Kropatsch / Bischof Digital Image Analysis

Selected Techniques and Applications
1. Auflage 2006
ISBN: 978-0-387-21643-0
Verlag: Springer US
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, 513 Seiten

This book presents a broad-ranging edited survey of computational and analytical methods and tools for digital image analysis and interpretation. The book brings together the recent results and methods in a uniform manner, thereby making the information accessible to nonspecialists and specialists alike.

Topics and features:
* Diverse topics are treated in an integrative style, using a common notation
* With theory and applications covered in a single volume, the reader sees immediately that the proposed methods also work in practice
* Overview of some key research in digital image processing and pattern recognition methods and tools
* Up-to-date coverage of current topics: information fusion, stochastic shape theory, graph-based image analysis and hierarchical systems

The book offers a uniquely comprehensive technical survey that not only provides in-depth coverage of the fundamental topics in the field, but also incorporates the newest developments that have arisen. It serves as an excellent and current resource for researchers, practitioners and professionals in computer science and electrical engineering focusing on methodology for digital imaging and analysis.

Written for: Researchers, practitioners, professionals

Autoren/Hrsg.

Kropatsch, Walter

Bischof, Horst

Inhaltsverzeichnis

1;Preface;6
1.1;About This Book;6
1.2;The Compact Disc;7
1.3;Acknowledgments;8
2;Contributors;10
3;Contents;16
4;List of Figures;23
5;List of Tables;30
6;Part I Mathematical Methods for Image Analysis;31
6.1;Introduction to Part I;32
6.2;1 Numerical Harmonic Analysis and Image Processing;35
6.2.1;1.1 Gabor Analysis and Digital Signal Processing;35
6.2.1.1;1.1.1 From Fourier to Gabor Expansions;36
6.2.1.2;1.1.2 Local Time-Frequency Analysis and Short-Time Fourier Transform;43
6.2.1.3;1.1.3 Fundamental Properties of Gabor Frames;45
6.2.1.4;1.1.4 Commutation Relations of the Gabor Frame Operator;46
6.2.1.5;1.1.5 Critical Sampling, Oversampling, and the Balian-Low Theorem;46
6.2.1.6;1.1.6 Wexler-Raz Duality Condition;51
6.2.1.7;1.1.7 Gabor Analysis on LCA Groups;52
6.2.1.8;1.1.8 Numerical Gabor Analysis;58
6.2.1.9;1.1.9 Image Representation and Gabor Analysis;62
6.2.2;1.2 Signal and Image Reconstruction;62
6.2.2.1;1.2.1 Notation;63
6.2.2.2;1.2.2 Signal Reconstruction and Frames;64
6.2.2.3;1.2.3 Numerical Methods for Signal Reconstruction;65
6.2.3;1.3 Examples and Applications;68
6.2.3.1;1.3.1 Object Boundary Recovery in Echocardiography;71
6.2.3.2;1.3.2 Image Reconstruction in Exploration Geophysics;72
6.2.3.3;1.3.3 Reconstruction of Missing Pixels in Images;74
6.3;2 Stochastic Shape Theory;76
6.3.1;2.1 Shape Analysis;76
6.3.2;2.2 Contour Line Parameterization;78
6.3.3;2.3 Deformable Templates;79
6.3.3.1;2.3.1 Stochastic Planar Deformation Processes;80
6.3.3.2;2.3.2 Gaussian Isotropic Random Planar Deformations;81
6.3.3.3;2.3.3 The Deformable Templates Model;82
6.3.3.4;2.3.4 Maximum Likelihood Classi.cation;83
6.3.4;2.4 The Wavelet Transform;85
6.3.4.1;2.4.1 Atomic Decompositions and Group Theory;86
6.3.4.2;2.4.2 Discrete Wavelets and Multiscale Analysis;89
6.3.4.3;2.4.3 Wavelet Packets;94
6.3.5;2.5 Wavelet Packet Descriptors;99
6.3.6;2.6 Global Nonlinear Optimization;101
6.3.6.1;2.6.1 Multilevel Single-Linkage Global Optimization;102
6.3.6.2;2.6.2 Implementation;104
6.4;3 Image Compression and Coding;107
6.4.1;3.1 Image Compression;107
6.4.1.1;3.1.1 Lossy Compression and Machine Vision;108
6.4.1.2;3.1.2 Multilevel Polynomial Interpolation;116
6.4.1.3;3.1.3 Enhancing the FBI Fingerprint Compression Standard;121
6.4.2;3.2 Multimedia Data Encryption;128
6.4.2.1;3.2.1 Symmetric Product Ciphers;128
6.4.2.2;3.2.2 Permutation by Chaotic Kolmogorov Flows;129
6.4.2.3;3.2.3 Substitution by AWC or SWB Generators;134
6.4.2.4;3.2.4 Security Considerations;137
6.4.2.5;3.2.5 Encryption Experiments;137
6.4.2.6;3.2.6 Encryption Summary;140
6.5;References;141
7;Part II Data Handling;157
7.1;Introduction to Part II;158
7.2;4 Parallel and Distributed Processing;159
7.2.1;4.1 Dealing with Large Remote Sensing Image Data Sets;159
7.2.1.1;4.1.1 Demands of Earth Observation;159
7.2.1.2;4.1.2 Processing Radar-Data of the Magellan Venus Probe;161
7.2.2;4.2 Parallel Radar Signal Processing;162
7.2.2.1;4.2.1 Parallelization Strategy;162
7.2.2.2;4.2.2 Evaluation of Parallelization Tools;163
7.2.2.3;4.2.3 Program Analysis and Parallelization;165
7.2.3;4.3 Parallel Radar Image Processing;167
7.2.3.1;4.3.1 Data Decomposition and Halo Handling;168
7.2.3.2;4.3.2 Dynamic Load Balancing and Communication Overloading;169
7.2.3.3;4.3.3 Performance Assessment;170
7.2.4;4.4 Distributed Processing;173
7.2.4.1;4.4.1 Front End;174
7.2.4.2;4.4.2 Back End;174
7.2.4.3;4.4.3 Broker;175
7.2.4.4;4.4.4 Experiences;177
7.3;5 Image Data Catalogs;178
7.3.1;5.1 Online Access to Remote Sensing Imagery;179
7.3.1.1;5.1.1 Remote Sensing Data Management;179
7.3.1.2;5.1.2 Image Data Information and Request System;181
7.3.1.3;5.1.3 Online Product Generation and Delivery;182
7.3.2;5.2 Content-Based Image Database Indexing and Retrieval;184
7.3.2.1;5.2.1 The Miniature Portrait Database;186
7.3.2.2;5.2.2 The Eigen Approach;189
7.3.2.3;5.2.3 Experiments;191
7.4;References;193
8;Part II Robust and Adaptive Image Understanding;197
8.1;Introduction to Part III;198
8.2;6 Graphs in Image Analysis;200
8.2.1;6.1 From Pixels to Graphs;200
8.2.1.1;6.1.1 Graphs in the Square Grid;201
8.2.1.2;6.1.2 Run Graphs;201
8.2.1.3;6.1.3 Area Voronoi Diagram;205
8.2.2;6.2 Graph Transformations in Image Analysis;212
8.2.2.1;6.2.1 Arrangements of Image Elements;212
8.2.2.2;6.2.2 Dual Graph Contraction;214
8.3;7 Hierarchies;219
8.3.1;7.1 Regular Image Pyramids;219
8.3.1.1;7.1.1 Structure;221
8.3.1.2;7.1.2 Contents;223
8.3.1.3;7.1.3 Processing;223
8.3.1.4;7.1.4 Fuzzy Curve Pyramid;225
8.3.2;7.2 Irregular Graph Pyramids;228
8.3.2.1;7.2.1 Computational Complexity;229
8.3.2.2;7.2.2 Irregular Pyramids by Hop.eld Networks;230
8.3.2.3;7.2.3 Equivalent Contraction Kernels;233
8.3.2.4;7.2.4 Extensions to 3D;236
8.4;8 Robust Methods;239
8.4.1;8.1 The Role of Robustness in Computer Vision;239
8.4.2;8.2 Parametric Models;240
8.4.2.1;8.2.1 Robust Estimation Methods;240
8.4.3;8.3 Robust Methods in Vision;241
8.4.3.1;8.3.1 Recover-and-Select Paradigm;241
8.4.3.2;8.3.2 Recover-and-Select applied to;247
8.5;9 Structural Object Recognition;256
8.5.1;9.1 2-D and 3-D Structural Features;256
8.5.2;9.2 Feature Selection;257
8.5.3;9.3 Matching Structural Descriptions;257
8.5.4;9.4 Reducing Search Complexity;258
8.5.5;9.5 Grouping and Indexing;258
8.5.5.1;9.5.1 Early Search Termination;259
8.5.6;9.6 Detection of Polymorphic Features;260
8.5.7;9.7 Polymorphic Grouping;260
8.5.8;9.8 Indexing and Matching;261
8.5.9;9.9 Polymorphic Features;261
8.5.10;9.10 3-D Object Recognition Example;262
8.5.10.1;9.10.1 The IDEAL System;262
8.5.10.2;9.10.2 Initial Structural Part Decomposition;263
8.5.10.3;9.10.3 Part Adjacency and Compatibility Graphs;264
8.5.10.4;9.10.4 Automatic Model Acquisition;266
8.5.10.5;9.10.5 Object Recognition from Appearances;267
8.5.10.6;9.10.6 Experiments;268
8.6;10 Machine Learning;269
8.6.1;10.1 What Is Machine Learning?;269
8.6.1.1;10.1.1 What Do Machine Learning Algorithms Need?;270
8.6.1.2;10.1.2 One Method Solves All? Use of Multistrategy;270
8.6.2;10.2 Methods;271
8.6.3;10.3 Operational;272
8.6.3.1;10.3.1 Discrimination and Classi.cation;274
8.6.3.2;10.3.2 Optimization and Search;274
8.6.3.3;10.3.3 Functional Relationship;275
8.6.3.4;10.3.4 Logical Operations;275
8.6.4;10.4 Object-Oriented Generalization;275
8.6.5;10.5 Generalized Logical Structures;276
8.6.5.1;10.5.1 Reformulation;277
8.6.5.2;10.5.2 Object-Oriented Implementation;278
8.6.6;10.6 Generalized Clustering Algorithms;279
8.6.6.1;10.6.1 Function Overloading;280
8.6.7;Conclusion;281
8.7;References;282
9;Part IV Information Fusion and Radiometric Models for Image Understanding;298
9.1;Introduction to Part IV;299
9.2;11 Information Fusion in Image Understanding;300
9.2.1;11.1 Active Fusion;301
9.2.2;11.2 Active Object Recognition;302
9.2.2.1;11.2.1 Related Research;304
9.2.3;11.3 Feature Space Active Recognition;305
9.2.3.1;11.3.1 Object Recognition in Parametric Eigenspace;306
9.2.3.2;11.3.2 Probability Distributions in Eigenspace;307
9.2.3.3;11.3.3 View Classi.cation and Pose Estimation;308
9.2.3.4;11.3.4 Information Integration;309
9.2.3.5;11.3.5 View Planning;310
9.2.3.6;11.3.6 The Complexity of the Algorithm;311
9.2.3.7;11.3.7 Experiments;312
9.2.3.8;11.3.8 A Counterexample for Conditional Independence in Equation ( 11.5);318
9.2.3.9;11.3.9 Conclusion;319
9.2.4;11.4 Reinforcement Learning for Active Object Recognition;320
9.2.4.1;11.4.1 Adaptive Generation of Object Hypotheses;322
9.2.4.2;11.4.2 Learning Recognition Control;325
9.2.4.3;11.4.3 Experiments;327
9.2.4.4;11.4.4 Discussion and Outlook;332
9.2.5;11.5 Generic Active Object Recognition;332
9.2.5.1;11.5.1 Object Models;333
9.2.5.2;11.5.2 Recognition System;334
9.2.5.3;11.5.3 Hypothesis Generation;334
9.2.5.4;11.5.4 Visibility Space;338
9.2.5.5;11.5.5 Viewpoint Estimation;341
9.2.5.6;11.5.6 Viewpoints and Actions;344
9.2.5.7;11.5.7 Motion Planning;346
9.2.5.8;11.5.8 Object Hypotheses Fusion;348
9.2.5.9;11.5.9 Conclusion;349
9.3;12 Image Understanding Methods for Remote Sensing;351
9.3.1;12.1 Radiometric Models;353
9.3.2;12.2 Subpixel Analysis of Remotely Sensed Images;360
9.3.3;12.3 Segmentation of Remotely Sensed Images;364
9.3.4;12.4 Land-Cover Classi.cation;367
9.3.5;12.5 Information Fusion for Remote Sensing;369
9.4;References;372
10;Part V 3D Reconstruction;380
10.1;Introduction to Part V;381
10.2;13 Fundamentals;384
10.2.1;13.1 Image Acquisition Aspects;384
10.2.1.1;13.1.1 Video Cameras;385
10.2.1.2;13.1.2 Amateur Cameras with CCD Sensors;385
10.2.1.3;13.1.3 Analog Metric Cameras;385
10.2.1.4;13.1.4 Remote Sensing Scanners;386
10.2.1.5;13.1.5 Other Visual Sensor Systems;387
10.2.2;13.2 Perspective Transformation;387
10.2.3;13.3 Stereo Reconstruction;391
10.2.4;13.4 Bundle Block Con.gurations;393
10.2.5;13.5 From Points and Lines to Surfaces;394
10.2.5.1;13.5.1 Representation of Irregular Object Surfaces;396
10.2.5.2;13.5.2 Representation of Man-Made Objects;399
10.2.5.3;13.5.3 Hybrid Representation of Object Surfaces;401
10.3;14 Image Matching Strategies;403
10.3.1;14.1 Raster-Based Matching Techniques;405
10.3.1.1;14.1.1 Cross Correlation;405
10.3.1.2;14.1.2 Least Squares Matching;407
10.3.2;14.2 Feature-Based Matching Techniques;409
10.3.2.1;14.2.1 Feature Extraction;409
10.3.2.2;14.2.2 Matching Homologous Image Features;412
10.3.3;14.3 Hierarchical Feature Vector Matching (HFVM);416
10.3.3.1;14.3.1 Feature Vector Matching (FVM);416
10.3.3.2;14.3.2 Subpixel Matching;419
10.3.3.3;14.3.3 Consistency Check;419
10.3.3.4;14.3.4 Hierarchical Feature Vector Matching;419
10.4;15 Precise Photogrammetric Measurement: Location of Targets and Reconstruction of Object Surfaces;421
10.4.1;15.1 Automation in Photogrammetric Plotting;423
10.4.1.1;15.1.1 Automation of Inner Orientation;424
10.4.1.2;15.1.2 Automation of Outer Orientation;424
10.4.2;15.2 Location of Targets;425
10.4.2.1;15.2.1 Location of Circular Symmetric Targets by Intersection of Gradient Vectors;426
10.4.2.2;15.2.2 Location of Arbitrarily Shaped Targets;427
10.4.2.3;15.2.3 The OEEPE Test on Digital Aerial Triangulation;429
10.4.2.4;15.2.4 Deformation Analysis of Wooden Doors;430
10.4.3;15.3 A General Framework for Object Reconstruction;432
10.4.3.1;15.3.1 Hierarchical Object Reconstruction;433
10.4.3.2;15.3.2 Mathematical Formulation of the Object Models;437
10.4.3.3;15.3.3 Robust Hybrid Adjustment;439
10.4.3.4;15.3.4 DEM Generation for Topographic Mapping;440
10.4.4;15.4 Semiautomatic Building Extraction;441
10.4.4.1;15.4.1 Building Models;443
10.4.4.2;15.4.2 Interactive Determination of Approximations;444
10.4.4.3;15.4.3 Automatic Fine Reconstruction;446
10.4.5;15.5 State of Work;447
10.5;16 3D Navigation and Reconstruction;448
10.5.1;16.1 High Accurate Stereo Reconstruction of Naturally Textured Surfaces for Navigation and 3D- Modeling;448
10.5.1.1;16.1.1 Reconstruction of Arbitrary Shapes Using the Locus Method;448
10.5.1.2;16.1.2 Using the locus Method for Cavity Inspection;452
10.5.1.3;16.1.3 Stereo Reconstruction Using Remote Sensing Images;456
10.5.1.4;16.1.4 Stereo Reconstruction for Space Research;459
10.5.1.5;16.1.5 Operational Industrial Stereo Vision Systems;459
10.5.2;16.2 A Framework for Vision-Based Navigation;461
10.5.2.1;16.2.1 Vision Sensor Systems;462
10.5.2.2;16.2.2 Closed-Loop Solution for Autonomous Navigation;463
10.5.2.3;16.2.3 Risk Map Generation;464
10.5.2.4;16.2.4 Local Path Planning;464
10.5.2.5;16.2.5 Path Execution and Navigation on the DEM;465
10.5.2.6;16.2.6 Prototype Software for Closed-Loop Vehicle Navigation;467
10.5.2.7;16.2.7 Simulation Results;468
10.6;17 3D Object Sensing Using Rotating CCD Cameras;474
10.6.1;17.1 Concept of Image-Based Theodolite Measurement Systems;474
10.6.2;17.2 The Videometric Imaging System;476
10.6.2.1;17.2.1 The Purpose of the Videometric Imaging System;476
10.6.2.2;17.2.2 An Interactive Measurement System–A First Step;479
10.6.2.3;17.2.3 An Automatic System–A Second Step;482
10.6.3;17.3 Conversion of the Measurement System into a Robot System;490
10.6.4;17.4 Decision Making;491
10.6.5;17.5 Outlook;495
10.7;References;497
11;Index;507

Leseproben

2 Stochastic Shape Theory (P. 49)

Christian Cenker

Georg Pflug

Manfred Mayer

Stochastic models and statistical procedures are essential for pattern recognition. Linear discriminant analysis, parametric and nonparametric density estimation, maximumlikelihood classiffication, supervised and nonsupervised learning, neural nets, parametric, nonparametric, and fuzzy clustering, principal component analysis, simulated annealing are only some of the well-known statistical techniques used for pattern recognition. Markov models and other stochastic models are often used to describe statistical characteristics of patterns in the pattern space.

We want to concentrate on modeling and feature extraction using new techniques.We do not model the characteristics of the pattern space but the generation of the patterns, i.e., modeling the pattern generation process via stochastic processes. Furthermore, wavelets and wavelet packets will help us to construct a feature extractor. Applying our models to a sample application we noticed the lack of global non-linear optimization algorithms. Thus, we added a section on optimization, in which we present a modiffi- cation of a multi-level single-linkage technique that can be used in high-dimensional feature spaces.

2.1 Shape Analysis

A project on o.ine signature verification shows the need for new approaches. Standard methods do not show the wanted accuracy, nevertheless, they have been implemented at a first stage in order to compare the results. As all signatures of one person are of di.erent but similar shape we look for a description of the similarity and the difference. First, a signature is a special form of curve, we discard all color, thickness and "pressure" information from the scanned signature (cf. (AYF86)), leaving only a thinned polygonal shape. We have a connected skeleton of the "contour".

The first problem to solve is the parameterization of the curve, i.e., to get a onedimensional function that represents the two-dimensional signature, as our constraints are on the one hand to use as little data for storage of the signatures as possible and, on the other hand, to develop fast algorithms. Thus, using only one-dimensional objects (functions) seem to be a feasible solution. We choose a change-in-angle parameterization of the curve, which has the advantages of shift, rotation and scale invariance (cf. (Nie90)).

Features are then extracted forming a sampled version of the contour, stored in a k-dimensional vector, and used for discrimination and classiffication. Based on the change-in-angle parameterization we present three different approaches to match the patterns. Starting with the description of classes of signatures and their similarity by stochastic processes, i.e., stochastic deformation processes, describe the generation process of the signatures of an individual (see Section 2.3).

Secondly, we want to use new "standard" signal analysis methods to analyze the curve or polygonal shape, i.e., wavelet and frame methods, as they provide fast algorithms that produce patterns that have a nice easy interpretation (see Section 2.5).