Algorithms for Image Processing and Computer Vision

J. R. Parker is a full professor working in the Art department at the University of Calgary. His major research projects include live performance in online virtual spaces, the design and construction of kinetic games, and the portrayal of Canadian history and culture in digital and online form.

• Preface xxiChapter 1 Practical Aspects of a Vision System — Image Display, Input/Output, and Library Calls 1OpenCV 2The Basic OpenCV Code 2The IplImage Data Structure 3Reading and Writing Images 6Image Display 7An Example 7Image Capture 10Interfacing with the AIPCV Library 14Website Files 18References 18Chapter 2 Edge-Detection Techniques 21The Purpose of Edge Detection 21Traditional Approaches and Theory 23Models of Edges 24Noise 26Derivative Operators 30Template-Based Edge Detection 36Edge Models: The Marr-Hildreth Edge Detector 39The Canny Edge Detector 42The Shen-Castan (ISEF) Edge Detector 48A Comparison of Two Optimal Edge Detectors 51Color Edges 53Source Code for the Marr-Hildreth Edge Detector 58Source Code for the Canny Edge Detector 62Source Code for the Shen-Castan Edge Detector 70Website Files 80References 82Chapter 3 Digital Morphology 85Morphology Defined 85Connectedness 86Elements of Digital Morphology — Binary Operations 87Binary Dilation 88Implementing Binary Dilation 92Binary Erosion 94Implementation of Binary Erosion 100Opening and Closing 101MAX — A High-Level Programming Language for Morphology 107The ‘‘Hit-and-Miss’’ Transform 113Identifying Region Boundaries 116Conditional Dilation 116Counting Regions 119Grey-Level Morphology 121Opening and Closing 123Smoothing 126Gradient 128Segmentation of Textures 129Size Distribution of Objects 130Color Morphology 131Website Files 132References 135Chapter 4 Grey-Level Segmentation 137Basics of Grey-Level Segmentation 137Using Edge Pixels 139Iterative Selection 140The Method of Grey-Level Histograms 141Using Entropy 142Fuzzy Sets 146Minimum Error Thresholding 148Sample Results From Single Threshold Selection 149The Use of Regional Thresholds 151Chow and Kaneko 152Modeling Illumination Using Edges 156Implementation and Results 159Comparisons 160Relaxation Methods 161Moving Averages 167Cluster-Based Thresholds 170Multiple Thresholds 171Website Files 172References 173Chapter 5 Texture and Color 177Texture and Segmentation 177A Simple Analysis of Texture in Grey-Level Images 179Grey-Level Co-Occurrence 182Maximum Probability 185Moments 185Contrast 185Homogeneity 185Entropy 186Results from the GLCM Descriptors 186Speeding Up the Texture Operators 186Edges and Texture 188Energy and Texture 191Surfaces and Texture 193Vector Dispersion 193Surface Curvature 195Fractal Dimension 198Color Segmentation 201Color Textures 205Website Files 205References 206Chapter 6 Thinning 209What Is a Skeleton? 209The Medial Axis Transform 210Iterative Morphological Methods 212The Use of Contours 221Choi/Lam/Siu Algorithm 224Treating the Object as a Polygon 226Triangulation Methods 227Force-Based Thinning 228Definitions 229Use of a Force Field 230Subpixel Skeletons 234Source Code for Zhang-Suen/Stentiford/Holt Combined Algorithm 235Website Files 246References 247Chapter 7 Image Restoration 251Image Degradations — The Real World 251The Frequency Domain 253The Fourier Transform 254The Fast Fourier Transform 256The Inverse Fourier Transform 260Two-Dimensional Fourier Transforms 260Fourier Transforms in OpenCV 262Creating Artificial Blur 264The Inverse Filter 270The Wiener Filter 271Structured Noise 273Motion Blur — A Special Case 276The Homomorphic Filter — Illumination 277Frequency Filters in General 278Isolating Illumination Effects 280Website Files 281References 283Chapter 8 Classification 285Objects, Patterns, and Statistics 285Features and Regions 288Training and Testing 292Variation: In-Class and Out-Class 295Minimum Distance Classifiers 299Distance Metrics 300Distances Between Features 302Cross Validation 304Support Vector Machines 306Multiple Classifiers — Ensembles 309Merging Multiple Methods 309Merging Type 1 Responses 310Evaluation 311Converting Between Response Types 312Merging Type 2 Responses 313Merging Type 3 Responses 315Bagging and Boosting 315Bagging 315Boosting 316Website Files 317References 318Chapter 9 Symbol Recognition 321The Problem 321OCR on Simple Perfect Images 322OCR on Scanned Images — Segmentation 326Noise 327Isolating Individual Glyphs 329Matching Templates 333Statistical Recognition 337OCR on Fax Images — Printed Characters 339Orientation — Skew Detection 340The Use of Edges 345Handprinted Characters 348Properties of the Character Outline 349Convex Deficiencies 353Vector Templates 357Neural Nets 363A Simple Neural Net 364A Backpropagation Net for Digit Recognition 368The Use of Multiple Classifiers 372Merging Multiple Methods 372Results From the Multiple Classifier 375Printed Music Recognition — A Study 375Staff Lines 376Segmentation 378Music Symbol Recognition 381Source Code for Neural Net Recognition System 383Website Files 390References 392Chapter 10 Content-Based Search — Finding Images by Example 395Searching Images 395Maintaining Collections of Images 396Features for Query by Example 399Color Image Features 399Mean Color 400Color Quad Tree 400Hue and Intensity Histograms 401Comparing Histograms 402Requantization 403Results from Simple Color Features 404Other Color-Based Methods 407Grey-Level Image Features 408Grey Histograms 409Grey Sigma — Moments 409Edge Density — Boundaries Between Objects 409Edge Direction 410Boolean Edge Density 410Spatial Considerations 411Overall Regions 411Rectangular Regions 412Angular Regions 412Circular Regions 414Hybrid Regions 414Test of Spatial Sampling 414Additional Considerations 417Texture 418Objects, Contours, Boundaries 418Data Sets 418Website Files 419References 420Systems 424Chapter 11 High-Performance Computing for Vision and Image Processing 425Paradigms for Multiple-Processor Computation 426Shared Memory 426Message Passing 427Execution Timing 427Using clock() 428Using QueryPerformanceCounter 430The Message-Passing Interface System 432Installing MPI 432Using MPI 433Inter-Process Communication 434Running MPI Programs 436Real Image Computations 437Using a Computer Network — Cluster Computing 440A Shared Memory System — Using the PC Graphics Processor 444GLSL 444OpenGL Fundamentals 445Practical Textures in OpenGL 448Shader Programming Basics 451Vertex and Fragment Shaders 452Required GLSL Initializations 453Reading and Converting the Image 454Passing Parameters to Shader Programs 456Putting It All Together 457Speedup Using the GPU 459Developing and Testing Shader Code 459Finding the Needed Software 460Website Files 461References 461Index 465