Representation and recognition in vision / Shimon Edelman.

Format:

Book

Author/Creator:

Edelman, Shimon, author.

Language:

English

Subjects (All):

Categorization (Psychology).

Cognitive science.

Mental representation.

Recognition (Psychology).

Visual perception.

Visualization.

Physical Description:

1 online resource (xxiii, 335 pages) : illustrations

Other Title:

MIT Press CogNet.

Place of Publication:

Cambridge, Massachusetts : The MIT Press, [1999]

System Details:

text file

Summary:

Researchers have long sought to understand what the brain does when we see an object, what two people have in common when they see the same object, and what a "seeing" machine would need to have in common with a human visual system. Recent neurobiological and computational advances in the study of vision have now brought us close to answering these and other questions about representation.

In Representation and Recognition in Vision, Shimon Edelman bases a comprehensive approach to visual representation on the notion of correspondence between proximal (internal) and distal similarities in objects. This leads to a computationally feasible and formally veridical representation of distal objects that addresses the needs of shape categorization and can be used to derive models of perceived similarity.

Contents:

1 The Problem of Representation 1

1.1 A Vision of Representation 1

1.2 Reconstruction 4

1.3 Representation without Reconstruction 6

1.3.1 The Feature Detector Redux 7

1.3.2 The Challenge 9

2 Theories of Representation and Object Recognition 11

2.1 Recognition-Related Tasks That Require Representation 11

2.1.1 Identification and Generalization 13

2.1.2 Categorization 15

2.1.3 Analogy 17

2.2 A Formalization of the Notion of Representation 19

2.2.1 The Problem of Representation 19

2.2.2 Representation as a Mapping 20

2.2.3 First- and Second-Order Isomorphism 21

2.3 Computational Theories of Recognition 27

2.3.1 Reconstructionist Theories: A Brief Historical Perspective 27

2.3.2 Structural Decomposition Theories 29

2.3.3 Theories Based on Geometric Constraints 33

2.3.4 Multidimensional Feature Spaces 35

3 S-isomorphism: The Theory 43

3.1 Similarity as Proximity in a Metric Space 45

3.1.1 Some Common Objections 45

3.1.2 A Metric Similarity Space as a Working Hypothesis 49

3.2 Shape Spaces 49

3.2.1 Kendall's Shape Space 49

3.2.2 Transformations and Deformations 50

3.2.3 Best-Correspondence Distance 52

3.2.4 An Objective Shape Space 53

3.3 Parameterization of Distal Shape Space 53

3.3.1 Scope of Parameterization 55

3.3.2 Dimensionality of Parameterization 58

3.4 The Distal to Proximal Mapping 62

3.4.1 Levels of Representation of Similarity 63

3.4.2 The Components of the Mapping F 64

3.4.3 Constraints on F 66

3.4.4 Implications 68

4 S-isomorphism: An Implementation 75

4.1 Task-Dependent Treatment of the Measurement Space 78

4.1.1 Identification ("Is this an image of object X?") 78

4.1.2 Recognition ("Is this an image of something I know?") 86

4.1.3 Categorization ("What is this thing?") 87

4.2 Categorization as Navigation in Shape Space 94

4.2.1 Defining the Frame of Reference 95

4.2.2 Active Landmarks 96

4.3 Tuned Units as Active Landmarks 98

4.3.1 Relevance 98

4.3.2 Coverage 98

4.3.3 Smoothness 98

4.3.4 Invariance to Irrelevant Dimensions 99

4.3.5 Monotonicity 101

4.3.6 Learnability 103

4.4 The Chorus of Prototypes 105

4.4.1 Persistent and Ephemeral Representations 106

6.4.2 Mental Rotation 165

6.4.3 Nonlinearities and the Effects of Practice 169

6.4.4 A Model of Canonical Views and Mental Rotation 173

6.4.5 Viewpoint Invariance as a Tool for Testing Theories of Representation 176

6.5 Psychophysics: Veridicality of Shape Representation 197

6.5.1 Perceived Similarity as Proximity in an Internal Metric Space 199

6.5.2 Veridical Perception of Distal Shape Contrasts 200

6.6.1 Translation Invariance for Line Drawings of Common Objects 208

6.6.2 Translation Invariance for Controlled Synthetic Animal-like Shapes 209

7 Dialogues on Representation and Recognition 219

7.1 On the Problem of Representation 220

7.2 On Theories of Representation 227

7.3 On S-isomorphism 229

7.4 On the Chorus of Prototypes 232

7.5 On the Performance of Chorus 234

7.6 On Representation and Recognition in Biological Vision 237

7.7 On What Has Been Left Out of This Book 243

Appendix A The Measurement Space and the Distal to Proximal Mapping 257

A.1 The View Space 259

A.2 The Shape Space 259

Appendix B Representation by Distances to Prototypes 263

B.1 The Chorus Transform 263

B.1.1 Distance Rank Preservation by CT 263

B.1.2 Bounds on Dimensionality Reduction: The Embedding Theorems 264

B.1.3 CT As an Implementation of Bourgain's Method 265

B.1.4 Non-expansion of Distances by CT 267

B.1.5 Preservation of Voronoi Structure by CT 268

B.2 Representation by Rank Order of Distances to Prototypes 269.

Notes:

"A Bradford book."

Includes bibliographical references (pages 303-329) and index.

Description based on print version record.

Other Format:

Print version: Edelman, Shimon. Representation and recognition in vision.

ISBN:

9780262272155

0262272156

Access Restriction:

Restricted for use by site license.