Applications of Evolutionary Computation in Image Processing and Pattern Recognition (eBook)

Foreword 6
Preface 8
Contents 12
1 Introduction 17
Abstract 17
1.1 Definition of an Optimization Problem 17
1.2 Classical Optimization 18
1.3 Evolutionary Computation Methods 21
1.3.1 Structure of an Evolutionary Computation Algorithm 22
References 24
2 Image Segmentation Based on Differential Evolution Optimization 25
Abstract 25
2.1 Introduction 25
2.2 Gaussian Approximation 26
2.3 Differential Evolution Algorithms 27
2.4 Determination of Thresholding Values 29
2.5 Experimental Results 30
2.6 Conclusions 36
References 37
3 Motion Estimation Based on Artificial Bee Colony (ABC) 39
Abstract 39
3.1 Introduction 39
3.2 Artificial Bee Colony (ABC) Algorithm 43
3.2.1 Biological Bee Profile 43
3.2.2 Description of the ABC Algorithm 43
3.2.3 Initializing the Population 44
3.2.4 Send Employed Bees 44
3.2.5 Select the Food Sources by the Onlooker Bees 45
3.2.6 Determine the Scout Bees 45
3.3 Fitness Approximation Method 45
3.3.1 Updating the Individual Database 46
3.3.2 Fitness Calculation Strategy 46
3.3.3 Presented ABC Optimization Method 48
3.4 Motion Estimation and Block Matching 50
3.5 BM Algorithm Based on ABC with the Estimation Strategy 51
3.5.1 Initial Population 52
3.5.2 The ABC-BM Algorithm 54
3.6 Experimental Results 55
3.6.1 ABC-BM Results 55
3.6.1.1 Distortion Performance 57
3.6.1.2 Search Efficiency 59
3.6.2 Results on H.264 59
3.6.3 Experiments with High Definition Sequences 62
3.7 Conclusions 64
References 65
4 Ellipse Detection on Images Inspired by the Collective Animal Behavior 68
Abstract 68
4.1 Introduction 68
4.2 Collective Animal Behavior Algorithm (CAB) 71
4.2.1 Description of the CAB Algorithm 71
4.2.1.1 Initializing the Population 71
4.2.1.2 Keep the Position of the Best Individuals 72
4.2.1.3 Move from or to Nearby Neighbors 72
4.2.1.4 Move Randomly 73
4.2.1.5 Compete for the Space Within of a Determined Distance (Update the Memory) 73
4.2.1.6 Computational Procedure 74
4.3 Ellipse Detection Using CAB 75
4.3.1 Data Preprocessing 75
4.3.2 Individual Representation 75
4.3.3 Objective Function 77
4.3.4 Implementation of CAB for Ellipse Detection 79
4.4 The Multiple Ellipse Detection Procedure 80
4.5 Experimental Results 81
4.5.1 Ellipse Localization 82
4.5.1.1 Synthetic Images 82
4.5.1.2 Natural Images 82
4.5.2 Shape Discrimination Tests 83
4.5.3 Ellipse Approximation: Occluded Ellipse and Ellipsoidal Detection 84
4.5.4 Performance Comparison 84
4.6 Conclusions 90
References 91
5 Template Matching by Using the States of Matter Algorithm 93
Abstract 93
5.1 Introduction 93
5.2 States of Matter 95
5.3 States of Matter Search (SMS) 97
5.3.1 Definition of Operators 97
5.3.1.1 Direction Vector 97
5.3.1.2 Collision 98
5.3.1.3 Random Positions 99
5.3.1.4 Best Element Updating 99
5.3.2 SMS Algorithm 100
5.3.2.1 General Procedure 100
5.3.2.2 The Complete Algorithm 100
5.3.2.3 Initialization 102
5.3.2.4 Gas State 103
5.3.2.5 Liquid State 103
5.3.2.6 Solid State 104
5.4 Fitness Approximation Method 104
5.4.1 Updating Individual Database 105
5.4.2 Fitness Calculation Strategy 105
5.4.3 Presented Optimization SMS Method 108
5.5 Template Matching Process 109
5.6 TM Algorithm Based on SMS with the Estimation Strategy 110
5.6.1 The SMS-TM Algorithm 111
5.7 Experimental Results 113
5.8 Conclusions 117
References 118
6 Estimation of Multiple View Relations Considering Evolutionary Approaches 120
Abstract 120
6.1 Introduction 120
6.2 View Relations from Point Correspondences 123
6.3 Random Sampling Consensus (RANSAC) Algorithm 126
6.4 Clonal Selection Algorithm (CSA) 127
6.4.1 Definitions 127
6.4.2 CSA Operators 128
6.4.3 Clonal Proliferation Operator (T_{/rm{P}}^{/rm{C}} ) 128
6.4.4 Affinity Maturation Operator (T_{M}^{/rm{A}} ) 129
6.4.5 Clonal Selection Operator (T_{/rm{S}}^{/rm{C}} ) 130
6.5 Method for Geometric Estimation Using CSA 130
6.5.1 Computational Procedure 132
6.6 Experimental Results 135
6.6.1 Fundamental Matrix Estimation with Synthetic Data 136
6.6.2 Fundamental Matrix Estimation with Real Images 139
6.6.3 Homography Estimation with Synthetic Data 142
6.6.4 Homography Estimation with Real Images 145
6.7 Conclusions 147
References 149
7 Circle Detection on Images Based on an Evolutionary Algorithm that Reduces the Number of Function Evaluations 152
Abstract 152
7.1 Introduction 153
7.2 The Adaptive Population with Reduced Evaluations (APRE) Algorithm 155
7.2.1 Initialization 156
7.2.2 Selecting the Population to Be Evolved 157
7.2.3 The Number of Elements N_{e}^{k} to Be Selected 157
7.2.4 Selection Strategy for Building {{/bf P}}^{k} 159
7.2.5 Exploration Operation 160
7.2.6 DE Mutation Operator 161
7.2.7 Trigonometric Mutation Operator 161
7.2.7.1 Computational Procedure 161
7.2.8 Fitness Estimation Strategy 163
7.2.9 Memory Updating 166
7.2.10 Exploitation Operation 166
7.2.11 Computational Procedure 168
7.3 Implementation of APRE-Based Circle Detector 169
7.3.1 Individual Representation 169
7.3.2 Objective Function 171
7.3.3 The Multiple Circle Detection Procedure 171
7.4 Results on Multi-circle Detection 172
7.5 Conclusions 178
References 179
8 Otsu and Kapur Segmentation Based on Harmony Search Optimization 181
Abstract 181
8.1 Introduction 181
8.2 Harmony Search Algorithm 183
8.2.1 The Harmony Search Algorithm 183
8.2.1.1 Initializing the Problem and the Algorithm Parameters 184
8.2.1.2 Harmony Memory Initialization 184
8.2.1.3 Improvisation of New Harmony Vectors 185
8.2.1.4 Updating the Harmony Memory 185
8.2.1.5 Computational Procedure 186
8.3 Image Multilevel Thresholding (MT) 186
8.3.1 Between---Class Variance (Otsu's Method) 187
8.3.2 Entropy Criterion Method (Kapur's Method) 189
8.4 Multilevel Thresholding Using Harmony Search Algorithm (HSMA) 191
8.4.1 Harmony Representation 191
8.4.2 HMA Implementation 191
8.4.3 Parameter Setting 193
8.5 Experimental Results 194
8.5.1 Otsu's Results 196
8.5.2 Kapur's Results 196
8.5.3 Comparisons 202
8.5.3.1 Comparison Between Otsu and Kapur HSMA 205
8.5.3.2 Comparison Among HSMA and Other MT Approaches 207
8.6 Conclusions 210
References 213
9 Leukocyte Detection by Using Electromagnetism-like Optimization 215
Abstract 215
9.1 Introduction 216
9.2 Electromagnetism-like Optimization Algorithm (EMO) 218
9.3 Circle Detection Using EMO 221
9.3.1 Data Preprocessing 221
9.3.2 Particle Representation 221
9.3.3 Objective Function 222
9.3.4 EMO Implementation 223
9.4 The White Blood Cell Detector 225
9.4.1 Image Preprocessing 226
9.4.2 The Modified EMO-Based Circle Detector 227
9.4.3 Numerical Example 229
9.5 Experimental Results 231
9.6 Comparisons to Other Methods 232
9.6.1 Detection Comparison 232
9.6.2 Robustness Comparison 233
9.6.3 Stability Comparison 234
9.7 Conclusions 237
References 238
10 Automatic Segmentation by Using an Algorithm Based on the Behavior of Locust Swarms 240
Abstract 240
10.1 Introduction 240
10.2 Biological Fundamentals and Mathematical Models 243
10.2.1 Solitary Phase 244
10.2.2 Social Phase 246
10.3 The Locust Search (LS) Algorithm 246
10.3.1 Solitary Operation (A) 248
10.3.2 Social Operation (B) 250
10.3.3 Complete LS Algorithm 252
10.3.4 Discussion About the LS Algorithm 253
10.4 Numerical Experiments over Benchmark Functions 254
10.4.1 Uni-modal Test Functions 255
10.4.2 Multimodal Test Functions 257
10.5 Gaussian Mixture Modelling 258
10.6 Segmentation Algorithm Based on LS 259
10.6.1 New Objective Function Jnew 260
10.6.2 Complete Segmentation Algorithm 262
10.7 Segmentation Results 264
10.7.1 Performance of LS Algorithm in Image Segmentation 264
10.7.1.1 First Image 264
10.7.1.2 Second Image 265
10.7.2 Histogram Approximation Comparisons 267
10.7.2.1 Convergence 269
10.7.2.2 Accuracy 270
10.7.2.3 Computational Cost 272
10.7.3 Performance Evaluation of the Segmentation Results 274
10.7.3.1 Evaluation Criteria 275
10.7.3.2 Experimental Protocol 275
10.8 Conclusions 277
References 278
Appendix A: RANSAC Algorithm 281
Appendix B: List of Benchmark Functions 283