SHAPE ANALYSIS AND CLASSIFICATION: THEORY AND PRACTICE
by
Luciano
da Fontoura Costa
and
Roberto M.
Cesar Junior
CRC Press Book Series
on Image Processing - CRC
Press
Table of Contents
1. INTRODUCTION
-
1.1 INTRODUCTION TO SHAPE ANALYSIS
-
1.2 CASE STUDIES
-
1.2.1 Case Study: Morphology of Plant Leaves
-
1.2.2 Case Study: Morphometric Classification of Ganglion Cells
-
1.3 COMPUTATIONAL SHAPE ANALYSIS
-
1.3.1 SHAPE PRE-PROCESSING
-
1.3.2 SHAPE TRANSFORMATIONS
-
1.3.3 SHAPE CLASSIFICATION
-
1.4 ORGANIZATION OF THE BOOK
2. MATHEMATICAL CONCEPTS
-
2.1 BASIC CONCEPTS
-
2.1.1 Propositional Logic
-
2.1.2 Functions
-
2.1.3 Free Variable Transformations
-
2.1.4 Some Special Real Functions
-
2.1.5 Complex Functions
-
2.2 LINEAR ALGEBRA
-
2.2.1 Scalars, Vectors, and Matrices
-
2.2.2 Vector Spaces
-
2.2.3 Linear Transformations
-
2.2.4 Metric Spaces, Inner Products, and Orthogonality
-
2.2.5 More about Vectors and Matrices
-
2.3 DIFFERENTIAL GEOMETRY
-
2.3.1 2D Parametric Curves
-
2.3.2 Arc Length, Speed and Tangent Fields
-
2.3.3 Normal Fields and Curvature
-
2.4 MULTIVARIATE CALCULUS
-
2.4.1 Multivariate Functions
-
2.4.2 Directional, Partial, and Total Derivatives
-
2.4.3 Differential Operators
-
2.5 CONVOLUTION AND CORRELATION
-
2.5.1 Continuous Convolution and Correlation
-
2.5.2 Discrete Convolution and Correlation
-
2.5.3 Non-Linear Correlation as a Coincidence Operator
-
2.6 PROBABILITY AND STATISTICS
-
2.6.1 Events and Probability
-
2.6.2 Random Variables and Probability Distributions
-
2.6.3 Random Vectors and Joint Distributions
-
2.6.4 Estimation
-
2.6.5 Autocorrelation 1242.6.6 The Karhunen-Loève Transform
-
2.7 FOURIER ANALYSIS
-
2.7.1 Brief Historical Remarks
-
2.7.2 The Fourier Series
-
2.7.3 Cookie Recipe
-
2.7.4 Fourier Series
-
2.7.5 The Continuous One-Dimensional Fourier Transform
-
2.7.6 Frequency Filtering
-
2.7.7 The Discrete One-Dimensional Fourier Transform
-
2.7.8 Matrix Formulation of the DFT
-
2.7.9 Applying the DFT
-
2.7.10 The Fast Fourier Transform
-
2.7.11 Discrete Convolution Performed in the Frequency Domain
3. IMAGE ACQUISITION AND PROCESSING
-
3.1 IMAGE REPRESENTATION
-
3.1.1 Image Formation and Gray Level Images
-
3.1.2 Sampling and Quantization
-
3.1.3 Shape Sampling
-
3.1.4 Binary Images
-
3.1.5 Color Digital Images
-
3.1.6 Video Sequences
-
3.1.7 Multispectral Images
-
3.1.8 Voxels
-
3.2 IMAGE PROCESSING AND FILTERING
-
3.2.1 Histograms and Pixel Manipulation
-
3.2.2 Local or neighborhood processing
-
3.2.3 Average Filtering
-
3.2.4 Linear Filtering
-
3.2.5 Linear Filtering
-
3.2.6 Gaussian Smoothing
-
3.2.7 Fourier-based filtering
-
3.2.8 Median and other Non-Linear Filters
-
3.3 IMAGE SEGMENTATION: EDGE DETECTION
-
3.3.1 Edge Detection in Binary Images
-
3.3.2 Gray-Level Edge Detection
-
3.3.3 Gradient-Based Edge Detection
-
3.3.4 Roberts Operator
-
3.3.5 Sobel, Prewitt and Kirsch Operator
-
3.3.6 Second Order Operators: Laplacian
-
3.3.7 Fourier-Based Edge Detection
-
3.3.8 Multiscale Edge Detection: The Marr-Hildreth Transform
-
3.4 IMAGE SEGMENTATION: FURTHER ALGORITHMS
-
3.4.1 Image Thresholding
-
3.4.2 Region-Growing
-
3.5 BINARY MATHEMATICAL MORPHOLOGY
-
3.5.1 Image Dilation
-
3.5.2 Image Erosion
-
3.6 CONCLUDING REMARKS
4. SHAPE CONCEPTS
-
4.1 INTRODUCTION TO TWO-DIMENSIONAL SHAPES
-
4.2 CONTINUOUS TWO-DIMENSIONAL SHAPES
-
4.2.1 Continuous Shapes and their Types
-
4.3 PLANAR SHAPE TRANSFORMATIONS
-
4.4 CHARACTERIZING 2D SHAPES IN TERMS OF FEATURES
-
4.5 CLASSIFYING 2D SHAPES
-
4.6 REPRESENTING 2D SHAPES
-
4.6.1 General Shape Representations
-
4.6.2 Landmark Representations
-
4.8 SHAPE METRICS
-
4.8.1 The 2n Euclidean Norm
-
4.8.2 The Mean Size
-
4.8.3 Alternative Shape Sizes
-
4.8.4 Which Size?
-
4.8.5 Distances Between Shapes
-
4.9 MORPHIC TRANSFORMATIONS
-
4.9.1 Affine Transformations
-
4.9.2 Euclidean Motions
-
4.9.3 Rigid Body Transformations
-
4.9.4 Similarity Transformations
-
4.9.5 Other Transformations and Some Important Remarks
-
4.9.6 Thin-Plate Splines
-
4.9.6.1 Single Thin-Plate Splines
-
4.9.6.2 Pairs of Thin-Plate Splines
5. TWO-DIMENSIONAL SHAPE REPRESENTATION
-
5.1 INTRODUCTION
-
5.2 PARAMETRIC CONTOUR
-
5.2.1 Contour Extraction
-
5.2.2 A Contour Following Algorithm
-
5.2.3 Contour Representation by Vectors and Complex Signals
-
5.2.4 Contour Representation based on the Chain Code
-
5.3 SET OF CONTOUR POINTS
-
5.4 CURVE APPROXIMATION
-
5.4.1 Poligonal Approximation
-
5.4.2 Ramer Algorithm for Polygonal Approximation
-
5.4.3 Split-and-Merge Algorithm for Polygonal Approximation
-
5.5 DIGITAL STRAIGHT LINES
-
5.5.1 Straight Lines and Segments
-
5.5.2 Generating Digital Straight Lines and Segments
-
5.5.3 Recognizing an Isolated Digital Straight Segment
-
5.6 HOUGH TRANSFORMS
-
5.6.1 Continuous Hough Transforms
-
5.6.2 Discrete Image and Continuous Parameter Space
-
5.6.3 Discrete Image and Parameter Space
-
5.6.4 Backmapping
-
5.6.5 Problems with the Hough Transform
-
5.6.6 Improving the Hough Transform
-
5.6.7 General Remarks on the Hough Transform
-
5.7 EXACT DILATIONS
-
5.8 DISTANCE TRANSFORMS
-
5.9 EXACT DISTANCE TRANSFORM THROUGH EXACT DILATIONS
-
5.10 VORONOI DIAGRAMS
-
5.11 SCALE SPACE SKELETONIZATION
6. SHAPE CHARACTERIZATION
-
6.1 STATISTICS FOR SHAPE DESCRIPTORS
-
6.2 SOME GENERAL DESCRIPTORS
-
6.2.1 Perimeter
-
6.2.2 Area
-
6.2.3 Centroid (center of mass)
-
6.2.4 Maximum and Minimum Distance to Centroid
-
6.2.5 Distance to the Boundary
-
6.2.6 Diameter
-
6.2.7 Maximum Chord
-
6.2.8 Norm Sizes
-
6.2.9 Maximum Arc Length
-
6.2.10 Major and Minor Axis
-
6.2.11 Thickness
-
6.2.12 Holes-based Shape Features
-
6.2.13 Topological Descriptors
-
6.2.14 Polygonal Approximation-Based Shape Descriptors
-
6.2.15 Shape Descriptors based on Regions and on Graphs
-
6.2.16 Complexity Descriptors 203
-
6.3 FRACTAL GEOMETRY FOR COMPLEXITY DESCRIPTORS
-
6.3.1 Preliminary Considerations and Definitions
-
6.3.2 The Box-Counting Approach
-
6.3.3 Case Example: The Classical Koch Curve
-
6.3.4 Implementing the Box-Counting Method
-
6.3.5 The Minkowski Sausage or Dilation Method
-
6.4 CURVATURE
-
6.4.1 Biological Motivation
-
6.4.2 Simple Approaches to Curvature
-
6.4.3 Curvature-Based Shape Descriptors
-
6.4.4 c-Curvature
-
6.5 SHAPE SIGNATURES
-
6.6 FOURIER DESCRIPTORS
-
6.6.1 Alternative Fourier Descriptors
7. MULTISCALE SHAPE CHARACTERIZATION
-
7.1 MULTISCALE TRANSFORMS
-
7.1.1 The Scale-Space
-
7.1.2 Time-Frequency Transforms
-
7.1.3 Gabor Filters
-
7.1.4 Time-Scale Transforms or Wavelets
-
7.1.5 A Unified Approach to Linear Multiscale Transforms
-
7.1.6 Case Study: Interpreting the Transforms
-
7.1.7 Analyzing the Multiscale Transforms
-
7.2 A FOURIER APPROACH TO MULTISCALE CURVATURE
-
7.2.1 Curvature Estimation using a Fourier Property
-
7.2.2 Numerical Differentiation using the Fourier Property
-
7.2.3 Gaussian Filtering and the Multiscale Approach
-
7.2.4 Some Simple Solutions for the Shrinking Problem
-
7.2.5 The Curvegram
-
7.2.6 Some Experimental Results
-
7.2.7 Curvature-Scale Space
-
7.3 MULTISCALE CONTOUR ANALYSIS USING WAVELETS
-
7.3.1 Preliminary Considerations
-
7.3.2 The W-Representation
-
7.3.3 Choosing the Analyzing Wavelet
-
7.3.4 Shape Analysis From the W-Representation
-
7.3.5 Dominant Point Detection using the W-Representation
-
7.3.6 Local Frequencies and Natural Scales
-
7.3.7 Contour Analysis using the Gabor Transform
-
7.3.8 Comparing and Integrating the Multiscale Representations
-
7.4 MULTISCALE ENERGIES
-
7.4.1 The Multiscale Bending Energy
-
7.4.2 The Multiscale Bending Energy
-
7.4.3 Neuromorphometry with Bending Energy
-
7.4.4 The Multiscale Wavelet Energy
-
7.4.5 Case of Study: Classification of Ganglion Cells
-
7.4.6 The Feature Space
-
7.4.7 Feature Selection and Dimensionality Reduction
8. SHAPE RECOGNITION AND CLASSIFICATION
-
8.1 INTRODUCTION TO SHAPE CLASSIFICATION
-
8.1.1 The Importance of Classification
-
8.1.2 Some Basic Concepts in Classification
-
8.1.3 A Simple Case Study in Classification
-
8.1.4 Some Additional Concepts in Classification
-
8.1.5 Feature Extraction
-
8.1.6 Feature Normalization
-
8.2 SUPERVISED PATTERN CLASSIFICATION
-
8.2.1 Bayes Decision Theory Principles
-
8.2.2 Bayesian Classification Involving Multiple Classes and Dimensions
-
8.2.3 Bayesian Classification of Leaves
-
8.2.4 Nearest Neighbours
-
8.3 UNSUPERVISED CLASSIFICATION AND CLUSTERING
-
8.3.1 Basic Concepts and Issues
-
8.3.2 Scatter Matrices and Dispersion Measures
-
8.3.3 Partitional Clustering
-
8.3.4 Hierarchical Clustering
-
8.4 A CASE STUDY: LEAVES CLASSIFICATION
-
8.4.1 Choice of Method
-
8.4.2 Choice of Metric
-
8.4.3 Choice of Features
-
8.4.4 Effects of Unit Variance Normalization and Principal Component Analysis
-
8.4.5 Validation Considering the Cophenetic Correlation Coefficient
-
8.5 EVALUATING CLASSIFICATION METHODS
9. FUTURE TRENDS IN SHAPE ANALYSIS AND CLASSIFICATION
References