Hybrid Intelligence for Image Analysis and Understanding

PROF. (DR.) SIDDHARTHA BHATTACHARYYA (SMIEEE, SMACM, LMCSI, LMOSI, LMISTE, MIAENG, MIRSS, MACSE, MIAASSE) obtained his Bachelors in Physics, Optics and Optoelectronics and Masters in Optics and Optoelectronics from the University of Calcutta, India, in 1995, 1998 and 2000 respectively. He completed a PhD in Computer Science and Engineering from Jadavpur University, India, in 2008. He is currently the Professor and Head of Information Technology at the RCC Institute of Information Technology, Kolkata, India. He is also the Dean (Research & Development) of the institute. He is a co-author of 3 books and co-editor of 5 books and more than 135 research publications. DR. INDRAJIT PAN did his Bachelors in Computer Science Engineering in 2005 at The University of Burdwan, India, and completed his Masters in Information Technology at Bengal Engineering and Science University, Shibpur. He got a University Medal for his performance in his Masters. Later, he was awarded a PhD in Engineering from the Indian Institute of Engineering, Science and Technology (IIEST). He has more than 10 years' experience teaching in undergraduate and postgraduate engineering in IT and allied field. Currently, he is an Assistant Professor of Information Technology at the RCC Institute of Information Technology. His research interests include CAD, Computer Security, Soft Computing Applications and Cloud Computing. DR. ANIRBAN MUKHERJEE did his Bachelors in Civil Engineering in 1994 at Jadavpur University, Kolkata. He completed his PhD on 'Automatic Diagram Drawing based on Natural Language Text Understanding' at the Indian Institute of Engineering, Science and Technology (IIEST), Shibpur, in 2014. He has more than 20 years' experience in teaching undergraduate and postgraduate engineering in IT and allied field. Currently, he is an Associate Professor and HOD of Engineering Science & Management at the RCC Institute of Information Technology. He has experience of working in computer aided design and engineering analysis and also of teaching on CAD courses. His research interests include Computer Graphics & CAD, Soft Computing Applications and Assistive Technology. He has co-authored two UG engineering textbooks: a popular one on 'Computer Graphics and Multimedia' and another on 'Engineering Mechanics'. He has also co-authored more than 15 books on Computer Graphics/Multimedia for distance learning professional courses at different Universities in India. PROF. (DR.) PARAMARTHA DUTTA has a B. Stat. (Hons.), M. Stat., M. Tech in Computer Science, and a PhD (Engineering) in Computer Science and Technology. With around 23 years of research and academic experience, Professor Dutta is currently serving as a Professor in the Department of Computer and System Sciences, Visva Bharati University. Professor Dutta is a senior Member of IEEE and ACM. He has executed almost 200 projects funded by the Govt. of India. Professor Dutta has remained associated with various Universities and Institutes as Visiting/Guest faculty. To date, Professor Dutta has more than 6 authored and 6 edited books in addition to around 180 papers, published in different International Journals and in International/National conference proceedings.

• Editor Biographies xviiList of Contributors xxiForeword xxviiPreface xxxiAbout the Companion website xxxv1 Multilevel Image Segmentation UsingModified Genetic Algorithm (MfGA)-based Fuzzy C-Means 1Sourav De, Sunanda Das, Siddhartha Bhattacharyya, and Paramartha Dutta1.1 Introduction 11.2 Fuzzy C-Means Algorithm 51.3 Modified Genetic Algorithms 61.4 Quality Evaluation Metrics for Image Segmentation 81.4.1 Correlation Coefficient 81.4.2 Empirical Measure Q(I) 81.5 MfGA-Based FCM Algorithm 91.6 Experimental Results and Discussion 111.7 Conclusion 22References 222 Character Recognition Using Entropy-Based Fuzzy C-Means Clustering 25B. Kondalarao, S. Sahoo, and D.K. Pratihar2.1 Introduction 252.2 Tools and Techniques Used 272.2.1 Fuzzy Clustering Algorithms 272.2.1.1 Fuzzy C-means Algorithm 282.2.1.2 Entropy-based Fuzzy Clustering 292.2.1.3 Entropy-based Fuzzy C-Means Algorithm 292.2.2 Sammon’s Nonlinear Mapping 302.3 Methodology 312.3.1 Data Collection 312.3.2 Preprocessing 312.3.3 Feature Extraction 322.3.4 Classification and Recognition 342.4 Results and Discussion 342.5 Conclusion and Future Scope ofWork 38References 39Appendix 413 A Two-Stage Approach to Handwritten Indic Script Identification 47Pawan Kumar Singh, Supratim Das, Ram Sarkar, andMita Nasipuri3.1 Introduction 473.2 Review of RelatedWork 483.3 Properties of Scripts Used in the PresentWork 513.4 ProposedWork 523.4.1 DiscreteWavelet Transform 533.4.1.1 HaarWavelet Transform 553.4.2 Radon Transform (RT) 573.5 Experimental Results and Discussion 633.5.1 Evaluation of the Present Technique 653.5.1.1 Statistical Significance Tests 663.5.2 Statistical Performance Analysis of SVM Classifier 683.5.3 Comparison with Other RelatedWorks 713.5.4 Error Analysis 733.6 Conclusion 74Acknowledgments 75References 754 Feature Extraction and Segmentation Techniques in a Static Hand Gesture Recognition System 79Subhamoy Chatterjee, Piyush Bhandari, and Mahesh Kumar Kolekar4.1 Introduction 794.2 Segmentation Techniques 814.2.1 Otsu Method for Gesture Segmentation 814.2.2 Color Space–Based Models for Hand Gesture Segmentation 824.2.2.1 RGB Color Space–Based Segmentation 824.2.2.2 HSI Color Space–Based Segmentation 834.2.2.3 YCbCr Color Space–Based Segmentation 834.2.2.4 YIQ Color Space–Based Segmentation 834.2.3 Robust Skin Color Region Detection Using K-Means Clustering and Mahalanobish Distance 844.2.3.1 Rotation Normalization 854.2.3.2 Illumination Normalization 854.2.3.3 Morphological Filtering 854.3 Feature Extraction Techniques 864.3.1 Theory of Moment Features 864.3.2 Contour-Based Features 884.4 State of the Art of Static Hand Gesture Recognition Techniques 894.4.1 Zoning Methods 904.4.2 F-Ratio-BasedWeighted Feature Extraction 904.4.3 Feature Fusion Techniques 914.5 Results and Discussion 924.5.1 Segmentation Result 934.5.2 Feature Extraction Result 944.6 Conclusion 974.6.1 FutureWork 99Acknowledgment 99References 995 SVM Combination for an Enhanced Prediction ofWriters’ Soft Biometrics 103Nesrine Bouadjenek, Hassiba Nemmour, and Youcef Chibani5.1 Introduction 1035.2 Soft Biometrics and Handwriting Over Time 1045.3 Soft Biometrics Prediction System 1065.3.1 Feature Extraction 1075.3.1.1 Local Binary Patterns 1075.3.1.2 Histogram of Oriented Gradients 1085.3.1.3 Gradient Local Binary Patterns 1085.3.2 Classification 1095.3.3 Fuzzy Integrals–Based Combination Classifier 1115.3.3.1 g�� Fuzzy Measure 1115.3.3.2 Sugeno’s Fuzzy Integral 1135.3.3.3 Fuzzy Min-Max 1135.4 Experimental Evaluation 1135.4.1 Data Sets 1135.4.1.1 IAM Data Set 1135.4.1.2 KHATT Data Set 1145.4.2 Experimental Setting 1145.4.3 Gender Prediction Results 1175.4.4 Handedness Prediction Results 1175.4.5 Age Prediction Results 1185.5 Discussion and Performance Comparison 1185.6 Conclusion 120References 1216 Brain-Inspired Machine Intelligence for Image Analysis: Convolutional Neural Networks 127Siddharth Srivastava and Brejesh Lall6.1 Introduction 1276.2 Convolutional Neural Networks 1296.2.1 Building Blocks 1306.2.1.1 Perceptron 1346.2.2 Learning 1356.2.2.1 Gradient Descent 1366.2.2.2 Back-Propagation 1366.2.3 Convolution 1396.2.4 Convolutional Neural Networks:The Architecture 1416.2.4.1 Convolution Layer 1426.2.4.2 Pooling Layer 1456.2.4.3 Dense or Fully Connected Layer 1466.2.5 Considerations in Implementation of CNNs 1466.2.6 CNN in Action 1476.2.7 Tools for Convolutional Neural Networks 1486.2.8 CNN Coding Examples 1486.2.8.1 MatConvNet 1486.2.8.2 Visualizing a CNN 1496.2.8.3 Image Category Classification Using Deep Learning 1536.3 Toward Understanding the Brain, CNNs, and Images 1576.3.1 Applications 1576.3.2 Case Studies 1586.4 Conclusion 159References 1597 Human Behavioral Analysis Using Evolutionary Algorithms and Deep Learning 165Earnest Paul Ijjina and Chalavadi Krishna Mohan7.1 Introduction 1657.2 Human Action Recognition Using Evolutionary Algorithms and Deep Learning 1677.2.1 Evolutionary Algorithms for Search Optimization 1687.2.2 Action Bank Representation for Action Recognition 1687.2.3 Deep Convolutional Neural Network for Human Action Recognition 1697.2.4 CNN Classifier Optimized Using Evolutionary Algorithms 1707.3 Experimental Study 1707.3.1 Evaluation on the UCF50 Data Set 1707.3.2 Evaluation on the KTH Video Data Set 1727.3.3 Analysis and Discussion 1767.3.4 Experimental Setup and Parameter Optimization 1777.3.5 Computational Complexity 1827.4 Conclusions and FutureWork 183References 1838 Feature-Based Robust Description andMonocular Detection: An Application to Vehicle Tracking 187Ramazan Yíldíz and Tankut Acarman8.1 Introduction 1878.2 Extraction of Local Features by SIFT and SURF 1888.3 Global Features: Real-Time Detection and Vehicle Tracking 1908.4 Vehicle Detection and Validation 1948.4.1 X-Analysis 1948.4.2 Horizontal Prominent Line Frequency Analysis 1958.4.3 Detection History 1968.5 Experimental Study 1978.5.1 Local Features Assessment 1978.5.2 Global Features Assessment 1978.5.3 Local versus Global Features Assessment 2018.6 Conclusions 201References 2029 A GIS Anchored Technique for Social Utility Hotspot Detection 205Anirban Chakraborty, J.K.Mandal, Arnab Patra, and JayatraMajumdar9.1 Introduction 2059.2 The Technique 2079.3 Case Study 2099.4 Implementation and Results 2219.5 Analysis and Comparisons 2249.6 Conclusions 229Acknowledgments 229References 23010 Hyperspectral Data Processing: Spectral Unmixing, Classification, and Target Identification 233Vaibhav Lodhi, Debashish Chakravarty, and PabitraMitra10.1 Introduction 23310.2 Background and Hyperspectral Imaging System 23410.3 Overview of Hyperspectral Image Processing 23610.3.1 Image Acquisition 23710.3.2 Calibration 23710.3.3 Spatial and Spectral preprocessing 23810.3.4 Dimension Reduction 23910.3.4.1 Transformation-Based Approaches 23910.3.4.2 Selection-Based Approaches 23910.3.5 postprocessing 24010.4 Spectral Unmixing 24010.4.1 Unmixing Processing Chain 24010.4.2 Mixing Model 24110.4.2.1 Linear Mixing Model (LMM) 24210.4.2.2 Nonlinear Mixing Model 24210.4.3 Geometrical-Based Approaches to Linear Spectral Unmixing 24310.4.3.1 Pure Pixel-Based Techniques 24310.4.3.2 Minimum Volume-Based Techniques 24410.4.4 Statistics-Based Approaches 24410.4.5 Sparse Regression-Based Approach 24510.4.5.1 Moore–Penrose Pseudoinverse (MPP) 24510.4.5.2 Orthogonal Matching Pursuit (OMP) 24610.4.5.3 Iterative Spectral Mixture Analysis (ISMA) 24610.4.6 Hybrid Techniques 24610.5 Classification 24710.5.1 Feature Mining 24710.5.1.1 Feature Selection (FS) 24810.5.1.2 Feature Extraction 24810.5.2 Supervised Classification 24810.5.2.1 Minimum Distance Classifier 24910.5.2.2 Maximum Likelihood Classifier (MLC) 25010.5.2.3 Support Vector Machines (SVMs) 25010.5.3 Hybrid Techniques 25010.6 Target Detection 25110.6.1 Anomaly Detection 25110.6.1.1 RX Anomaly Detection 25210.6.1.2 Subspace-Based Anomaly Detection 25310.6.2 Signature-Based Target Detection 25310.6.2.1 Euclidean distance 25410.6.2.2 Spectral Angle Mapper (SAM) 25410.6.2.3 Spectral Matched Vilter (SMF) 25410.6.2.4 Matched Subspace Detector (MSD) 25510.6.3 Hybrid Techniques 25510.7 Conclusions 256References 25611 A Hybrid Approach for Band Selection of Hyperspectral Images 263Aditi Roy Chowdhury, Joydev Hazra, and Paramartha Dutta11.1 Introduction 26311.2 Relevant Concept Revisit 26611.2.1 Feature Extraction 26611.2.2 Feature Selection Using 2D PCA 26611.2.3 Immune Clonal System 26711.2.4 Fuzzy KNN 26811.3 Proposed Algorithm 27111.4 Experiment and Result 27111.4.1 Description of the Data Set 27211.4.2 Experimental Details 27411.4.3 Analysis of Results 27511.5 Conclusion 278References 27912 Uncertainty-Based Clustering Algorithms for Medical Image Analysis 283Deepthi P. Hudedagaddi and B.K. Tripathy12.1 Introduction 28312.2 Uncertainty-Based Clustering Algorithms 28312.2.1 Fuzzy C-Means 28412.2.2 Rough Fuzzy C-Means 28512.2.3 Intuitionistic Fuzzy C-Means 28512.2.4 Rough Intuitionistic Fuzzy C-Means 28612.3 Image Processing 28612.4 Medical Image Analysis with Uncertainty-Based Clustering Algorithms 28712.4.1 FCM with Spatial Information for Image Segmentation 28712.4.2 Fast and Robust FCM Incorporating Local Information for Image Segmentation 29012.4.3 Image Segmentation Using Spatial IFCM 29112.4.3.1 Applications of Spatial FCM and Spatial IFCM on Leukemia Images 29212.5 Conclusions 293References 29313 An Optimized Breast Cancer Diagnosis SystemUsing a Cuckoo Search Algorithm and Support Vector Machine Classifier 297Manoharan Prabukumar, Loganathan Agilandeeswari, and Arun Kumar Sangaiah13.1 Introduction 29713.2 Technical Background 30113.2.1 Morphological Segmentation 30113.2.2 Cuckoo Search Optimization Algorithm 30213.2.3 Support Vector Machines 30313.3 Proposed Breast Cancer Diagnosis System 30313.3.1 Preprocessing of Breast Cancer Image 30313.3.2 Feature Extraction 30413.3.2.1 Geometric Features 30413.3.2.2 Texture Features 30513.3.2.3 Statistical Features 30613.3.3 Features Selection 30613.3.4 Features Classification 30713.4 Results and Discussions 30713.5 Conclusion 31013.6 FutureWork 310References 31014 Analysis of Hand Vein Images Using Hybrid Techniques 315R. Sudhakar, S. Bharathi, and V. Gurunathan14.1 Introduction 31514.2 Analysis of Vein Images in the Spatial Domain 31814.2.1 Preprocessing 31814.2.2 Feature Extraction 31914.2.3 Feature-Level Fusion 32014.2.4 Score Level Fusion 32014.2.5 Results and Discussion 32214.2.5.1 Evaluation Metrics 32314.3 Analysis of Vein Images in the Frequency Domain 32614.3.1 Preprocessing 32614.3.2 Feature Extraction 32614.3.3 Feature-Level Fusion 33014.3.4 Support Vector Machine Classifier 33114.3.5 Results and Discussion 33114.4 Comparative Analysis of Spatial and Frequency Domain Systems 33214.5 Conclusion 335References 33515 Identification of Abnormal Masses in Digital Mammogram Using Statistical Decision Making 339Indra Kanta Maitra and Samir Kumar Bandyopadhyay15.1 Introduction 33915.1.1 Breast Cancer 33915.1.2 Computer-Aided Detection/Diagnosis (CAD) 34015.1.3 Segmentation 34015.2 PreviousWorks 34115.3 Proposed Method 34315.3.1 Preparation 34315.3.2 Preprocessing 34515.3.2.1 Image Enhancement and Edge Detection 34615.3.2.2 Isolation and Suppression of Pectoral Muscle 34815.3.2.3 Breast Contour Detection 35115.3.2.4 Anatomical Segmentation 35315.3.3 Identification of Abnormal Region(s) 35415.3.3.1 Coloring of Regions 35415.3.3.2 Statistical Decision Making 35515.4 Experimental Result 35815.4.1 Case Study with Normal Mammogram 35815.4.2 Case Study with Abnormalities Embedded in Fatty Tissues 35815.4.3 Case Study with Abnormalities Embedded in Fatty-Fibro-Glandular Tissues 35915.4.4 Case Study with Abnormalities Embedded in Dense-Fibro-Glandular Tissues 35915.5 Result Evaluation 36015.5.1 Statistical Analysis 36115.5.2 ROC Analysis 36115.5.3 Accuracy Estimation 36515.6 Comparative Analysis 36615.7 Conclusion 366Acknowledgments 366References 36716 Automatic Detection of Coronary Artery Stenosis Using Bayesian Classification and Gaussian Filters Based on Differential Evolution 369Ivan Cruz-Aceves, Fernando Cervantes-Sanchez, and Arturo Hernandez-Aguirre16.1 Introduction 36916.2 Background 37016.2.1 Gaussian Matched Filters 37116.2.2 Differential Evolution 37116.2.2.1 Example: Global Optimization of the Ackley Function 37316.2.3 Bayesian Classification 37516.2.3.1 Example: Classification Problem 37516.3 Proposed Method 37716.3.1 Optimal Parameter Selection of GMF Using Differential Evolution 37716.3.2 Thresholding of the Gaussian Filter Response 37816.3.3 Stenosis Detection Using Second-Order Derivatives 37816.3.4 Stenosis Detection Using Bayesian Classification 37916.4 Computational Experiments 38116.4.1 Results of Vessel Detection 38216.4.2 Results of Vessel Segmentation 38216.4.3 Evaluation of Detection of Coronary Artery Stenosis 38416.5 Concluding Remarks 386Acknowledgment 388References 38817 Evaluating the Efficacy of Multi-resolution Texture Features for Prediction of Breast Density UsingMammographic Images 391Kriti, Harleen Kaur, and Jitendra Virmani17.1 Introduction 39117.1.1 Comparison of Related Methods with the Proposed Method 39717.2 Materials and Methods 39817.2.1 Description of Database 39817.2.2 ROI Extraction Protocol 39817.2.3 Workflow for CAD System Design 39817.2.3.1 Feature Extraction 40017.2.3.2 Classification 40717.3 Results 41017.3.1 Results Based on Classification Performance of the Classifiers (Classification Accuracy and Sensitivity) for Each Class 41117.3.1.1 Experiment I: To Determine the Performance of Different FDVs Using SVM Classifier 41117.3.1.2 Experiment II: To Determine the Performance of Different FDVs Using SSVM Classifier 41217.3.2 Results Based on Computational Efficiency of Classifiers for Predicting 161 Instances of Testing Dataset 41217.4 Conclusion and Future Scope 413References 415Index 423