algorithm, 264–267 dynamic topology and classification capability of SOTM, 267–268 model architecture, 263–264 summary, 268–269 Semantic gap, 18, 314 Semiautomatic interaction procedure, 290 Semiblind deconvolution, 74, 80 Semiblind restoration, 74–76 Shen–Castan edge detector, 239, 258 Small sample problem, 315, 329 Soft-decision method, 215–228 blur compensation, 226–228 blur identification by conjugate gradient optimization, 223–225 hierarchical neural network for image restoration, 217–222 recursive subspace optimization, 215–217 soft parametric blur estimator, 222–223 361 Soft parametric blur estimator, 222–223 Soft relevance function, 320 SOM, see Self-organizing map SOTM, see Self-organizing tree map Space variant adaptive restoration, 73 Spatial domain, 301 Spatially adaptive image restoration, 49–88 adaptive constraint extension of Lagrange model, 80–87 conditions for KKT theory to hold, 85–87 problem formulation, 80–83 problem solution, 83–85 adaptive constraint extension of penalty function model, 53–69 gradient-based method, 56–64 local statistics analysis, 64–69 motivation, 54–56 correcting spatially variant distortion using adaptive constraints, 69–73 dealing with spatially variant distortion, 51–53 implementation considerations, 78–79 numerical examples (more), 79–80 application example, 80 efficiency, 79–80 semiblind restoration using adaptive constraints, 74–76 summary, 88 Spatially adaptive parameters, Spatially invariant degradations, 23, 47, 130 Spatially invariant distortion, 45, 52 Spatially variant degradations, 23, 47 Spatially variant distortion, 88 SPIHT, 291 Squash function, 228 Stationary process, 52 Stationary random processes, Subnetworks, 125 Subtractive clustering, 319, 320 Support vector machine (SVM), 314 SVM, see Support vector machine Symmetrical double regularization (SDR), 216 362 Index T V Texture, 4–5 Texture images, retrieval of in compressed domain, 292–298 noninteractive retrieval versus automatic interactive retrieval, 293–294 user interaction versus semiautomatic retrieval, 294–297 user subjectivity tests, 298 Threshold parameter, Toeplitz matrix, 28 Tree map, self-organizing, 16–17 Two-stage clustering, 319, 320, 322, 331 Vision importance of, 1–2 system, Visual features, 326 Visual Texture Coding (VTC), 291 VTC, see Visual Texture Coding U Uniform blur, 232–234 Unsupervised competitive learning, 10 User subjectivity, 298 W Wavelet transform technique, Gabor, 292 Weighted order statistic (WOS) filter, 134 Weighted probability density error measure, 178 Weighting mask, 45, 53 Weighting matrix, 44, 45, 51 Weight-parameterized model-based neuron, 124–125 Weight vectors, 11 WOS filter, see Weighted order statistic filter ... main emphasis is on two specific adaptive imageprocessing systems and their associated algorithms: the adaptive image- restoration algorithm and the adaptive edge-characterization Adaptive image. .. considered as a particular implementation of the stages of segmentation and characterization in the overall adaptive image- processing scheme, it can also be regarded as a self-contained adaptive image- processing.. . resolution and color In humans the senses of smell and hearing have taken second place to vision Humans have more facial muscles than any other animal, Adaptive Image Processing: A Computational Intelligence