Hyperspectral Data Processing E-Book

Chein-I Chang

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 1900

Veröffentlichungsjahr: 2013

Beschreibung

Hyperspectral Data Processing: Algorithm Design and Analysis is a culmination of the research conducted in the Remote Sensing Signal and Image Processing Laboratory (RSSIPL) at the University of Maryland, Baltimore County. Specifically, it treats hyperspectral image processing and hyperspectral signal processing as separate subjects in two different categories. Most materials covered in this book can be used in conjunction with the author's first book, Hyperspectral Imaging: Techniques for Spectral Detection and Classification, without much overlap. Many results in this book are either new or have not been explored, presented, or published in the public domain. These include various aspects of endmember extraction, unsupervised linear spectral mixture analysis, hyperspectral information compression, hyperspectral signal coding and characterization, as well as applications to conceal target detection, multispectral imaging, and magnetic resonance imaging. Hyperspectral Data Processing contains eight major sections: * Part I: provides fundamentals of hyperspectral data processing * Part II: offers various algorithm designs for endmember extraction * Part III: derives theory for supervised linear spectral mixture analysis * Part IV: designs unsupervised methods for hyperspectral image analysis * Part V: explores new concepts on hyperspectral information compression * Parts VI & VII: develops techniques for hyperspectral signal coding and characterization * Part VIII: presents applications in multispectral imaging and magnetic resonance imaging Hyperspectral Data Processing compiles an algorithm compendium with MATLAB codes in an appendix to help readers implement many important algorithms developed in this book and write their own program codes without relying on software packages. Hyperspectral Data Processing is a valuable reference for those who have been involved with hyperspectral imaging and its techniques, as well those who are new to the subject.

Leseprobe

Contents

Cover

Title Page

Copyright

Dedication

Preface

Chapter 1: Overview and Introduction

1.1 Overview

1.2 Issues of Multispectral and Hyperspectral Imageries

1.3 Divergence of Hyperspectral Imagery from Multispectral Imagery

1.4 Scope of This Book

1.5 Book's Organization

1.6 Laboratory Data to be Used in This Book

1.7 Real Hyperspectral Images to be Used in this Book

1.8 Notations and Terminologies to be Used in this Book

I: Preliminaries

Chapter 2: Fundamentals of Subsample and Mixed Sample Analyses

2.1 Introduction

2.2 Subsample Analysis

2.3 Mixed Sample Analysis

2.4 Kernel-Based Classification

2.5 Conclusions

Chapter 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

3.6 Examples

3.7 Conclusions

Chapter 4: Design of Synthetic Image Experiments

4.1 Introduction

4.2 Simulation of Targets of Interest

4.3 Six Scenarios of Synthetic Images

4.4 Applications

4.5 Conclusions

Chapter 5: Virtual Dimensionality of Hyperspectral Data

5.1 Introduction

5.2 Reinterpretation of VD

5.3 VD Determined by Data Characterization-Driven Criteria

5.4 VD Determined by Data Representation-Driven Criteria

5.5 Synthetic Image Experiments

5.6 VD Estimated for Real Hyperspectral Images

5.7 Conclusions

Chapter 6: Data Dimensionality Reduction

6.1 Introduction

6.2 Dimensionality Reduction by Second-Order Statistics-Based Component Analysis Transforms

6.3 Dimensionality Reduction by High-Order Statistics-Based Components Analysis Transforms

6.4 Dimensionality Reduction by Infinite-Order Statistics-Based Components Analysis Transforms

6.5 Dimensionality Reduction by Projection Pursuit-Based Components Analysis Transforms

6.6 Dimensionality Reduction by Feature Extraction-Based Transforms

6.7 Dimensionality Reduction by Band Selection

6.8 Constrained Band Selection

6.9 Conclusions

II: Endmember Extraction

Chapter 7: Simultaneous Endmember Extraction Algorithms (SM-EEAs)

7.1 Introduction

7.2 Convex Geometry-Based Endmember Extraction

7.3 Second-Order Statistics-Based Endmember Extraction

7.4 Automated Morphological Endmember Extraction (AMEE)

7.5 Experiments

7.6 Conclusions

Chapter 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

8.6 High-Order Statistics-Based SQ-EEAS

8.7 Experiments

8.8 Conclusions

Chapter 9: Initialization-Driven Endmember Extraction Algorithms (ID-EEAs)

9.1 Introduction

9.2 Initialization Issues

9.3 Initialization-Driven EEAs

9.4 Experiments

9.5 Conclusions

Chapter 10: Random Endmember Extraction Algorithms (REEAs)

10.1 Introduction

10.2 Random PPI (RPPI)

10.3 Random VCA (RVCA)

10.4 Random N-FINDR (RN-FINDR)

10.5 Random SGA (RSGA)

10.6 Random ICA-Based EEA (RICA-EEA)

10.7 Synthetic Image Experiments

10.8 Real Image Experiments

10.9 Conclusions

Chapter 11: Exploration on Relationships among Endmember Extraction Algorithms

11.1 Introduction

11.2 Orthogonal Projection-Based EEAs

11.3 Comparative Study and Analysis Between SGA and VCA

11.4 Does an Endmember Set Really Yield Maximum Simplex Volume?

11.5 Impact of Dimensionality Reduction on EEAs

11.6 Conclusions

III: Supervised Linear Hyperspectral Mixture Analysis

Chapter 12: Orthogonal Subspace Projection Revisited

12.1 Introduction

12.2 Three Perspectives to Derive OSP

12.3 Gaussian Noise in OSP

12.4 OSP Implemented with Partial Knowledge

12.5 OSP Implemented Without Knowledge

12.6 Conclusions

Chapter 13: Fisher's Linear Spectral Mixture Analysis

13.1 Introduction

13.2 Feature Vector-Constrained FLSMA (FVC-FLSMA)

13.3 Relationship Between FVC-FLSMA and LCMV, TCIMF, and CEM

13.4 Relationship Between FVC-FLSMA and OSP

13.5 Relationship Between FVC-FLSMA and LCDA

13.6 Abundance-Constrained Least Squares FLDA (ACLS-FLDA)

13.7 Synthetic Image Experiments

13.8 Real Image Experiments

13.9 Conclusions

Chapter 14: Weighted Abundance-Constrained Linear Spectral Mixture Analysis

14.1 Introduction

14.2 Abundance-Constrained LSMA (AC-LSMA)

14.3 Weighted Least-Squares Abundance-Constrained LSMA

14.4 Synthetic Image-Based Computer Simulations

14.5 Real Image Experiments

14.6 Conclusions

Chapter 15: Kernel-Based Linear Spectral Mixture Analysis

15.1 Introduction

15.2 Kernel-Based LSMA (KLSMA)

15.3 Synthetic Image Experiments

15.4 AVIRIS Data Experiments

15.5 HYDICE Data Experiments

15.6 Conclusions

IV: Unsupervised Hyperspectral Image Analysis

Chapter 16: Hyperspectral Measures

16.1 Introduction

16.2 Signature Vector-Based Hyperspectral Measures for Target Discrimanition and Identification

16.3 Correlation-Weighted Hyperspectral Measures for Target Discrimanition and Identification

16.4 Experiments

16.5 Conclusions

Chapter 17: Unsupervised Linear Hyperspectral Mixture Analysis

17.1 Introduction

17.2 Least Squares-Based ULSMA

17.3 Component Analysis-Based ULSMA

17.4 Synthetic Image Experiments

17.5 Real-Image Experiments

17.6 ULSMA Versus Endmember Extraction

17.7 Conclusions

Chapter 18: Pixel Extraction and Information

18.1 Introduction

18.2 Four Types of Pixels

18.3 Algorithms Selected to Extract Pixel Information

18.4 Pixel Information Analysis via Synthetic Images

18.5 Real Image Experiments

18.6 Conclusions

V: Hyperspectral Information Compression

Chapter 19: Exploitation-Based Hyperspectral Data Compression

19.1 Introduction

19.2 Hyperspectral Information Compression Systems

19.3 Spectral/Spatial Compression

19.4 Progressive Spectral/Spatial Compression

19.5 3D Compression

19.6 Exploration-Based Applications

19.7 Experiments

19.8 Conclusions

Chapter 20: Progressive Spectral Dimensionality Process

20.1 Introduction

20.2 Dimensionality Prioritization

20.3 Representation of Transformed Components for DP

20.4 Progressive Spectral Dimensionality Process

20.5 Hyperspectral Compression by PSDP

20.6 Experiments for PSDP

20.7 Conclusions

Chapter 21: Progressive Band Dimensionality Process

21.1 Introduction

21.2 Band Prioritization

21.3 Criteria for Band Prioritization

21.4 Experiments for BP

21.5 Progressive Band Dimensionality Process

21.6 Hyperspectral Compresssion by PBDP

21.7 Experiments for PBDP

21.8 Conclusions

Chapter 22: Dynamic Dimensionality Allocation

22.1 Introduction

22.2 Dynamic Dimensionality Allocaction

22.3 Signature Discriminatory Probabilties

22.4 Coding Techniques for Determining DDA

22.5 Experiments for Dynamic Dimensionality Allocation

22.6 Conclusions

Chapter 23: Progressive Band Selection

23.1 Introduction

23.2 Band De-Corrleation

23.3 Progressive Band Selection

23.4 Experiments for Progressive Band Selection

23.5 Endmember Extraction

23.6 Land Cover/Use Classification

23.7 Linear Spectral Mixture Analysis

23.8 Conclusions

VI: Hyperspectral Signal Coding

Chapter 24: Binary Coding For Spectral Signatures

24.1 Introduction

24.2 Binary Coding

24.3 Spectral Feature-Based Coding

24.4 Experiments

24.5 Conclusions

Chapter 25: Vector Coding for Hyperspectral Signatures

25.1 Introduction

25.2 Spectral Derivative Feature Coding

25.3 Spectral Feature Probabilistic Coding

25.4 Real Image Experiments

25.5 Conclusions

Chapter 26: Progressive Coding for Spectral Signatures

26.1 Introduction

26.2 Multistage Pulse Code Modulation

26.3 MPCM-Based Progressive Spectral Signature Coding

26.4 NIST-GAS Data Experiments

26.5 Real Image Hyperspectral Experiments

26.6 Conclusions

VII: Hyperspectral Signal Characterization

Chapter 27: Variable-Number Variable-Band Selection for Hyperspectral Signals

27.1 Introduction

27.2 Orthogonal Subspace Projection-Based Band Prioritization Criterion

27.3 Variable-Number Variable-Band Selection

27.4 Experiments

27.5 Selection of Reference Signatures

27.6 Conclusions

Chapter 28: Kalman Filter-Based Estimation for Hyperspectral Signals

28.1 Introduction

28.2 Kalman Filter-Based Linear Unmixing

28.3 Kalman Filter-Based Spectral Characterization Signal-Processing Techniques

28.4 Computer Simulations Using AVIRIS Data

28.5 Computer Simulations Using NIST-Gas Data

28.6 Real Data Experiments

28.7 Conclusions

Chapter 29: Wavelet Representation for Hyperspectral Signals

29.1 Introduction

29.2 Wavelet Analysis

29.3 Wavelet-Based Signature Characterization Algorithm

29.4 Synthetic Image-Based Computer Simulations

29.5 Real Image Experiments

29.6 Conclusions

VIII: Applications

Chapter 30: Applications of Target Detection

30.1 Introduction

30.2 Size Estimation of Subpixel Targets

30.3 Experiments

30.4 Concealed Target Detection

30.5 Computer-Aided Detection and Classification Algorithm for Concealed Targets

30.6 Experiments for Concealed Target Detection

30.7 Conclusions

Chapter 31: Nonlinear Dimensionality Expansion to Multispectral Imagery

31.1 Introduction

31.2 Band Dimensionality Expansion

31.3 Hyperspectral Imaging Techniques Expanded by BDE

31.4 Feature Dimensionality Expansion by Nonlinear Kernels

31.5 BDE in Conjunction with FDE

31.6 Multispectral Image Experiments

31.7 Conclusion

Chapter 32: Multispectral Magnetic Resonance Imaging

32.1 Introduction

32.2 Linear Spectral Mixture Analysis for MRI

32.3 Linear Spectral Random Mixture Analysis for MRI

32.4 Kernel-Based Linear Spectral Mixture Analysis

32.5 Synthetic MR Brain Image Experiments

32.6 Real MR Brain Image Experiments

32.7 Conclusions

Chapter 33: Conclusions

33.1 Design Principles for Nonliteral Hyperspectral Imaging Techniques

33.2 Endemember Extraction

33.3 Linear Spectral Mixture Analysis

33.4 Anomaly Detection

33.5 Support Vector Machines and Kernel-Based Approaches

33.6 Hyperspectral Compression

33.7 Hyperspectral Signal Processing

33.8 Applications

33.9 Further Topics

Glossary

Appendix: Algorithm Compendium

References

Index

Copyright © 2013 by John Wiley & Sons, Inc. All rights reserved

Published by John Wiley & Sons, Inc., Hoboken, NJ

Published simultaneously in Canada

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, [phone], fax [phone], or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, [phone], fax [phone], or online at http://www.wiley.com/go/permission.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at [phone], outside the United States at [phone] or fax [phone].

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Chang, Chein-I.

Hyperspectral data processing : algorithm design and analysis / Chein-I Chang.

p. cm.

Includes bibliographical references and index.

ISBN 978-0-471-69056-6 (hardback)

1. Image processing–Digital techniques. 2. Spectroscopic imaging. 3. Signal processing. I. Chang, Chein-I. Hyperspectral imaging. II. Title.

TA1637.C4776 2012

621.39'94–dc23

2011043896

This book is dedicated to members of my family, specifically my mother who provided me with her timeless support and encouragement during the course of preparing this book. It is also dedicated to all of my students who have contributed to this book.

Preface

Hyperspectral imaging has witnessed tremendous growth over the past few years. Still its applications to new areas are yet to be explored. Many hyperspectral imaging techniques have been developed and reported in various venues. My first book, Hyperspectral Imaging: Techniques for Spectral Detection and Classification, referenced as Chang (2003a), was written in an attempt to summarize the research conducted at that time in my laboratory (remote sensing signal and image processing laboratory, RSSIPL) and to provide readers with a peek of this fascinating and exciting area. With rapid advancement in this area many signal processing techniques have been developed for hyperspectral signal and image processing. This book has been written with four goals in mind. One is to continuously explore new statistical signal processing algorithms in this area for various applications. Many results in this book are new, particularly some in Chapters 2, 4, 5–6, 11, 16, 18–19, 23, 24, 29, 30–31, and 33. A second goal is to supplement Chang (2003a), where many potential research efforts were only briefly mentioned (in Chapter 18 of the book). A third goal is to distinguish this book from Chang (2003a) in many ways. Unlike Chang (2003a) where the main theme was hyperspectral target detection and classification from a viewpoint of subpixel and mixed pixel analysis, this book is focused on more in-depth treatment of hyperspectral signal and image processing from a statistical signal processing point of view. A fourth and last goal is to focus on several unsettled but very important issues that have been avoided and never addressed in the past.

One issue is “how many spectral signatures are required to unmix data?” arising in linear hyperspectral unmixing. This has been a long-standing and unresolved issue in remote sensing image processing, specifically hyperspectral imaging, since the number of signatures to be used for data unmixing has a significant impact on image analysis while its accurate number is never known in real applications. Another is “how many pure spectral signatures, referred to as endmembers, are supposed to be present in the data to be processed?” It is common practice to assume that the number of signatures used for spectral unmixing is the same number of endmembers. Unfortunately, such a claim, which has been widely accepted by the community, is not always true in practical applications (see Chapter 17). The issue of endmembers has not received much interest in multispectral image analysis because of its low spectral and spatial resolutions that generally result in mixed data sample vectors. However, due to recent advances in hyperspectral imaging sensors with hundreds of contiguous spectral bands endmember extraction has become increasingly important since endmembers provide crucial “nonliteral” information in spectral interpretation, characterization, and analysis. Interestingly, this issue has never been seriously addressed until recently when it has been investigated by a series of papers (Chang, 2006ab; Chang and Plaza, 2006; Chang et al., 2006; Plaza and Chang 2006) by introducing a new concept of virtual dimensionality (VD). Besides, some controversial issues result from misinterpreting VD. Therefore, one of the major chapters in this book is Chapter 5, which revisits VD to explore its utility in various applications. Unlike the intrinsic dimensionality (ID), also known as effective dimensionality (ED), which is somewhat abstract and defined as the minimum number of parameters to represent general high-dimensional multivariate data, VD is more practical and realistic. It is defined as the number of “spectrally” distinct signatures particularly developed for hyperspectral data in which the non-literal (spectral) information is more crucial and vital than information provided by other dimensions such as spatial information. In particular, an issue arises in how to define the spectral distinction among signatures in VD estimation. Furthermore, unlike ID that is a one-size-fits-all definition for all data sets, VD should adapt to data sets used for different applications as well as vary with the techniques used to estimate VD. In order to address this issue, Chapter 5 explores two types of VD criteria, data characterization-driven criteria and data representation-driven criteria, to define spectrally distinct signatures, and further decouples the concept of VD from the techniques used to estimate VD. Consequently, when VD is poorly estimated by one technique for a particular data set, it is not the definition of VD to be blamed, but rather the technique used for VD estimation that is not applicable to this particular data set. In addition, an issue related to VD is “characterization of pixel information.” For example, an anomaly is not necessarily an endmember and vice versa. So, the issues “what is the distinction between these two?” and “how do we characterize these two?” become interesting issues in hyperspectral data exploitation to be discussed in Chapter 18.

Another interesting topic presented in this book is a new concept of “hyperspectral information compression” introduced in Chapters 19–23. It is different from the commonly used so-called hyperspectral data compression in the sense that hyperspectral information compression is generally performed based on the information required to be retained rather than the size of hyperspectral data to be compressed. Therefore, a more appropriate term to be used is “exploitation-based lossy hyperspectral data compression.” Nevertheless, it should be noted that the definitions and terminologies used in these chapters are by no means standard.

Finally, an issue of “multispectral imagery versus hyperspectral imagery” is also investigated. It seems that there is no cut-and-dried definition to distinguish these two terminologies. A general understanding of distinguishing these two is that a hyperspectral image is acquired by hundreds of contiguous spectral channels/bands with very high spectral resolution, while a multispectral image is collected by tens of discrete spectral channels/bands with low spectral resolution. If this interpretation is used, we run into a dilemma, “how many spectral channels/bands are enough for a remotely sensed image to be called a hyperspectral image?” or “how fine the spectral resolution should be for a remote sensing image to be considered as a hyperspectral image?” For example, if we take a small set of hyperspectral band images with spectral resolution 10 nm, say five spectral band images, to form a five-dimensional image cube, do we still consider this new-formed five-dimensional image cube as a hyperspectral image or simply a multispectral image? If we adopt the former definition based on the number of bands, this five-dimensional image cube should be viewed as a multispectral image. On the other hand, if we adopt the latter definition based on spectral resolution, the five-dimensional image cube should be considered as a hyperspectral image. Thus far, it seems that there is no general consensus on this issue. In Chapter 31, an attempt is made to address this issue from a viewpoint of how two versions of independent component analysis (ICA), over-complete ICA, and under-complete ICA can be used to resolve this long-debated issue in the context of linear spectral mixture analysis (LSMA). After all, some of these issues may never be settled or standardized for years to come. Many researchers can always argue differently at their discretion and provide their own versions of interpretation. I have no intention of disputing any of them, but rather respect their opinions.

Since processing hyperspectral signatures as one-dimensional signals and processing hyperspectral images as three-dimensional image cubes are rather different, this book makes a distinction by treating hyperspectral image processing and hyperspectral signal processing in two separate categories to avoid confusion. To this end, three categories are specifically outlined in this book: Category A: hyperspectral image processing; Category B: hyperspectral signal processing; and Category C: applications.

For better understanding, a set of six chapters is included in PART I as preliminaries that cover fundamentals and provide a basic background required for readers to follow algorithm design and development discussed in this book. Category A is made up of 15 chapters (Chapters 7–23) treated separately in four different parts, Part II to Part V. Category B consists of six chapters (Chapters 24–29) in two separate parts, Part VI and Part VII. Finally, applications make up Category C.

It is worth noting that many materials presented in this book have been only available after Chang (2003a). Theses include endmember extraction (Chapters 7–11), algorithm design using different levels of information (supervised linear hyperspectral mixture analysis in Chapters 12–15), pixel characterization and analysis (unsupervised hyperspectral analysis in Chapters 16–18), exploitation-based hyperspectral information compression (Chapters 19–23), hyperspectral signature coding and characterization (Chapters 24–29), and applications (Chapters 30–32) in Category C.

There are three unique features in this book that cannot be found in Chang (2003a): (1) Part I: preliminaries (Chapters 2–6); (2) extensive studies of synthetic image-based experiments for performance evaluation; and (3) an appendix on algorithm compendium that compiles recently developed signal processing algorithms developed in the RSSIPL, all of which are believed to be useful and beneficial to those who design and develop algorithms for hyperspectral signal/image processing. Because this book also addresses many issues that were not explored in Chang (2003a), it can be used in conjunction with Chang (2003a) without much overlap, where the latter provides necessary basic background in design and development of statistical signal processing algorithms for hyperspectral image analysis, especially for subpixel detection and mixed pixel classification. Therefore, on one end, those who have been involved in hyperspectral imaging and are familiar with hyperspectral imaging techniques will find this book useful as reference material. On the other end, those who are new will find this book a good and valuable guide on the topics that may interest them.

I would like to thank the Spectral Information Technology Applications Center (SITAC) that provides its HYDICE data to be used for experiments in this book. I would also like to acknowledge the use of Purdue's Indiana Indian Pine test site and the AVIRIS Cuprite image data website.

I owe my sincere gratitude and deepest appreciation to my former Ph.D. students, Drs. Sumit Chakravarty, Hsian-Min Chen, Yingzi Du, Qian Du, Mingkai Hsueh, Baohoing Ji, Xiaoli Jiao, Keng-Hao Liu, Weimin Liu, Bharath Ramakrishna, Hsuan Ren, Haleh Safavi, Chiun-Mu Wang, Jianwei Wang, Jing Wang, Su Wang, Englin Wong, Chao-Cheng Wu, Wei Xiong, and MS student, Ms. Farzeen Chaudhary as well as my current Ph.D. student, Shih-Yu Chen. My appreciation is also extended to my colleagues, Professor Chinsu Lin with the Department of Forestry and Natural Resources at National Chiayi University, Dr. Ching Wen Yang who is the Director of Computer Center, Taichung Veterans General Hospital, and Professor Ching Tsorng Tsai with the Computer Science Department at Tunghai University. I would like to thank particularly my former Ph.D. students, Dr. Chao-Cheng Wu who carried out most of the experiments presented in Chapters 7–11, Dr. Ken-Hao Liu who performed many experiments described in Chapters 21–23, Dr. Su Wang who did all the work mentioned in Chapter 29, Dr. Englin Wong who performed all the experiments described in Chapter 32, and Professor Antonio J. Plaza who contributed to some part of Chapter 18 when he was on sabbatical leave from the Computer Science Department, University of Extremadura, Spain, in 2004 to visit my laboratory. This book could not have been completed without their contributions.

I would also like to thank the Ministry of Education in Taiwan for supporting me as a Distinguished Lecture Chair within the Department of Electrical Engineering from 2005 to 2006, a Chair Professorship of Reduction Technology within the Environmental Restoration and Disaster Reduction Research Center and Department of Electrical Engineering from 2006 to 2009, and a Chair Professorship of Remote Sensing Technology within the Department of Electrical Engineering from 2009 to 2012, at National Chung Hsing University where Professor Yen-Chieh Ouyang of Electrical Engineering has been a very supportive host during my visit. In particular, during the period 2009–2010, I was on sabbatical leave from UMBC to visit National Chung Hsing University where my appointment as a distinguished visiting fellow/fellow professor was supported and funded by the National Science Council in Taiwan under projects of NSC 98-2811-E-005-024 and NSC 98-2221-E-005-096. All their support is highly appreciated.

Last but not least, I would also like to thank my friends, Dr. San-Kan Lee (Deputy Superintendent of Taichung Veterans General Hospital (TCVGH)), Dr. Clayton Chi-Chang Chen (Chairman of Radiology at TCVGH), Dr. Jyh-Wen Chai (Section Chief of Radiology at TCVGH), and Dr. Yong Kie Wong (Head of Dental Department at TCVGH) who have selflessly provided their expertise and resources, especially an excellent testbed environment to help me use hyperspectral imaging techniques in magnetic resonance imaging (MRI). Chapter 32 is indeed a culmination of such a great working relationship.

As a final note, I would like to share that this book was supposed to be delivered by 2008. The most important factor that caused the delay is the urge to include the latest reports on hyperspectral data analysis. It is very difficult and challenging to keep a track of such new developments. Nevertheless, this book has grown three times larger than what I had originally proposed. Those who are interested in my forthcoming 2013 book can have a quick peek of these topics briefly discussed in Chapter 33, which includes a new development of target-characterized virtual dimensionality (VD), real-time and progressive processing of endmember extraction, unsupervised target detection, anomaly detection, as well as their field programmable gate array (FPGA) implementation.

Chein-I Chang ()

Professor of Electrical Engineering

Remote Sensing Signal and Image Processing Laboratory (RSSIPL)

University of Maryland, Baltimore County

Baltimore, Maryland

USA

Chair Professor of Remote Sensing Technology

National Chung Hsing University

Taichung, Taiwan

Republic of China

Distinguished Professor

Providence University

Taichung, Taiwan

Republic of China

International Chair Professor

National Taipei University of Technology

Taipei, Taiwan

Republic of China

Technical Advisor

Center for QUantitative Imaging in Medicine (CQUIM)

Taichung Veterans General Hospital

Taichung, Taiwan

Republic of China

Fall 2012

1

Overview and Introduction

The past few years have witnessed tremendous advances in hyperspectral imaging where statistical signal processing has played a pivotal role in driving algorithm design and development for hyperspectral data exploitation. It has attracted attention of those who come from different disciplinary areas by exploring new applications and making connections between remote sensing and other engineering fields. In recent years, there has been a significant increase in participation in various conferences and venues related to this area, which in turn has provided evidence that hyperspectral signal and image processing has broken away from traditional spatial domain analysis–based remote sensing and successfully branched out to stand alone as a single research topic, similar to signal processing that evolved as a separate area from communications in the late 1970s. On the contrary issues related to high spectral resolution provided by hyperspectral imaging sensors have also changed the ways in which algorithms are designed and developed. As a consequence, many problems such as subpixels and mixed pixels that are generally encountered in hyperspectral image processing, but do not occur in classical two-dimensional (2D) image processing, have become major issues for traditional spatial domain-based techniques. This is because the concept of “seeing-is-believing” by visual inspection, which has been widely used in image processing, cannot resolve issues of targets that are completely embedded in a single pixel or partially but do not fully occupy a single pixel, in which case only spectral properties can be used to characterize such targets for data analysis. Therefore, to distinguish such spectral characterization-based analysis from the traditional spatial domain–based analysis, the former is referred to as nonliteral analysis as opposed to the latter termed as literal analysis.

Due to complicated environments in real-world problems many uncontrollable parameters are also beyond our grip. In order to explore insights into algorithm design, the use of synthetic images to simulate various scenarios to substantiate designed algorithms for performance analysis becomes an effective proof-of-concept evaluation tool. Such synthetic images can be simulated by either real image scenes or laboratory data sets for various applications. Unfortunately, such synthetic image-based computer simulations have received little attention in the past. Accordingly, one of the major features that readers will find in this book is an extensive use of synthetic image-based experiments in algorithm design and analysis for qualitative as well as quantitative performance evaluation. Another unique feature of this book is that the algorithms derived and developed in this book can be implemented with little difficulty via the MATLAB, a widely accepted software package developed by the MATHWORK for engineering applications. This advantage allows readers to implement their algorithms. To further facilitate this benefit MATLAB codes of many popular algorithms developed in this book are also made available in the appendix at the end of this book. Most importantly, this book has also expanded its use of images in Chang (2003a) to include two more popular image scenes, Purdue's Indian Pine test site in Indiana and Cuprite image scene in Nevada, both of which are available on web site so that they can be used by those who develop new algorithms, to validate and evaluate their designed algorithm for performance analysis as well as to conduct their own comparative study. Last but not the least, this book includes an appendix that compiles many algorithms developed in the Remote Sensing Signal and Image Processing Laboratory (RSSIPL) at the University of Maryland, Baltimore County (UMBC). Such an algorithm compendium should serve as a valuable guide for people who are interested in applications of hyperspectral data processing.