HYPERSPECTRAL DATA PROCESSING: Algorithm Design and Analysis -- CONTENTS -- PREFACE -- 1 OVERVIEWAND INTRODUCTION -- 1.1 Overview -- 1.2 Issues of Multispectral and Hyperspectral Imageries -- 1.3 Divergence of Hyperspectral Imagery from Multispectral Imagery -- 1.3.1 Misconception: Hyperspectral Imaging is a Natural Extension of Multispectral Imaging -- 1.3.2 Pigeon-Hole Principle: Natural Interpretation of Hyperspectral Imaging -- 1.4 Scope of This Book -- 1.5 Book's Organization -- 1.5.1 Part I: Preliminaries -- 1.5.2 Part II: Endmember Extraction -- 1.5.3 Part III: Supervised Linear Hyperspectral Mixture Analysis -- 1.5.4 Part IV: Unsupervised Hyperspectral Analysis -- 1.5.5 Part V: Hyperspectral Information Compression -- 1.5.6 Part VI: Hyperspectral Signal Coding -- 1.5.7 Part VII: Hyperspectral Signal Feature Characterization -- 1.5.8 Applications -- 1.5.8.1 Chapter 30: Applications of Target Detection -- 1.5.8.2 Chapter 31: Nonlinear Dimensionality Expansion to Multispectral Imagery -- 1.5.8.3 Chapter 32: Multispectral Magnetic Resonance Imaging -- 1.6 Laboratory Data to be Used in This Book -- 1.6.1 Laboratory Data -- 1.6.2 Cuprite Data -- 1.6.3 NIST/EPA Gas-Phase Infrared Database -- 1.7 Real Hyperspectral Images to be Used in this Book -- 1.7.1 AVIRIS Data -- 1.7.1.1 Cuprite Data -- 1.7.1.2 Purdue's Indiana Indian Pine Test Site -- 1.7.2 HYDICE Data -- 1.8 Notations and Terminologies to be Used in this Book -- I: PRELIMINARIES -- 2 FUNDAMENTALS OF SUBSAMPLE AND MIXED SAMPLE ANALYSES -- 2.1 Introduction -- 2.2 Subsample Analysis -- 2.2.1 Pure-Sample Target Detection -- 2.2.2 Subsample Target Detection -- 2.2.2.1 Adaptive Matched Detector (AMD) -- 2.2.2.2 Adaptive Subspace Detector (ASD) -- 2.2.3 Subsample Target Detection: Constrained Energy Minimization (CEM) -- 2.3 Mixed Sample Analysis -- 2.3.1 Classification with Hard Decisions.

2.3.1.1 Fisher's Linear Discriminant Analysis (FLDA) -- 2.3.1.2 Support Vector Machines (SVM) -- 2.3.2 Classification with Soft Decisions -- 2.3.2.1 Orthogonal Subspace Projection (OSP) -- 2.3.2.2 Target-Constrained Interference-Minimized Filter (TCIMF) -- 2.4 Kernel-Based Classification -- 2.4.1 Kernel Trick Used in Kernel-Based Methods -- 2.4.2 Kernel-Based Fisher's Linear Discriminant Analysis (KFLDA) -- 2.4.3 Kernel Support Vector Machine (K-SVM) -- 2.5 Conclusions -- 3 THREE-DIMENSIONAL RECEIVER OPERATING CHARACTERISTICS (3D ROC) ANALYSIS -- 3.1 Introduction -- 3.2 Neyman-Pearson Detection Problem Formulation -- 3.3 ROC Analysis -- 3.4 3D ROC Analysis -- 3.5 Real Data-Based ROC Analysis -- 3.5.1 How to Generate ROC Curves from Real Data -- 3.5.2 How to Generate Gaussian-Fitted ROC Curves -- 3.5.3 How to Generate 3D ROC Curves -- 3.5.4 How to Generate 3D ROC Curves for Multiple Signal Detection and Classification -- 3.6 Examples -- 3.6.1 Hyperspectral Imaging -- 3.6.1.1 Hyperspectral Target Detection -- 3.6.1.2 Linear Hyperspectral Mixture Analysis -- 3.6.2 Magnetic Resonance (MR) Breast Imaging -- 3.6.2.1 Breast Tumor Detection -- 3.6.2.2 Brain Tissue Classification -- 3.6.3 Chemical/Biological Agent Detection -- 3.6.4 Biometric Recognition -- 3.7 Conclusions -- 4 DESIGN OF SYNTHETIC IMAGE EXPERIMENTS -- 4.1 Introduction -- 4.2 Simulation of Targets of Interest -- 4.2.1 Simulation of Synthetic Subsample Targets -- 4.2.2 Simulation of Synthetic Mixed-Sample Targets -- 4.3 Six Scenarios of Synthetic Images -- 4.3.1 Panel Simulations -- 4.3.2 Three Scenarios for Target Implantation (TI) -- 4.3.2.1 Scenario TI1 (Clean Panels Implanted into Clean Background) -- 4.3.2.2 Scenario TI2 (Clean Panels Implanted into Noisy Background) -- 4.3.2.3 Scenario TI3 (Gaussian Noise Added to Clean Panels Implanted into Clean Background).

4.3.3 Three Scenarios for Target Embeddedness (TE) -- 4.3.3.1 Scenario TE1 (Clean Panels Embedded in Clean Background) -- 4.3.3.2 Scenario TE2 (Clean Panels Embedded in Noisy Background) -- 4.3.3.3 Scenario TE3 (Gaussian Noise Added to Clean Panels Embedded in Background) -- 4.4 Applications -- 4.4.1 Endmember Extraction -- 4.4.2 Linear Spectral Mixture Analysis (LSMA) -- 4.4.2.1 Mixed Pixel Classification -- 4.4.2.2 Mixed Pixel Quantification -- 4.4.3 Target Detection -- 4.4.3.1 Subpixel Target Detection -- 4.4.3.2 Anomaly Detection -- 4.5 Conclusions -- 5 VIRTUAL DIMENSIONALITY OF HYPERSPECTRAL DATA -- 5.1 Introduction -- 5.2 Reinterpretation of VD -- 5.3 VD Determined by Data Characterization-Driven Criteria -- 5.3.1 Eigenvalue Distribution-Based Criteria -- 5.3.1.1 Thresholding Energy Percentage -- 5.3.1.2 Thresholding Difference between Normalized Correlation Eigenvalues and Normalized Covariance Eigenvalues -- 5.3.1.3 Finding First Sudden Drop in the Normalized Eigenvalue Distribution -- 5.3.2 Eigen-Based Component Analysis Criteria -- 5.3.2.1 Singular Value Decomposition (SVD) -- 5.3.2.2 Principal Components Analysis (PCA) -- 5.3.3 Factor Analysis: Malinowski's Error Theory -- 5.3.4 Information Theoretic Criteria (ITC) -- 5.3.4.1 AIC -- 5.3.4.2 MDL -- 5.3.5 Gershgorin Radius-Based Methods -- 5.3.5.1 Thresholding Gershgorin Radii -- 5.3.5.2 Thresholding Difference Gershgorin Radii between RLxL and KLxL -- 5.3.6 HFC Method -- 5.3.7 Discussions on Data Characterization-Driven Criteria -- 5.4 VD Determined by Data Representation-Driven Criteria -- 5.4.1 Orthogonal Subspace Projection (OSP) -- 5.4.2 Signal Subspace Estimation (SSE) -- 5.4.3 Discussions on OSP and SSE/HySime -- 5.5 Synthetic Image Experiments -- 5.5.1 Data Characterization-Driven Criteria -- 5.5.1.1 Target Implantation (TI) Scenarios -- 5.5.1.2 Target Embeddedness (TE) Scenarios.

5.5.2 Data Representation-Driven Criteria -- 5.6 VD Estimated for Real Hyperspectral Images -- 5.7 Conclusions -- 6 DATA DIMENSIONALITY REDUCTION -- 6.1 Introduction -- 6.2 Dimensionality Reduction by Second-Order Statistics-Based Component Analysis Transforms -- 6.2.1 Eigen Component Analysis Transforms -- 6.2.1.1 Principal Components Analysis -- 6.2.1.2 Standardized Principal Components Analysis -- 6.2.1.3 Singular Value Decomposition -- 6.2.2 Signal-to-Noise Ratio-Based Components Analysis Transforms -- 6.2.2.1 Maximum Noise Fraction Transform -- 6.2.2.2 Noise-Adjusted Principal Component Transform -- 6.3 Dimensionality Reduction by High-Order Statistics-Based Components Analysis Transforms -- 6.3.1 Sphering -- 6.3.2 Third-Order Statistics-Based Skewness -- 6.3.3 Fourth-Order Statistics-Based Kurtosis -- 6.3.4 High-Order Statistics -- 6.3.5 Algorithm for Finding Projection Vectors -- 6.4 Dimensionality Reduction by Infinite-Order Statistics-Based Components Analysis Transforms -- 6.4.1 Statistics-Prioritized ICA-DR (SPICA-DR) -- 6.4.2 Random ICA-DR -- 6.4.3 Initialization Driven ICA-DR -- 6.5 Dimensionality Reduction by Projection Pursuit-Based Components Analysis Transforms -- 6.5.1 Projection Index-Based Projection Pursuit -- 6.5.2 Random Projection Index-Based Projection Pursuit -- 6.5.3 Projection Index-Based Prioritized Projection Pursuit -- 6.5.4 Initialization Driven Projection Pursuit -- 6.6 Dimensionality Reduction by Feature Extraction-Based Transforms -- 6.6.1 Fisher's Linear Discriminant Analysis -- 6.6.2 Orthogonal Subspace Projection -- 6.7 Dimensionality Reduction by Band Selection -- 6.8 Constrained Band Selection -- 6.9 Conclusions -- II: ENDMEMBER EXTRACTION -- 7 SIMULTANEOUS ENDMEMBER EXTRACTION ALGORITHMS (SM-EEAs) -- 7.1 Introduction -- 7.2 Convex Geometry-Based Endmember Extraction.

7.2.1 Convex Geometry-Based Criterion: Orthogonal Projection -- 7.2.2 Convex Geometry-Based Criterion: Minimal Simplex Volume -- 7.2.2.1 Minimal-Volume Transform (MVT) -- 7.2.2.2 Convex Cone Analysis (CCA) -- 7.2.3 Convex Geometry-Based Criterion: Maximal Simplex Volume -- 7.2.3.1 Simultaneous N-FINDR (SM N-FINDR) -- 7.2.3.2 Iterative N-FINDR (IN-FINDR) -- 7.2.3.3 Various Versions of Implementing IN-FINDR -- 7.2.3.4 Discussions on Various Implementation Versions of IN-FINDR -- 7.2.3.5 Comparative Study Among Various Versions of IN-FINDR -- 7.2.3.6 Alternative SM N-FINDR -- 7.2.4 Convex Geometry-Based Criterion: Linear Spectral Mixture Analysis -- 7.3 Second-Order Statistics-Based Endmember Extraction -- 7.4 Automated Morphological Endmember Extraction (AMEE) -- 7.5 Experiments -- 7.5.1 Synthetic Image Experiments -- 7.5.1.1 Scenario TI1 (Endmembers Implanted in a Clean Background) -- 7.5.1.2 Scenario TI2 (Endmembers Implanted in a Noisy Background) -- 7.5.1.3 Scenario TI3 (Noisy Endmembers Implanted in a Noisy Background) -- 7.5.1.4 Scenario TE1 (Endmembers Embedded into a Clean Background) -- 7.5.1.5 Scenario TE2 (Endmembers Embedded into a Noisy Background) -- 7.5.1.6 Scenario TE3 (Noisy Endmembers Embedded into a Noisy Background) -- 7.5.2 Cuprite Data -- 7.5.3 HYDICE Data -- 7.6 Conclusions -- 8 SEQUENTIAL ENDMEMBER EXTRACTION ALGORITHMS (SQ-EEAs) -- 8.1 Introduction -- 8.2 Successive N-FINDR (SC N-FINDR) -- 8.3 Simplex Growing Algorithm (SGA) -- 8.4 Vertex Component Analysis (VCA) -- 8.5 Linear Spectral Mixture Analysis-Based SQ-EEAs -- 8.5.1 Automatic Target Generation Process-EEA (ATGP-EEA) -- 8.5.2 Unsupervised Nonnegativity Constrained Least-Squares-EEA (UNCLS-EEA) -- 8.5.3 Unsupervised Fully Constrained Least-Squares-EEA (UFCLS-EEA) -- 8.5.4 Iterative Error Analysis-EEA (IEA-EEA) -- 8.6 High-Order Statistics-Based SQ-EEAS.

8.6.1 Third-Order Statistics-Based SQ-EEA.