Energy Minimization Methods in Computer Vision and Pattern Recognition: 6th International Conference, EMMCVPR 2007, Ezhou, China, August 27-29, 2007, ... (Lecture Notes in Computer Science, 4679)

✍ Scribed by Alan L. Yuille (editor), Song-Chun Zhu (editor), Daniel Cremers (editor), Yongtian Wang (editor)

• Publisher: Springer
• Year: 2007
• Tongue: English
• Leaves: 505
• Category: Library

✦ Synopsis

This volume contains the papers presented at the Sixth International Conference on Energy Minimization Methods on Computer Vision and Pattern Recognition (EMMCVPR 2007), held at the Lotus Hill Institute, Ezhou, Hubei, China, August 27–29, 2007. The motivation for this conference is the realization that many problems in computer vision and pattern recognition can be formulated in terms of probabilistic inference or optimization of energy functions. EMMCVPR 2007 addressed the critical issues of representation, learning, and inference. Important new themes include pr- abilistic grammars, image parsing, and the use of datasets with ground-truth to act as benchmarks for evaluating algorithms and as a way to train learning algorithms. Other themes include the development of efficient inference algorithms using advanced techniques from statistics, computer science, and applied mathematics. We received 140 submissions for this workshop. Each paper was reviewed by three committee members. Based on these reviews we selected 22 papers for oral presen- tion and 15 papers for poster presentation. This book makes no distinction between oral and poster papers. We have organized these papers in seven sections on al- rithms, applications, image parsing, image processing, motion, shape, and thr- dimensional processing. Finally, we thank those people who helped make this workshop happen. We - knowledge the Program Committee for their help in reviewing the papers.

✦ Table of Contents

Title Page
Preface
Organization
Table of Contents
An Effective Multi-level Algorithm Based on Simulated Annealing for Bisecting Graph
Introduction
Mathematical Description
Motivation
An Effective Multi-level Simulated Annealing Refinement Algorithm
Experimental Results
Conclusions
References
Szemer´edi’s Regularity Lemma and Its Applications to Pairwise Clustering and Segmentation
Introduction
Szemer´edi’s Regularity Lemma
Finding Regular Partitions in Polynomial Time
The Regularity Lemma and Pairwise Clustering
Experimental Results
Conclusions
References
Exact Solution of Permuted Submodular MinSum Problems
Introduction
Notations and Definitions
Transforming a Permuted Submodular Problem into a Submodular One
Conclusion
References
Efficient Shape Matching Via Graph Cuts
Introduction
Metrics for Shapes
Related Work and Contribution
Shape Matching Via Graph Cuts
Connection Between Matchings and Graph Cuts
Graph Construction
Equivalence to Shortest Path Formulation
Integral Invariants and Curvature
Experimental Results
Matching with Articulated Parts
Robustness to Noise
Comparison with Dynamic Time Warping
Graph Cut Algorithm
Conclusion and Future Work
References
Simulating Classic Mosaics with Graph Cuts
Introduction
Energy Optimisation with Graph Cuts
Simulating Classic Mosaics with Graph Cuts
Notation and Definitions
Generating Tile Orientations
Generating Mosaic Layers
Stitching Two Mosaic Layers
Experimental Results
References
An Energy Minimisation Approach to Attributed Graph Regularisation
Introduction
Graph Regularisation and Manifold-Constrained Energy Minimisation
Riemannian Manifolds
The Ginzburg-Landau Functional
Energy Minimisation on Graphs
Introducing Constraints
Implementation Issues
Applications
Robust Normal Estimation for Photometric Stereo
Semi-supervised Image Segmentation
Experiments
Photometric Stereo
Image Segmentation
Conclusions
References
A Pupil Localization Algorithm Based on Adaptive Gabor Filtering and Negative Radial Symmetry
Introduction
Eye Detection
Face Tilt Detection and Adjustment
Y-Coordinate Localization
X-Coordinate Localization Based on Gabor Filtering
Improvement on X-Coordinate Localization of Eye Windows
Enhancing the Double Peak Property of the Y-Directional Integral Projection
Automatic Gabor Filter Selection Based on PCA (Principle Component Analysis)
Pupil Localization
Experiment Result
Conclusion
References
Decomposing Document Images by Heuristic Search
Introduction
Document Image Representation
Data Preparation
Features
GeneratingWord-Graph – The Primitive Layer
Generating Zone Hypotheses
Models and Learning
A Likelihood Model for Zones
A Prior Model for Zone-Map
Zone Inference by Heuristic Search
Generating Costs and Constraints from Learned Statistics
The $A^$ Algorithm
Data Set and Experimental Results
Conclusions
References
CIDER: Corrected Inverse-Denoising Filter for Image Restoration
Introduction
Wiener Filter and Wavelet-Domain Wiener Filter
Wiener Filter
Wavelet-Domain Denoising
AnalysisofCIDER
Mean Squares Error
Regularization Parameter
Simulation Results
Conclusion
References
Skew Detection Algorithm for Form Document Based on Elongate Feature
Introduction
Skew Detection for Form Document Based on Elongate Feature
Extracting Connected Components with DRG
Optimal Horizontal Line Extraction
Skew Estimation
Experiment Result Analysis
Conclusion
References
Active Appearance Models Fitting with Occlusion
Introduction
AAMs Construction with Occlusion
AAMs Fitting with Occlusion
Basic AAMs Fitting
Adaptive Fitting Algorithm
Experiments
Fitting Example
Fitting Evaluation
Discussion
References
Combining Left and Right Irises for Personal Authentication
Introduction
Feature Extraction and Matching
Iris Preprocessing
IrisCode Extraction
IrisCode Comparison
ScoreFusion
Experimental Results and Analysis
Database
Difference Between Left and Right Irises
Matching Tests
Accuracy Tests
Conclusions
References
Bottom-Up Recognition and Parsing of the Human Body
Introduction
Overview of Our Parsing Method
Multiple Segmentations
Shape Comparison
Parse Rule Application Procedure
Learning with Other Features
Features
Learning the Classifiers
Results Using Shape Only
Segmentation Scoring
Joint Position Scoring
Results with Learning
Conclusion
References
Introduction to a Large-Scale General Purpose Ground Truth Database: Methodology, Annotation Tool and Benchmarks
Introduction
Methodology: Representation and Organization of Labeling Information
Region Segmentation and Semantic Annotation
Sketch Graph Representation
Hierarchical Decomposition and Parsing Graph
And-Or Graph Knowledgebase and Bottom-Up/Top-Down Labeling Procedure
Global 3D Geometry Information
Annotation Tool: Integrating Functional Modules
Database Statistics, Subsets and Benchmarks
Conclusions and Future Works
References
An Automatic Portrait System Based on And-Or Graph Representation
Introduction
The And-Or Graph for Portrait
The Automatic Portrait System
Face Subsystem
Hair and Collar Subsystem
Multiple Style Rendering
Experiments
References
Object Category Recognition Using Generative Template Boosting
Introduction
Generative Template Synthesis
The Learnable And-Or Graph
Template Synthesise
Inference and Experiments
Discriminative Prune
Top-Down Verification
Comparison
Summary
References
Bayesian Inference for Layer Representation with Mixed Markov Random Field
Introduction
Problem Formulation of the 2.1D Sketch
Graph Representation
Definition of the Problem Domain
Bayesian Formulation
Inference
Discriminative Probabilities on Edges
Algorithm
Experiments
Discussion
References
Dichromatic Reflection Separation from a Single Image
Introduction
Reflection Image Decomposition
Information Extraction
Specular Estimation Map
Diffuse Estimation Map
Boundary Classification
Fuzzy Integral
Fuzzy Decision
Defect Compensation
Experimental Results
Concluding Remarks and Future Works
References
Noise Removal and Restoration Using Voting-Based Analysis and Image Segmentation Based on Statistical Models
Introduction
Related Work
Tensor Voting
Tensor Field in Physical Analogy
Tensor Encoding
Voting Communication
Noise Removal and Restoration Using Generalized Adaptive Vector Sigma Filters in Tensor Voting
Image Analysis in Tensor Voting Space
Region Restoration Using Vector Sigma Filter
Segmentation Using Statistical Method
Experimental Results
Conclusions
References
A Boosting Discriminative Model for Moving Cast Shadow Detection
Introduction
Related Work
Proposed Methods
Shadow Model
Discriminative Random Fields for Cast Shadow Detection
Data Potentials
Smoothness Potentials
Boosting Learning for Data Potentials
Color Invariance Subspace
Texture Invariance Subspace
Boosting for Data Potentials
Experiments Result
Performance Evaluate
Experiments
Conlusions
References
An a Contrario Approach for Parameters Estimation of a Motion-Blurred Image
Introduction
The Model
The Helmholtz Principle
$\epsilon$- Meaningful Parallel Segments
Adapted K-Means Cluster Algorithm
Experimental Results
Conclusion
References
Improved Object Tracking Using an AdaptiveColour Model
Introduction
Defining Similarity by Histograms
Selecting the Best Colour Space Model
Normalized Chamfer Distance Transform
Experimental Evaluation
Conclusions
References
Vehicle Tracking Based on Image Alignment in Aerial Videos
Introduction
Tracking Algorithm Based on Image Alignment
Outlier Rejection
Kalman Filter
Occlusion Treating
Occlusion Detection
Reappearance Detection
Reappearance Verification
Experiment Results and Conclusion
References
Probabilistic Fiber Tracking Using Particle Filtering and Von Mises-Fisher Sampling
Introduction
Tracking Algorithm
Global Tracking Model
Recursive Posterior Using Particle Filtering
Algorithm Ingredients
Observation Density
Prior Density
Importance Density Function
Algorithm Outline
Experimental Results
Synthetic Dataset
Brain Diffusion MRI
Conclusion
References
Compositional Object Recognition, Segmentation, and Tracking in Video
The Rational for Compositionality
Related Work
Compositional Approach to Video Analysis
Atomic Compositional Constituents
Compositions of Parts
Temporal Grouping of Compositions
Obtaining Multiple Segmentation Hypotheses
Model Selection to Identify Reliable Segmentation Hypotheses
Compositional Shape Model for Object Recognition
Evaluation
Evaluating the Building Blocks of the Composition System
Multi-object Recognition
Analyzing the Relevance of Compositions
Discussion
References
Bayesian Order-Adaptive Clustering for Video Segmentation
Introduction
Background
Clustering with Mixture Models
Conjugate Models
Dirichlet Process Mixture Models
Order-Adaptive Clustering in Time Series
MCMC Inference for High-Throughput Problems
Sampling in Time Series
Multiscale Sampling
Experiments
Conclusions
References
Dynamic Feature Cascade for Multiple Object Tracking with Trackability Analysis
Introduction
Multiple Object Tracking
Object Trackability
Definition of Trackability
Tracking Features
Computation of Trackability
Dynamic Feature Cascade
Experiments
Conclusion
References
Discrete Skeleton Evolution
Introduction
Related Work
Discrete Skeleton Evolution
Experiments and Discussions
Test on Kimia’s Dataset
Comparison with Other Methods
Test on Skeletons with Loops
Some Experimental Details
Conclusions and Future Work
References
Shape Classification Based on Skeleton Path Similarity
Introduction
Background
Obtain Skeleton by Skeleton Pruning
Shape Representation with Skeleton Paths
Bayesian Classification
Experiments
Conclusions
References
Removing Shape-Preserving Transformations in Square-Root Elastic (SRE) Framework for Shape Analysis of Curves
Introduction
Square-Root Elastic (SRE) Framework
Square-Root Representations of Curves
Path Straightening Flows for Computing Geodesics
Removing Shape Preserving Transformations
Elastic Shape Space ($\cal S$)
Geodesics in ${\cal C}/({\mathbb{S}^1} \times SO(n))$)
Geodesics in $\cal S$
Experimental Results
Summary
References
Shape Analysis of Open Curves in $\real^3$ with Applications
Introduction
Shape Spaces of Open Curves in $\real^3$
Representations of Open Curves
Re-parametrization and Orientation Orbits of Curves
Geodesics and Distances in Shape Space ${\cal S}$
Penalty Function on Optimal Matching $\phi^$
Discrete Implementation
Sample Statistics of Shapes of Curves
Clustering Bundles of Curves
Summary
References
Energy-Based Reconstruction of 3D Curves for Quality Control
Introduction
Problem Formulation
Properties of NURBS Curves
Projective Invariance
Control Point Insertion
Optimization
Distance Minimization
Gradient Energy Minimization
Distance Versus Gradient Energy
Curve Estimation
Optimization on the Control Points
Control Point Insertion
Algorithm
Experimental Evaluation
Virtual Images
Real Images
Conclusions
References
3D Computation of Gray Level Co-occurrence in Hyperspectral Image Cubes
Introduction
Hyperspectral Volumetric Texture Analysis
Methods and Materials
3D GLCM Computation
Semi-variance Analysis
Test Data
Results and Discussions
Conclusion and Future Work
References
Continuous Global Optimization in Multiview 3D Reconstruction
Introduction
Convex Formulations of Image Segmentation
A Continuous Energy Model for Multiview Reconstruction
Continuous Global Optimization
An Equivalent Convex Formulation
Fast Minimization by Successive Overrelaxation
Experiments
Summary
References
A New Bayesian Method for Range Image Segmentation
Introduction
Image Segmentation by Randomized Region Growing
Surface Modeling
Region Growing by Randomized Region Seed Sampling
Edge Regularization by Bayesian Inference
MAP-MRF Pixel Labeling
Computation of the Optimal Solution
Experimentation and Discussion
Evaluation Framework
Parameter Selection
Performance Evaluation and Comparison
Conclusion
References
Marked Point Process for Vascular Tree Extraction on Angiogram
Introduction
Model for Approximate Vessel Centerline Extraction
Double Area Model
The Data Term
Optimization by Data-Driven MCMC
Experiments on Simulation of the Point Process
Extraction of Coronary Artery Edge with Marker Watershed
Discussion
References
Surface Reconstruction from LiDAR Data with Extended Snake Theory
Introduction
Surface Reconstruction from LiDAR Data
Proposed Algorithm
Snake Theory
Extended Snake Theory
Experiments
Conclusions
References
Author Index