Description
Book SynopsisA synergy of techniques on hybrid intelligence for real-life image analysis Hybrid Intelligence for Image Analysis and Understanding brings together research on the latest results and progress in the development of hybrid intelligent techniques for faithful image analysis and understanding.
Table of ContentsEditor Biographies xvii
List of Contributors xxi
Foreword xxvii
Preface xxxi
About the Companion website xxxv
1 Multilevel Image Segmentation UsingModified Genetic Algorithm (MfGA)-based Fuzzy C-Means 1
Sourav De, Sunanda Das, Siddhartha Bhattacharyya, and Paramartha Dutta
1.1 Introduction 1
1.2 Fuzzy C-Means Algorithm 5
1.3 Modified Genetic Algorithms 6
1.4 Quality Evaluation Metrics for Image Segmentation 8
1.4.1 Correlation Coefficient 8
1.4.2 Empirical Measure Q(I) 8
1.5 MfGA-Based FCM Algorithm 9
1.6 Experimental Results and Discussion 11
1.7 Conclusion 22
References 22
2 Character Recognition Using Entropy-Based Fuzzy C-Means Clustering 25
B. Kondalarao, S. Sahoo, and D.K. Pratihar
2.1 Introduction 25
2.2 Tools and Techniques Used 27
2.2.1 Fuzzy Clustering Algorithms 27
2.2.1.1 Fuzzy C-means Algorithm 28
2.2.1.2 Entropy-based Fuzzy Clustering 29
2.2.1.3 Entropy-based Fuzzy C-Means Algorithm 29
2.2.2 Sammon’s Nonlinear Mapping 30
2.3 Methodology 31
2.3.1 Data Collection 31
2.3.2 Preprocessing 31
2.3.3 Feature Extraction 32
2.3.4 Classification and Recognition 34
2.4 Results and Discussion 34
2.5 Conclusion and Future Scope ofWork 38
References 39
Appendix 41
3 A Two-Stage Approach to Handwritten Indic Script Identification 47
Pawan Kumar Singh, Supratim Das, Ram Sarkar, andMita Nasipuri
3.1 Introduction 47
3.2 Review of RelatedWork 48
3.3 Properties of Scripts Used in the PresentWork 51
3.4 ProposedWork 52
3.4.1 DiscreteWavelet Transform 53
3.4.1.1 HaarWavelet Transform 55
3.4.2 Radon Transform (RT) 57
3.5 Experimental Results and Discussion 63
3.5.1 Evaluation of the Present Technique 65
3.5.1.1 Statistical Significance Tests 66
3.5.2 Statistical Performance Analysis of SVM Classifier 68
3.5.3 Comparison with Other RelatedWorks 71
3.5.4 Error Analysis 73
3.6 Conclusion 74
Acknowledgments 75
References 75
4 Feature Extraction and Segmentation Techniques in a Static Hand Gesture Recognition System 79
Subhamoy Chatterjee, Piyush Bhandari, and Mahesh Kumar Kolekar
4.1 Introduction 79
4.2 Segmentation Techniques 81
4.2.1 Otsu Method for Gesture Segmentation 81
4.2.2 Color Space–Based Models for Hand Gesture Segmentation 82
4.2.2.1 RGB Color Space–Based Segmentation 82
4.2.2.2 HSI Color Space–Based Segmentation 83
4.2.2.3 YCbCr Color Space–Based Segmentation 83
4.2.2.4 YIQ Color Space–Based Segmentation 83
4.2.3 Robust Skin Color Region Detection Using K-Means Clustering and Mahalanobish Distance 84
4.2.3.1 Rotation Normalization 85
4.2.3.2 Illumination Normalization 85
4.2.3.3 Morphological Filtering 85
4.3 Feature Extraction Techniques 86
4.3.1 Theory of Moment Features 86
4.3.2 Contour-Based Features 88
4.4 State of the Art of Static Hand Gesture Recognition Techniques 89
4.4.1 Zoning Methods 90
4.4.2 F-Ratio-BasedWeighted Feature Extraction 90
4.4.3 Feature Fusion Techniques 91
4.5 Results and Discussion 92
4.5.1 Segmentation Result 93
4.5.2 Feature Extraction Result 94
4.6 Conclusion 97
4.6.1 FutureWork 99
Acknowledgment 99
References 99
5 SVM Combination for an Enhanced Prediction ofWriters’ Soft Biometrics 103
Nesrine Bouadjenek, Hassiba Nemmour, and Youcef Chibani
5.1 Introduction 103
5.2 Soft Biometrics and Handwriting Over Time 104
5.3 Soft Biometrics Prediction System 106
5.3.1 Feature Extraction 107
5.3.1.1 Local Binary Patterns 107
5.3.1.2 Histogram of Oriented Gradients 108
5.3.1.3 Gradient Local Binary Patterns 108
5.3.2 Classification 109
5.3.3 Fuzzy Integrals–Based Combination Classifier 111
5.3.3.1 g�� Fuzzy Measure 111
5.3.3.2 Sugeno’s Fuzzy Integral 113
5.3.3.3 Fuzzy Min-Max 113
5.4 Experimental Evaluation 113
5.4.1 Data Sets 113
5.4.1.1 IAM Data Set 113
5.4.1.2 KHATT Data Set 114
5.4.2 Experimental Setting 114
5.4.3 Gender Prediction Results 117
5.4.4 Handedness Prediction Results 117
5.4.5 Age Prediction Results 118
5.5 Discussion and Performance Comparison 118
5.6 Conclusion 120
References 121
6 Brain-Inspired Machine Intelligence for Image Analysis: Convolutional Neural Networks 127
Siddharth Srivastava and Brejesh Lall
6.1 Introduction 127
6.2 Convolutional Neural Networks 129
6.2.1 Building Blocks 130
6.2.1.1 Perceptron 134
6.2.2 Learning 135
6.2.2.1 Gradient Descent 136
6.2.2.2 Back-Propagation 136
6.2.3 Convolution 139
6.2.4 Convolutional Neural Networks:The Architecture 141
6.2.4.1 Convolution Layer 142
6.2.4.2 Pooling Layer 145
6.2.4.3 Dense or Fully Connected Layer 146
6.2.5 Considerations in Implementation of CNNs 146
6.2.6 CNN in Action 147
6.2.7 Tools for Convolutional Neural Networks 148
6.2.8 CNN Coding Examples 148
6.2.8.1 MatConvNet 148
6.2.8.2 Visualizing a CNN 149
6.2.8.3 Image Category Classification Using Deep Learning 153
6.3 Toward Understanding the Brain, CNNs, and Images 157
6.3.1 Applications 157
6.3.2 Case Studies 158
6.4 Conclusion 159
References 159
7 Human Behavioral Analysis Using Evolutionary Algorithms and Deep Learning 165
Earnest Paul Ijjina and Chalavadi Krishna Mohan
7.1 Introduction 165
7.2 Human Action Recognition Using Evolutionary Algorithms and Deep Learning 167
7.2.1 Evolutionary Algorithms for Search Optimization 168
7.2.2 Action Bank Representation for Action Recognition 168
7.2.3 Deep Convolutional Neural Network for Human Action Recognition 169
7.2.4 CNN Classifier Optimized Using Evolutionary Algorithms 170
7.3 Experimental Study 170
7.3.1 Evaluation on the UCF50 Data Set 170
7.3.2 Evaluation on the KTH Video Data Set 172
7.3.3 Analysis and Discussion 176
7.3.4 Experimental Setup and Parameter Optimization 177
7.3.5 Computational Complexity 182
7.4 Conclusions and FutureWork 183
References 183
8 Feature-Based Robust Description andMonocular Detection: An Application to Vehicle Tracking 187
Ramazan Yíldíz and Tankut Acarman
8.1 Introduction 187
8.2 Extraction of Local Features by SIFT and SURF 188
8.3 Global Features: Real-Time Detection and Vehicle Tracking 190
8.4 Vehicle Detection and Validation 194
8.4.1 X-Analysis 194
8.4.2 Horizontal Prominent Line Frequency Analysis 195
8.4.3 Detection History 196
8.5 Experimental Study 197
8.5.1 Local Features Assessment 197
8.5.2 Global Features Assessment 197
8.5.3 Local versus Global Features Assessment 201
8.6 Conclusions 201
References 202
9 A GIS Anchored Technique for Social Utility Hotspot Detection 205
Anirban Chakraborty, J.K.Mandal, Arnab Patra, and JayatraMajumdar
9.1 Introduction 205
9.2 The Technique 207
9.3 Case Study 209
9.4 Implementation and Results 221
9.5 Analysis and Comparisons 224
9.6 Conclusions 229
Acknowledgments 229
References 230
10 Hyperspectral Data Processing: Spectral Unmixing, Classification, and Target Identification 233
Vaibhav Lodhi, Debashish Chakravarty, and PabitraMitra
10.1 Introduction 233
10.2 Background and Hyperspectral Imaging System 234
10.3 Overview of Hyperspectral Image Processing 236
10.3.1 Image Acquisition 237
10.3.2 Calibration 237
10.3.3 Spatial and Spectral preprocessing 238
10.3.4 Dimension Reduction 239
10.3.4.1 Transformation-Based Approaches 239
10.3.4.2 Selection-Based Approaches 239
10.3.5 postprocessing 240
10.4 Spectral Unmixing 240
10.4.1 Unmixing Processing Chain 240
10.4.2 Mixing Model 241
10.4.2.1 Linear Mixing Model (LMM) 242
10.4.2.2 Nonlinear Mixing Model 242
10.4.3 Geometrical-Based Approaches to Linear Spectral Unmixing 243
10.4.3.1 Pure Pixel-Based Techniques 243
10.4.3.2 Minimum Volume-Based Techniques 244
10.4.4 Statistics-Based Approaches 244
10.4.5 Sparse Regression-Based Approach 245
10.4.5.1 Moore–Penrose Pseudoinverse (MPP) 245
10.4.5.2 Orthogonal Matching Pursuit (OMP) 246
10.4.5.3 Iterative Spectral Mixture Analysis (ISMA) 246
10.4.6 Hybrid Techniques 246
10.5 Classification 247
10.5.1 Feature Mining 247
10.5.1.1 Feature Selection (FS) 248
10.5.1.2 Feature Extraction 248
10.5.2 Supervised Classification 248
10.5.2.1 Minimum Distance Classifier 249
10.5.2.2 Maximum Likelihood Classifier (MLC) 250
10.5.2.3 Support Vector Machines (SVMs) 250
10.5.3 Hybrid Techniques 250
10.6 Target Detection 251
10.6.1 Anomaly Detection 251
10.6.1.1 RX Anomaly Detection 252
10.6.1.2 Subspace-Based Anomaly Detection 253
10.6.2 Signature-Based Target Detection 253
10.6.2.1 Euclidean distance 254
10.6.2.2 Spectral Angle Mapper (SAM) 254
10.6.2.3 Spectral Matched Vilter (SMF) 254
10.6.2.4 Matched Subspace Detector (MSD) 255
10.6.3 Hybrid Techniques 255
10.7 Conclusions 256
References 256
11 A Hybrid Approach for Band Selection of Hyperspectral Images 263
Aditi Roy Chowdhury, Joydev Hazra, and Paramartha Dutta
11.1 Introduction 263
11.2 Relevant Concept Revisit 266
11.2.1 Feature Extraction 266
11.2.2 Feature Selection Using 2D PCA 266
11.2.3 Immune Clonal System 267
11.2.4 Fuzzy KNN 268
11.3 Proposed Algorithm 271
11.4 Experiment and Result 271
11.4.1 Description of the Data Set 272
11.4.2 Experimental Details 274
11.4.3 Analysis of Results 275
11.5 Conclusion 278
References 279
12 Uncertainty-Based Clustering Algorithms for Medical Image Analysis 283
Deepthi P. Hudedagaddi and B.K. Tripathy
12.1 Introduction 283
12.2 Uncertainty-Based Clustering Algorithms 283
12.2.1 Fuzzy C-Means 284
12.2.2 Rough Fuzzy C-Means 285
12.2.3 Intuitionistic Fuzzy C-Means 285
12.2.4 Rough Intuitionistic Fuzzy C-Means 286
12.3 Image Processing 286
12.4 Medical Image Analysis with Uncertainty-Based Clustering Algorithms 287
12.4.1 FCM with Spatial Information for Image Segmentation 287
12.4.2 Fast and Robust FCM Incorporating Local Information for Image Segmentation 290
12.4.3 Image Segmentation Using Spatial IFCM 291
12.4.3.1 Applications of Spatial FCM and Spatial IFCM on Leukemia Images 292
12.5 Conclusions 293
References 293
13 An Optimized Breast Cancer Diagnosis SystemUsing a Cuckoo Search Algorithm and Support Vector Machine Classifier 297
Manoharan Prabukumar, Loganathan Agilandeeswari, and Arun Kumar Sangaiah
13.1 Introduction 297
13.2 Technical Background 301
13.2.1 Morphological Segmentation 301
13.2.2 Cuckoo Search Optimization Algorithm 302
13.2.3 Support Vector Machines 303
13.3 Proposed Breast Cancer Diagnosis System 303
13.3.1 Preprocessing of Breast Cancer Image 303
13.3.2 Feature Extraction 304
13.3.2.1 Geometric Features 304
13.3.2.2 Texture Features 305
13.3.2.3 Statistical Features 306
13.3.3 Features Selection 306
13.3.4 Features Classification 307
13.4 Results and Discussions 307
13.5 Conclusion 310
13.6 FutureWork 310
References 310
14 Analysis of Hand Vein Images Using Hybrid Techniques 315
R. Sudhakar, S. Bharathi, and V. Gurunathan
14.1 Introduction 315
14.2 Analysis of Vein Images in the Spatial Domain 318
14.2.1 Preprocessing 318
14.2.2 Feature Extraction 319
14.2.3 Feature-Level Fusion 320
14.2.4 Score Level Fusion 320
14.2.5 Results and Discussion 322
14.2.5.1 Evaluation Metrics 323
14.3 Analysis of Vein Images in the Frequency Domain 326
14.3.1 Preprocessing 326
14.3.2 Feature Extraction 326
14.3.3 Feature-Level Fusion 330
14.3.4 Support Vector Machine Classifier 331
14.3.5 Results and Discussion 331
14.4 Comparative Analysis of Spatial and Frequency Domain Systems 332
14.5 Conclusion 335
References 335
15 Identification of Abnormal Masses in Digital Mammogram Using Statistical Decision Making 339
Indra Kanta Maitra and Samir Kumar Bandyopadhyay
15.1 Introduction 339
15.1.1 Breast Cancer 339
15.1.2 Computer-Aided Detection/Diagnosis (CAD) 340
15.1.3 Segmentation 340
15.2 PreviousWorks 341
15.3 Proposed Method 343
15.3.1 Preparation 343
15.3.2 Preprocessing 345
15.3.2.1 Image Enhancement and Edge Detection 346
15.3.2.2 Isolation and Suppression of Pectoral Muscle 348
15.3.2.3 Breast Contour Detection 351
15.3.2.4 Anatomical Segmentation 353
15.3.3 Identification of Abnormal Region(s) 354
15.3.3.1 Coloring of Regions 354
15.3.3.2 Statistical Decision Making 355
15.4 Experimental Result 358
15.4.1 Case Study with Normal Mammogram 358
15.4.2 Case Study with Abnormalities Embedded in Fatty Tissues 358
15.4.3 Case Study with Abnormalities Embedded in Fatty-Fibro-Glandular Tissues 359
15.4.4 Case Study with Abnormalities Embedded in Dense-Fibro-Glandular Tissues 359
15.5 Result Evaluation 360
15.5.1 Statistical Analysis 361
15.5.2 ROC Analysis 361
15.5.3 Accuracy Estimation 365
15.6 Comparative Analysis 366
15.7 Conclusion 366
Acknowledgments 366
References 367
16 Automatic Detection of Coronary Artery Stenosis Using Bayesian Classification and Gaussian Filters Based on Differential Evolution 369
Ivan Cruz-Aceves, Fernando Cervantes-Sanchez, and Arturo Hernandez-Aguirre
16.1 Introduction 369
16.2 Background 370
16.2.1 Gaussian Matched Filters 371
16.2.2 Differential Evolution 371
16.2.2.1 Example: Global Optimization of the Ackley Function 373
16.2.3 Bayesian Classification 375
16.2.3.1 Example: Classification Problem 375
16.3 Proposed Method 377
16.3.1 Optimal Parameter Selection of GMF Using Differential Evolution 377
16.3.2 Thresholding of the Gaussian Filter Response 378
16.3.3 Stenosis Detection Using Second-Order Derivatives 378
16.3.4 Stenosis Detection Using Bayesian Classification 379
16.4 Computational Experiments 381
16.4.1 Results of Vessel Detection 382
16.4.2 Results of Vessel Segmentation 382
16.4.3 Evaluation of Detection of Coronary Artery Stenosis 384
16.5 Concluding Remarks 386
Acknowledgment 388
References 388
17 Evaluating the Efficacy of Multi-resolution Texture Features for Prediction of Breast Density UsingMammographic Images 391
Kriti, Harleen Kaur, and Jitendra Virmani
17.1 Introduction 391
17.1.1 Comparison of Related Methods with the Proposed Method 397
17.2 Materials and Methods 398
17.2.1 Description of Database 398
17.2.2 ROI Extraction Protocol 398
17.2.3 Workflow for CAD System Design 398
17.2.3.1 Feature Extraction 400
17.2.3.2 Classification 407
17.3 Results 410
17.3.1 Results Based on Classification Performance of the Classifiers (Classification Accuracy and Sensitivity) for Each Class 411
17.3.1.1 Experiment I: To Determine the Performance of Different FDVs Using SVM Classifier 411
17.3.1.2 Experiment II: To Determine the Performance of Different FDVs Using SSVM Classifier 412
17.3.2 Results Based on Computational Efficiency of Classifiers for Predicting 161 Instances of Testing Dataset 412
17.4 Conclusion and Future Scope 413
References 415
Index 423
