Hyperspectral Data Processing

CHEIN-I CHANG, PhD, is a Professor in the Department of Computer Science and Electrical Engineering at the University of Maryland, Baltimore County. He established the Remote Sensing Signal and Image Processing Laboratory and conducts research in designing and developing signal processing algorithms for hyperspectral imaging, medical imaging, and documentation analysis. A Fellow of IEEE and SPIE, Dr. Chang has published over 125 refereed journal articles, including more than forty papers in the IEEE Transaction on Geoscience and Remote Sensing. In addition to authoring Hyperspectral Imaging: Techniques for Spectral Detection and Classification, as well as editing two books, Hyperspectral Data Exploitation: Theory and Applications and Recent Advances in Hyperspectral Signal and Imaging Processing and co-editing one book, High Performance Computing in Remote Sensing, he holds five patents and has several pending.

• Preface xxiii1 Overview and Introduction 11.1 Overview 21.2 Issues of Multispectral and Hyperspectral Imageries 31.3 Divergence of Hyperspectral Imagery from Multispectral Imagery 41.4 Scope of This Book 71.5 Book’s Organization 101.6 Laboratory Data to be Used in This Book 191.7 Real Hyperspectral Images to be Used in this Book 201.8 Notations and Terminologies to be Used in this Book 29I: Preliminaries 312 Fundamentals of Subsample and Mixed Sample Analyses 332.1 Introduction 332.2 Subsample Analysis 352.3 Mixed Sample Analysis 452.4 Kernel-Based Classification 572.5 Conclusions 603 Three-dimensional Receiver Operating Characteristics (3d Roc) Analysis 633.1 Introduction 633.2 Neyman–Pearson Detection Problem Formulation 653.3 ROC Analysis 673.4 3D ROC Analysis 693.5 Real Data-Based ROC Analysis 723.6 Examples 783.7 Conclusions 994 Design of Synthetic Image Experiments 1014.1 Introduction 1024.2 Simulation of Targets of Interest 1034.3 Six Scenarios of Synthetic Images 1044.4 Applications 1124.5 Conclusions 1235 Virtual Dimensionality of Hyperspectral Data 1245.1 Introduction 1245.2 Reinterpretation of VD 1265.3 VD Determined by Data Characterization-Driven Criteria 1265.4 VD Determined by Data Representation-Driven Criteria 1405.5 Synthetic Image Experiments 1445.6 VD Estimated for Real Hyperspectral Images 1555.7 Conclusions 1636 Data Dimensionality Reduction 1686.1 Introduction 1686.2 Dimensionality Reduction by Second-Order Statistics-Based Component Analysis Transforms 1706.3 Dimensionality Reduction by High-Order Statistics-Based Components Analysis Transforms 1796.4 Dimensionality Reduction by Infinite-Order Statistics-Based Components Analysis Transforms 1846.5 Dimensionality Reduction by Projection Pursuit-Based Components Analysis Transforms 1906.6 Dimensionality Reduction by Feature Extraction-Based Transforms 1956.7 Dimensionality Reduction by Band Selection 1966.8 Constrained Band Selection 1976.9 Conclusions 198II: Endmember Extraction 2017 Simultaneous Endmember Extraction Algorithms (SM-EEAs) 2077.1 Introduction 2087.2 Convex Geometry-Based Endmember Extraction 2097.3 Second-Order Statistics-Based Endmember Extraction 2287.4 Automated Morphological Endmember Extraction (AMEE) 2307.5 Experiments 2317.6 Conclusions 2398 Sequential Endmember Extraction Algorithms (SQ-EEAs) 2418.1 Introduction 2418.2 Successive N-FINDR (SC N-FINDR) 2448.3 Simplex Growing Algorithm (SGA) 2448.4 Vertex Component Analysis (VCA) 2478.5 Linear Spectral Mixture Analysis-Based SQ-EEAs 2488.6 High-Order Statistics-Based SQ-EEAS 2528.7 Experiments 2548.8 Conclusions 2629 Initialization-driven Endmember Extraction Algorithms (ID-EEAs) 2659.1 Introduction 2659.2 Initialization Issues 2669.3 Initialization-Driven EEAs 2719.4 Experiments 2789.5 Conclusions 28310 Random Endmember Extraction Algorithms (REEAs) 28710.1 Introduction 28710.2 Random PPI (RPPI) 28810.3 Random VCA (RVCA) 29010.4 Random N-FINDR (RN-FINDR) 29010.5 Random SGA (RSGA) 29210.6 Random ICA-Based EEA (RICA-EEA) 29210.7 Synthetic Image Experiments 29310.8 Real Image Experiments 30510.9 Conclusions 31311 Exploration on Relationships Among Endmember Extraction Algorithms 31611.1 Introduction 31611.2 Orthogonal Projection-Based EEAs 31811.3 Comparative Study and Analysis Between SGA and VCA 33011.4 Does an Endmember Set Really Yield Maximum Simplex Volume? 33911.5 Impact of Dimensionality Reduction on EEAs 34411.6 Conclusions 348III: Supervised Linear Hyperspectral Mixture Analysis 35112 Orthogonal Subspace Projection Revisited 35512.1 Introduction 35512.2 Three Perspectives to Derive OSP 35812.3 Gaussian Noise in OSP 36412.4 OSP Implemented with Partial Knowledge 37212.5 OSP Implemented Without Knowledge 38312.6 Conclusions 39013 Fisher’s Linear Spectral Mixture Analysis 39113.1 Introduction 39113.2 Feature Vector-Constrained FLSMA (FVC-FLSMA) 39213.3 Relationship Between FVC-FLSMA and LCMV, TCIMF, and CEM 39513.4 Relationship Between FVC-FLSMA and OSP 39613.5 Relationship Between FVC-FLSMA and LCDA 39613.6 Abundance-Constrained Least Squares FLDA (ACLS-FLDA) 39713.7 Synthetic Image Experiments 39813.8 Real Image Experiments 40213.9 Conclusions 40914 Weighted Abundance-constrained Linear Spectral Mixture Analysis 41114.1 Introduction 41114.2 Abundance-Constrained LSMA (AC-LSMA) 41314.3 Weighted Least-Squares Abundance-Constrained LSMA 41314.4 Synthetic Image-Based Computer Simulations 41914.5 Real Image Experiments 42614.6 Conclusions 43215 Kernel-based Linear Spectral Mixture Analysis 43415.1 Introduction 43415.2 Kernel-Based LSMA (KLSMA) 43615.3 Synthetic Image Experiments 44115.4 AVIRIS Data Experiments 44415.5 HYDICE Data Experiments 46015.6 Conclusions 462IV: Unsupervised Hyperspectral Image Analysis 46516 Hyperspectral Measures 46916.1 Introduction 46916.2 Signature Vector-Based Hyperspectral Measures for Target Discrimanition and Identification 47016.3 Correlation-Weighted Hyperspectral Measures for Target Discrimanition and Identification 47216.4 Experiments 47716.5 Conclusions 48217 Unsupervised Linear Hyperspectral Mixture Analysis 48317.1 Introduction 48317.2 Least Squares-Based ULSMA 48617.3 Component Analysis-Based ULSMA 48817.4 Synthetic Image Experiments 49017.5 Real-Image Experiments 50317.6 ULSMA Versus Endmember Extraction 51717.7 Conclusions 52418 Pixel Extraction and Information 52618.1 Introduction 52618.2 Four Types of Pixels 52718.3 Algorithms Selected to Extract Pixel Information 52818.4 Pixel Information Analysis via Synthetic Images 52818.5 Real Image Experiments 53418.6 Conclusions 539V: Hyperspectral Information Compression 54119 Exploitation-based Hyperspectral Data Compression 54519.1 Introduction 54519.2 Hyperspectral Information Compression Systems 54719.3 Spectral/Spatial Compression 54919.4 Progressive Spectral/Spatial Compression 55719.5 3D Compression 55719.6 Exploration-Based Applications 55919.7 Experiments 56119.8 Conclusions 58020 Progressive Spectral Dimensionality Process 58120.1 Introduction 58220.2 Dimensionality Prioritization 58420.3 Representation of Transformed Components for DP 58520.4 Progressive Spectral Dimensionality Process 58920.5 Hyperspectral Compression by PSDP 59720.6 Experiments for PSDP 59820.7 Conclusions 60821 Progressive Band Dimensionality Process 61321.1 Introduction 61421.2 Band Prioritization 61521.3 Criteria for Band Prioritization 61721.4 Experiments for BP 62421.5 Progressive Band Dimensionality Process 65121.6 Hyperspectral Compresssion by PBDP 65321.7 Experiments for PBDP 65621.8 Conclusions 66222 Dynamic Dimensionality Allocation 66422.1 Introduction 66422.2 Dynamic Dimensionality Allocaction 66522.3 Signature Discriminatory Probabilties 66722.4 Coding Techniques for Determining DDA 66722.5 Experiments for Dynamic Dimensionality Allocation 66922.6 Conclusions 68223 Progressive Band Selection 68323.1 Introduction 68323.2 Band De-Corrleation 68423.3 Progressive Band Selection 68623.4 Experiments for Progressive Band Selection 68823.5 Endmember Extraction 68823.6 Land Cover/Use Classification 69023.7 Linear Spectral Mixture Analysis 69423.8 Conclusions 715VI: Hyperspectral Signal Coding 71724 Binary Coding for Spectral Signatures 71924.1 Introduction 71924.2 Binary Coding 72024.3 Spectral Feature-Based Coding 72324.4 Experiments 72524.5 Conclusions 74025 Vector Coding for Hyperspectral Signatures 74125.1 Introduction 74125.2 Spectral Derivative Feature Coding 74325.3 Spectral Feature Probabilistic Coding 75525.4 Real Image Experiments 76425.5 Conclusions 77126 Progressive Coding for Spectral Signatures 77226.1 Introduction 77226.2 Multistage Pulse Code Modulation 77426.3 MPCM-Based Progressive Spectral Signature Coding 78326.4 NIST-GAS Data Experiments 78626.5 Real Image Hyperspectral Experiments 79026.6 Conclusions 796VII: Hyperspectral Signal Characterization 79727 Variable-number Variable-band Selection for Hyperspectral Signals 79927.1 Introduction 79927.2 Orthogonal Subspace Projection-Based Band Prioritization Criterion 80127.3 Variable-Number Variable-Band Selection 80327.4 Experiments 80627.5 Selection of Reference Signatures 81927.6 Conclusions 81928 Kalman Filter-based Estimation for Hyperspectral Signals 82028.1 Introduction 82028.2 Kalman Filter-Based Linear Unmixing 82228.3 Kalman Filter-Based Spectral Characterization Signal-Processing Techniques 82428.4 Computer Simulations Using AVIRIS Data 83128.5 Computer Simulations Using NIST-Gas Data 84328.6 Real Data Experiments 85228.7 Conclusions 85729 Wavelet Representation for Hyperspectral Signals 85929.1 Introduction 85929.2 Wavelet Analysis 86029.3 Wavelet-Based Signature Characterization Algorithm 86329.4 Synthetic Image-Based Computer Simulations 86829.5 Real Image Experiments 87129.6 Conclusions 875VIII: Applications 87730 Applications of Target Detection 87930.1 Introduction 87930.2 Size Estimation of Subpixel Targets 88030.3 Experiments 88130.4 Concealed Target Detection 89130.5 Computer-Aided Detection and Classification Algorithm for Concealed Targets 89230.6 Experiments for Concealed Target Detection 89330.7 Conclusions 89531 Nonlinear Dimensionality Expansion to Multispectral Imagery 89731.1 Introduction 89731.2 Band Dimensionality Expansion 89931.3 Hyperspectral Imaging Techniques Expanded by BDE 90231.4 Feature Dimensionality Expansion by Nonlinear Kernels 90431.5 BDE in Conjunction with FDE 90931.6 Multispectral Image Experiments 90931.7 Conclusion 91832 Multispectral Magnetic Resonance Imaging 92032.1 Introduction 92032.2 Linear Spectral Mixture Analysis for MRI 92332.3 Linear Spectral Random Mixture Analysis for MRI 92832.4 Kernel-Based Linear Spectral Mixture Analysis 93332.5 Synthetic MR Brain Image Experiments 93332.6 Real MR Brain Image Experiments 95132.7 Conclusions 95533 Conclusions 95633.1 Design Principles for Nonliteral Hyperspectral Imaging Techniques 95633.2 Endmember Extraction 96433.3 Linear Spectral Mixture Analysis 97033.4 Anomaly Detection 97433.5 Support Vector Machines and Kernel-Based Approaches 97733.6 Hyperspectral Compression 98133.7 Hyperspectral Signal Processing 98433.8 Applications 98733.9 Further Topics 987Glossary 993Appendix: Algorithm Compendium 997References 1052Index 1071

“I make a strong recommendation to anyone interested in hyperspectral image processing, and hyperspectral signal processing to make this book a common reference.”  (Photogrammetric Engineering and Remote Sensing, 1 June 2015)