Relational Matching

Front Cover
Springer Science & Business Media, Sep 10, 1992 - Computers - 190 pages
Relational matching is a method for finding the best correspondences betweenstructural descriptions. It is widely used in computer vision for the recognition and location of objects in digital images. For this purpose, the digital images and the object models are represented by structural descriptions. The matching algorithm then has to determine which image elements and object model parts correspond. This book is the result of abasic study of relational matching. The book focuses particularly on the evaluation of correspondences. In order to find the best match, one needs a measure to evaluate the quality of a match. The author reviews the evaluation measures that have been suggested over the past few decades and presents a new measure based on information theory. The resulting theorycombines matching strategies, information theory, and tree search methods. For the benefit of the reader, comprehensive introductions are given to all these topics.
 

What people are saying - Write a review

We haven't found any reviews in the usual places.

Contents

Introduction to relational matching
1
12 Relational matching theory
2
13 Organization of the thesis
3
A classification of matching methods
5
212 Relations
6
213 The image as a function of coordinates
7
22 The match evaluation function
8
222 Constraints
11
642 Advantages of the mutual information
81
643 Mutual information of uncertain attributes
83
644 Mutual information as a compatibility function for relaxation processes
84
Tree search methods and heuristics
87
71 Problem representations in a tree
88
72 Tree search methods
91
722 Informed search methods
94
723 Informed search with a merit function
99

23 Search methods
12
231 Tree search
13
232 Simulated annealing
15
234 Least squares
17
235 Relaxation labeling
18
24 Hierarchy
20
25 Examples
22
252 3Dimensional object recognition with relational matching
23
253 Stereo matching with simulated annealing
25
255 Least squares area based matching
26
256 Least squares feature based matching
28
257 Scene interpretation by relaxation labeling
30
258 Computational models of human stereo vision
31
26 Discussion
32
Formal description of relational matching
35
32 Compositions
37
33 Exact matching
38
34 Inexact matching
41
35 Tree search
42
36 Some problems using relational matching
43
Problem definition and contributions of the thesis
45
42 Tree search methods
47
Theory of relational matching
49
51 Information measures for discrete signals
51
52 Information measures for continuous signals
54
53 The minimum description length principle
57
532 Interpretation of noisy point distributions
59
533 Relation to maximum a posteriori and maximum likelihood estimation
61
54 Discretization of continuous signals
62
Evaluation of mappings between relational descriptions
67
62 Mapping as an information channel
69
622 Modeling the transfer of information over a communication channel
71
63 The conditional information as a distance function after Boyer and Kak
73
633 Analysis of the conditional information as an evaluation function
74
64 The mutual information as a merit function
79
641 A probabilistic view
80
73 Checking consistency of future instantiations
101
731 Forward checking and looking ahead
102
732 Relational matching with forward checking
104
74 Unit ordering
105
75 The necessity of stop criteria for the correspondence problem
107
752 Perceptual grouping of primitives
109
Object location by relational matching
111
Relational image and model description
112
82 Extraction of image features
115
822 Line extraction
116
823 Region extraction
117
824 Symbolic postprocessing
118
83 Used primitives and relations and their attributes
121
Evaluation functions for object location
123
911 Mutual information of line length measurements
124
912 Mutual information of angle measurements
129
913 Mutual information determined from training matches
131
92 The mutual information of the spatial resection
134
93 Construction of the evaluation function for object location
135
941 Reliability of transformation parameters
136
942 A statistical test on the amount of support
138
Strategy and performance of the tree search for object location
143
101 Estimation of the future merit
144
102 Heuristics for object location
146
1022 The usefulness of a known transformation
148
1023 Underestimation of future merit
150
103 Description of the objects and their images
152
104 Performance of the object location
155
Summary and discussion
163
Literature
169
Mutual information between a continuous signal and a discretized noisy observation
181
Distribution of the coordinates of points on a sphere
185
Conditional probability density function of the image line length
187
Tables with search results
189
Copyright

Other editions - View all

Common terms and phrases