دانلود کتاب Shape Perception in Human and Computer Vision. An Interdisciplinary Perspective

فهرست مطالب :

Shape Perception in Human and Computer Vision
Preface
Contents
Contributors
Chapter 1: The Role of Mid-Level Shape Priors in Perceptual Grouping and Image Abstraction
1.1 Introduction
1.2 Symmetric Part Detection and Grouping
1.3 Contour Closure
1.4 Abstract Part Recovery
1.5 Conclusions
References
Chapter 2: Symmetry Is the sine qua non of Shape
2.1 Introduction
2.2 Prior Definitions of Shape
2.3 Explanation of the New Definition and How We Worked It out
2.4 Symmetry Groups for 3D Shapes, Their Invariants and Invariants of the Perspective Projection
2.5 Inferring 3D Shape from a 3D Object
2.6 Computational and Psychophysical Implications of the New Definition
2.6.1 Veridical Perception of 3D Shapes
2.6.2 Shapes of Non-rigid Objects
2.6.3 Symmetry as an Objective, but Informative, Prior
2.6.4 Shape Constancy: View-Invariant vs. View-Dependent Shape Perception
2.7 Conclusion
References
Chapter 3: Flux Graphs for 2D Shape Analysis
3.1 Introduction
3.2 Full Boundary Reconstruction
3.2.1 Boundary Representation Through Regular Points with First-Order Approximation of the Tangent Vector
3.2.2 Full Boundary Reconstruction
3.3 Salient Parts of the Medial Axis
3.3.1 Simplifying the Skeleton
3.4 Flux Graphs
3.4.1 Nodes and Edges
3.4.2 Qualitative Stability with Viewpoint Changes
3.5 Flux Graphs for Matching
3.5.1 Topological and Geometrical Similarity
3.5.2 The DAG Matcher
3.5.3 The Dataset and Experimental Results
3.5.4 Flux Graphs versus Shock Graphs
3.5.5 Matching 2-D Views of 3-D Models
3.6 Conclusion
References
Chapter 4: An Integrated Bayesian Approach to Shape Representation and Perceptual Organization
4.1 Shape and Perceptual Organization
4.2 Bayesian Estimation of the Shape Skeleton
4.2.1 Sketch of the Theory
4.3 Applications and Extensions
4.3.1 Decomposing Shapes into Parts
4.3.2 Tuning the Shape Model to the Environment
4.3.3 Shape Similarity
4.3.4 Figure and Ground
4.3.5 3D Shape
4.4 Discussion and Conclusion
References
Chapter 5: Perceptual Organization of Shape
5.1 Introduction
5.2 Computational Models
5.3 Feedback in the Primate Object Pathway
5.4 Generative Models of Shape
5.4.1 Localized Diffeomorphisms: Formlets
5.5 Formlet Coding
5.5.1 Formlet Bases
5.5.2 Diffeomorphism Constraint
5.5.3 Formlet Composition
5.6 Formlet Pursuit
5.7 Evaluation
5.8 Discussion
References
Chapter 6: Two-Dimensional Shape as a Mid-Level Vision Gestalt
6.1 General Introduction
6.2 Part I. Contours of Outline Shapes Derived from Everyday Objects
6.2.1 Introduction
6.2.2 The Role of Curvature Singularities in Shape Perception
Silhouette and Outline Versions
Salient Points
Straight-Line Versions
Fragmented Versions
Microgenesis of Fragmented Picture Identification
6.2.3 Conclusion
6.3 Part II. From Mid- to Lower-Level Vision: Linking Shape Perception to Perceptual Organization
6.3.1 Introduction
6.3.2 Conclusion
6.4 Part III. From Mid- to Higher-Level Vision: Linking Shape Perception to Categorization
6.4.1 Introduction
6.4.2 Shape Features, Dimensions, and Generative Transformations
6.4.3 Shape Similarity and Categorization Learning
Shape Similarity of Fourier Boundary Descriptors
Shape Similarity and Categorization Learning
6.4.4 Within-Category Shape Discrimination
6.4.5 Shape Interpretation and Perceptual Switching
6.4.6 Conclusion
6.5 General Conclusion
References
Chapter 7: Shape Priors for Image Segmentation
7.1 Image Analysis and Prior Knowledge
7.2 Explicit versus Implicit Shape Representation
7.3 Statistical Shape Priors for Explicit Shape Representations
7.3.1 Linear Shape Priors
7.3.2 Nonlinear Shape Priors
7.4 Statistical Priors for Level-Set Representations
7.4.1 Nonparametric Shape Priors
7.4.2 Dynamical Shape Priors for Implicit Shapes
7.5 Parametric Representations Revisited: Combinatorial Solutions for Segmentation with Shape Priors
7.6 Conclusion
References
Chapter 8: Observations on Shape-from-Shading in Humans
8.1 Shape Properties
Summary
8.2 Veridicality and Stability
Summary
8.3 Boundary Conditions and Contours
Summary
8.4 The Role of Lighting
Summary
8.5 Estimating Lighting Direction and Diffuseness
Summary
8.6 Prior Assumptions
Summary
8.7 Computation of Shape-from-Shading
Summary
8.8 Operation in Sub-optimal Conditions
Summary
8.9 Conclusion
References
Chapter 9: Deformations and Lighting
9.1 Introduction
9.2 Background
9.3 Robust Image Metrics for Lighting and Deformation
9.3.1 Intensity Variation
9.3.2 Interaction Between Deformations and Intensity Change
9.4 Using These Metrics for Image Comparison
9.4.1 A Deformation and Lighting Insensitive Metric for Face Recognition
9.4.2 Geodesics for Image Comparison with a Lighting-Insensitive Metric
9.5 Conclusions
References
Chapter 10: The Shape of Space
10.1 The Shape of Space
10.2 Some Spaces of Interest
10.2.1 The Pinned down Observer
10.2.2 The Pinned down Observer, Fixating the Forward Direction
10.2.3 The Pictorial Observer
10.3 The \"Polarized\" Pinned down Observer
10.4 Reanalysis of Some Pertinent Empirical Data
External local sign
Failures of \"simple geodesic arcs\"
Pointing in circles
Pointing to opposite sides
The existence of planes
The curvature of large fronto-parallels
Pointing in pictorial space
10.5 Conclusion
References
Chapter 11: The Visual Hierarchy Mirage: Seeing Trees in a Graph
11.1 Introduction
11.2 What Is Intermediate-Level Vision?
11.3 Hierarchies and Trees
11.4 Neuroanatomy is a Graph
11.5 Border Ownership as Visual Inference
11.6 Summary
References
Chapter 12: Natural Selection and Shape Perception
12.1 Introduction
12.2 Bayesian Decision Theory
12.3 A General Framework for Perception and Its Evolution
12.4 Shape as a Code for Fitness
12.4.1 Implications for Shape Perception
12.4.2 The Role of Action in the Evolution of Shape Perception
12.4.3 Perceptual Organization of Shape
12.5 Discussion
Appendix: Relation to Quantum Bayesianism
References
Chapter 13: Shape as an Emergent Property
13.1 Shape Inference
13.2 Modelling Shape
13.3 The Classical Approach
13.3.1 Drawbacks of the Classical Approach
13.4 Nonlocal Interactions
13.4.1 Higher-Order Active Contours
13.4.2 Reformulation as a Network of Nodes
13.4.3 Nonlocality via Local Interactions
13.5 Discussion
References
Chapter 14: Representing 3D Shape and Location
14.1 A Primal Sketch That Survives Eye Rotation
14.2 Translation of the Optic Center
14.2.1 Representing Surface Slant and Depth Relief
14.2.2 Representing Location
14.3 Implementation of a Universal Primal Sketch
14.4 Apparent Paradoxes in the Representation of 3D Shape and Location
References
Chapter 15: Joint Registration and Shape Analysis of Curves and Surfaces
15.1 Introduction
15.2 Shape Analysis of Curves
Mathematical Representation
Shape Matching and Geodesics
Shape Statistics of Curves
15.3 Shape Analysis of Parameterized Surfaces
Mathematical Representation
Shape Matching and Geodesics
Shape Statistics of Surfaces
15.4 Conclusion
References
Chapter 16: The Statistics of Shape, Reflectance, and Lighting in Real-World Scenes
16.1 Introduction
16.2 Lighting
16.2.1 Lighting: Scene Statistics
16.2.2 Lighting: Psychophysics
16.3 Shape and Reflectance
16.3.1 Shape and Reflectance: Scene Statistics
16.3.2 Shape and Reflectance: Psychophysics
16.4 Conclusion
References
Chapter 17: Structure vs. Appearance and 3D vs. 2D? A Numeric Answer
17.1 Introduction
17.2 Information Projection
17.3 Case I: Combining Sketch and Texture
17.4 Case II: Mixing 3D and 2D Primitives
17.5 Discussion
References
Chapter 18: Challenges in Understanding Visual Shape Perception and Representation: Bridging Subsymbolic and Symbolic Coding
18.1 Introduction
18.2 Some Useful Examples
18.3 Interpolation Processes Underlying Object Perception
18.4 Contour and Surface Processes
18.5 Contour Interpolation
18.6 Triggering Contour Interpolation
18.7 Contour Relatability
18.8 Neural Models of Contour Interpolation
18.9 Shape Perception
18.10 Constant Curvature Coding: An Example of a Bridge Between Subsymbolic and Symbolic Shape Coding
18.11 Early Symbolic Encoding of Contours: Arclets
18.12 Evidence for Constant Curvature Coding in Human Shape Perception
18.13 Connecting Contour Interpolation and Shape Descriptions
18.14 Summary
References
Chapter 19: 3D Face Reconstruction from Single Two-Tone and Color Images
19.1 Introduction
19.2 Reconstruction Ambiguities in Two-Tone Images
19.3 Shape Reconstruction with a Prior Model
Step 1: Recovery of Lighting Coefficients
Step 2: Depth Recovery
Boundary Conditions for Depth Recovery
Step 3: Estimating Albedo
19.4 Experiments
19.5 Conclusion
References
Chapter 20: Perception and Action Without Veridical Metric Reconstruction: An Affine Approach
20.1 Veridical Metric Structure from Binocular Disparities and Retinal Velocities
20.2 Is Veridical Metric Structure Used or Needed?
20.2.1 Perceived 3D Structure from Stereo and Motion Signals
20.2.2 Reach-to-Grasp Without Visual or Haptic Feedback
20.2.3 Is Veridical Metric Reconstruction Needed?
20.3 Local Affine Information and Heuristic Scaling
(a) Local Affine Information
(b) Maximizing Precision of Affine Estimates
20.3.1 Metric Tasks and Heuristic Scaling
20.4 Conclusion
References
Chapter 21: A Stochastic Grammar for Natural Shapes
21.1 Introduction
21.2 Shape Grammar
21.3 Sampling Shapes from Images
21.4 Experimental Results
References
Chapter 22: Hard-Wired and Plastic Mechanisms in 3-D Shape Perception
22.1 Introduction
22.2 Orientation and Frequency Cues for 3-D Shape Perception
22.3 Folded Surfaces
22.4 Carved Surfaces
22.5 Perceptual Strategies
22.6 Cross-Orientation Inhibition
22.7 Hebbian Learning of Matched Filters
22.8 Hardwired Cortical Anisotropies
22.9 Plastic Processes in Shape from Texture
22.10 Conclusion
References
Chapter 23: Holistic Shape Recognition: Where-to-Look and How-to-Look
23.1 Introduction
23.1.1 What Have We Learned?
23.1.2 What Were the Challenges We Faced
23.1.3 How Has Our Thinking on the Problem Changed over the Course of Research? What\'s Worked and What Hasn\'t?
Deformable Graphical Model and Co-segmentation-Recognition
Shape Jigsaw Model
Many-to-Many Shape Packing Model
23.1.4 What Are the Obstacles to the Community\'s Success?
Shape Model Representation
Model Obstacles
Shape Features
Feature Obstacles
Shape Matching
Matching Obstacles
23.2 Many-to-Many Shape Packing
23.2.1 Many-to-Many Matching as a Packing Problem
Many-to-One Matching
23.3 Shape Detection and Segmentation
Placement Score
Deformation Score
Inference for Detection
Latent SVM for Discriminative Detector Learning
Joint Many-to-One Matching
Joint Matching and Final Evaluation
23.4 Experiments on ETHZ Shape
23.5 Comments
References
Chapter 24: Shape Processing as Inherently Three-Dimensional
24.1 The Inherently Three-Dimensional Demand Characteristics of Visual Encoding
24.2 Theoretical Analysis of Shape Representation as Surface Manifolds
24.3 Neural Aspects of 3D Shape Representation
24.4 Need for the Surface Representation of 3D Shape
24.5 Hypercyclopean Form Analysis
24.6 Metric Constraints on 3D Shape Perception
24.7 Cortical Organization of 3D Shape Representation
24.8 Conclusion
References
Chapter 25: The Role of Shape in Visual Recognition
25.1 Regularity, Structure, and Form
25.1.1 The Nature of Shape
The Special Role of Shape: Invariance and Emergence
25.2 Shape Representation for Visual Recognition
25.2.1 The Days of Geometry: Blocks, Cylinders, and Acronyms
25.2.2 The Dawn of Appearance
25.2.3 Textons Everywhere
25.2.4 Half a Century of Evolution-A Critique
Complexity of Shape Models
Real World Benchmarks and Performance
Dimensionality and Flexibility
25.3 Quo Vadis?
25.3.1 Shape: Representing Statistical Dependencies Between Parts
Compositionality
A Compositional Shortcut
25.4 Conclusion and Outlook
References
Chapter 26: Human Object Recognition: Appearance vs. Shape
26.1 Cortical Pathways for Visual Processing
26.2 The Ventral Pathway
26.2.1 The Lateral Occipital Complex (LOC): An Area Critical for Perceiving Shape
26.2.2 Coding for Shape vs. Surface in LOC
26.2.3 Different Subregions for the Processing Different Stimulus Dimensions
26.2.4 How Efficient are Line Drawings for Object Recognition?
26.3 What Aspects of Shape are Coded in LOC?
26.3.1 Evidence for the Representation of Objects in Terms of Parts Rather than Local Features, Templates or Concepts
26.3.2 Relations Between Parts and Between Objects
26.4 Coda
References
Chapter 27: Shape-Based Object Discovery in Images
27.1 Introduction
27.2 A Brief Review of Our Approach
27.3 Image Representation Using Shapes and Shape Description
27.4 Constructing the Graph of Pairs of Image Contours
Case 1
Case 2
27.5 Coordinate-Ascent Swendsen-Wang Cut
27.5.1 Bayesian Formulation
27.5.2 Inference Using the CASW Cut
27.6 Results
27.7 Conclusion
References
Chapter 28: Schema-Driven Influences in Recovering 3-D Shape from Motion in Human and Computer Vision
28.1 Introduction
28.2 The Hollow-Mask Illusion for Humans
28.3 The Hollow-Mask Illusion and Computer Vision
28.3.1 Model and Algorithm
28.3.2 Algorithm Input-Results
28.4 Discussion
References
Chapter 29: Detecting, Representing and Attending to Visual Shape
29.1 Introduction
29.2 An Early Use of Curvature: Curvature-Tuned Smoothing
29.3 2DSIL: End-Stopped and Curvature Computations for Silhouette Recognition
29.4 Conclusions
References
Chapter 30: Toward a Dynamical View of Object Perception
30.1 Object Perception: What Happens When?
30.2 Systematic Tests of Whether Past Experience Influences Figure-Ground Perception
30.3 Where Is Object Perception Accomplished?
30.4 Competition
30.5 High-Level Object Memories and Dynamical Interactions Between High and Low Levels
30.6 Conclusion
References
Chapter 31: Modeling Shapes with Higher-Order Graphs: Methodology and Applications
31.1 Introduction
31.1.1 Main Obstacle-Extrinsic Factors
31.1.2 Key Strategy-Encoding Shape Invariance in Higher-Order Graphs
31.2 Nonrigid 3D Surface Matching
31.3 Pose-Invariant Prior and Knowledge-Based Segmentation
31.4 Conclusion
References
Chapter 32: Multisensory Shape Processing
32.1 Introduction
32.2 Measuring Perceptual Spaces
32.3 Multisensory Perceptual Spaces
32.4 Visuo-haptic Face Recognition-The Role of Expertise
32.5 Summary and Open Questions
References
Chapter 33: Shape-Based Instance Detection Under Arbitrary Viewpoint
33.1 Introduction
33.2 Invariant Methods
33.3 Non-invariant (View-Based) Methods
33.4 Ambiguities
33.5 Conclusion
References
Index