Computer Vision - ACCV 2010

Name: Computer Vision - ACCV 2010 | 10th Asian Conference on Computer Vision, Queenstown, New Zealand, November 8-12, 2010, Revised Selected Papers, Part III
Brand: Springer
Price: 96.29 EUR
Availability: OnlineOnly

10th Asian Conference on Computer Vision, Queenstown, New Zealand, November 8-12, 2010, Revised Selected Papers, Part III

Reinhard Klette Ron Kimmel Akihiro Sugimoto(Herausgeber*in)

Springer (Verlag)

Erschienen am 17. Februar 2011

XXI, 790 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-642-19318-7 (ISBN)

96,29 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Title
Preface
Organization
Table of Contents
Posters on Day 2 of ACCV 2010
Approximate and SQP Two View Triangulation
Introduction
Optimal Triangulation
Approximate Triangulation
Theoretical Performance Bounds
Local SQP Correction
Empirical Performance Evaluation
Simulation Setup
Simulation Results
Computational Efficiency
Validation with Real World Data
Conclusions and Discussion
References
Adaptive Motion Segmentation Algorithm Based on the Principal Angles Configuration
Introduction
Our Proposal
Notation
Issues Regarding the Behaviour of Principal Angles
Rank Selection via Principal Angles Clusterization (PAC)
Sum of Clusterization-Based Affinity (SCbA)
Summary of Our Proposal
Experiments
Conclusions and Perspectives
References
Real-Time Detection of Small Surface Objects Using Weather Effects
Introduction
Weather Effects on Vision
Atmospheric Scattering Model
Degradation Analysis
Detection Using Weather Effects
Dark Channel Prior for Surface Objects
Depth Map from a Single Image
Detecting the Horizon
Spatial-Variant Thresholds
Experimental Results
Discussions and Conclusions
References
Automating Snakes for Multiple Objects Detection
Introduction
Quad Tree Based Snake Initialization
Snake Validation Using Boosting
Results and Discussions
Oil Sand Images
Leukocyte Images
Interpretation of Results
Conclusion and Fututre Works
References
Monocular Template-Based Reconstruction of Smooth and Inextensible Surfaces
Introduction
Related Work on Inextensible Surface Reconstruction
Convex Formulation of the Upper Bound Approach with Noise in All Images
Noise in the Template Only
Noise in Both the Template and the Input Images
Smooth and Inextensible Surface Reconstruction
Parametric Surface Model
Surface Reconstruction as a Least-Squares Problem
Experimental Results
Experiments on Synthetic Data
Experiments on Real Data
Conclusion
References
Multi-class Leveraged k-NN for Image Classification
Introduction
Method
Problem Statement and Notations
(Leveraged) k-Nearest Neighbors
Multiclass Surrogate Risk Minimization
MLNN: Multi-class Leveraged k-NN Rule
Properties of MLNN
Experiments
Scene Categorization
Categorization Results
Conclusion
References
Video Based Face Recognition Using Graph Matching
Introduction
Face Image Representation
Feature Point Extraction
Feature Description with Local Binary Pattern
Image Graph Construction
Probabilistic Graph Appearance Model
Adaptive Matching and Recognition
Experiments
Conclusion
References
A Hybrid Supervised-Unsupervised Vocabulary Generation Algorithm for Visual Concept Recognition
Introduction
Visual Word Generation
Hybrid Supervised-Unsupervised Approach
Relation to the Baseline Procedures
Datasets
Experimental Setup
Results
Concept Recognition Performance
Runtime Considerations
Label Distribution Statistics across Visual Words
Conclusion
References
Image Inpainting Based on Probabilistic Structure Estimation
Introduction
ProposedMethod
Initial Image Construction
Patch Selection
Experiments
Inpainted Results
Subjective Evaluation of Image Quality
Computational Time
Conclusion
References
Text Localization and Recognition in Complex Scenes Using Local Features
Introduction
Related Work
Local Feature-Based Approach
Method of Building TCFD
Voting and Geometric Verification Algorithm
Segmentation Algorithm and Multiple-Character Recognition
The Datasets
Experimental Results
Conclusion
References
Pyramid-Based Multi-structure Local Binary Pattern for Texture Classification
Introduction
Local Binary Pattern
The LBP Methods
Structures Extracted by Conventional LBP
Multi-structure Local Binary Pattern
Image Pyramid
Rotation Invariant Multi-structure Local Binary Pattern
Similarity Metric
Experimental Results
Calculation of Distance Weights
Implementation of Multi-structure Local Binary Pattern
Conclusions
References
Unsupervised Moving Object Detection with On-line Generalized Hough Transform
Introduction
ProposedWork
Automatic On-line Instance Selection
On-line Extremely Randomized Trees
Object Detection
Experiments
The PETS2006 Dataset
The i-LIDS Dataset
The Traffic Dataset
Conclusions
References
Interactive Event Search through Transfer Learning
Introduction
Overview and Contributions of Proposed Approach
Relation to Existing Work
Relevance Feedback for Video Retrieval
Basics of Transfer Learning
Transfer Learning with Multiple Source Tasks
Transfer Learning for Relevance Feedback Search
Relevance Feedback Using Transfer Learning for Activity Search
Scoring Videos and Relevance Feedback
Transferring Knowledge from a Source Task
Determining Which Source Task to Transfer From
Integrating Relevance Feedback with Transfer Learning
Experiments
Conclusion
References
A Compositional Exemplar-Based Model for Hair Segmentation
Introduction
Compositional Exemplar-Based Model
Learning Similarity Function by SLS Features
Introduce NLC Constraint into CEM
Segmentation with Dual-Level Conditional Random Field
Experimental Result
Conclusion
References
Descriptor Learning Based on Fisher Separation Criterion for Texture Classification
Introduction and Previous Work
Texture Image Representation by FBL Descriptor
The Learning Stage
The Training Stage
The Testing Stage
Experimental Design and Results
Rotation Invariant Texture Classification
Material Categorization
Texture Classification under Variant Imaging Conditions
Face Recognition
Discussion
Conclusions
References
Semi-supervised Neighborhood Preserving Discriminant Embedding: A Semi-supervised Subspace Learning Algorithm
Introduction
Graph Embedding and Graph Based NPE and LDA
Graph Embedding View of Subspace Learning
Graph View of NPE
Graph View of LDA
Semi-supervised Neighborhood Discriminant Embedding
The Objective Function
The Algorithm
Experiments and Discussions
Data Sets
Face Representation
Face Recognition
Conclusion
References
Segmentation via NCuts and Lossy Minimum Description Length: A Unified Approach
Introduction
Related Work
k-Way Normalized Cuts Criterion
Lossy Minimum Description Length Criterion
LMDL-NCuts: A Combined and Unified Criterion
The Combined Criterion
Initializing
Construct Affinity Matrix
The Algorithm
Experiments
Feature Construction
Adaptively Select the Number of Segments
Comparison with LMDL Based Algorithm
Qualitative and Quantitative Comparison
Conclusion
References
A Phase Discrepancy Analysis of Object Motion
Introduction
The Theory
Phase Discrepancy and Ego-motion
Approximating the Phase Discrepancy
Eliminating Boundary Effects
Experiments
Implementing the Phase Discrepancy Algorithm
A New Database for Moving Object Detection
Performance Evaluation
Comparison to Previous Methods
Database Consistency
Discussion and Future Work
Sources of Errors
Connections to Spectral Residual
Concluding Remarks
References
Image Classification Using Spatial Pyramid Coding and Visual Word Reweighting
Introduction
Related Work
Spatial Pyramid Coding and Visual Word Reweighting
Spatial Pyramid Coding
Visual Word Reweighting
Experiments
Conclusion
References
Class-Specific Low-Dimensional Representation of Local Features for Viewpoint Invariant Object Recognition
Introduction
Related Work
Low-Dimensional Feature Representation by Supervised Multidimensional Scaling (sMDS)
Experimental Results
Conclusions
References
Learning Non-coplanar Scene Models by Exploring the Height Variation of Tracked Objects
Introduction
Camera Projection Model
Image Patch Model
Processing of Tracking Observations
Motion Tracking
Walkable Regions
Height Variation across Multiple Planes
Image Segmentation to Scene Planes
Segmentation of Walkable Region
Global Motion Variety
3D Scene Model Estimation
Estimating Average Heights
Altitude Estimation
Dataset and Results
Dataset
Results and Evaluation
Conclusion and Future Work
References
Optimal Regions for Linear Model-Based 3D Face Reconstruction
Introduction
Preliminaries
Facial Data
Linear Subspaces for Reconstruction
Automatic Segmentation of Facial Regions
Optimal Region Blending
Algorithm Derivation
Final Algorithm
Evaluation
Experiments
Conclusion
References
Color Kernel Regression for Robust Direct Upsampling from Raw Data of General Color Filter Array
Introduction
Interpolation with Kernel Regression
Steering Kernel P418takeda2007kernel
Proposed Color Kernel Regression
Simultaneous Color Kernel Regression
Local Density Parameter Optimization
Iterative Color Kernel Regression
Experiments
Direct Upsampling from the Raw Data of the CFA Pattern
Interplation from Irregularly Sampled Data
Conclusions
References
The Large-Scale Crowd Density Estimation Based on Effective Region Feature Extraction Method
Introduction
System Architecture
Crowd Density Estimation
Region Feature Extracted and Modified
Crowd Feature Extraction Based on Region Extraction
Density Estimation Based on Support Vector Machine
Experiments and Results
Conclusion
References
TILT: Transform Invariant Low-Rank Textures
Introduction
Transform Invariant Low-Rank Textures
Low-Rank Textures
Deformed and Corrupted Low-Rank Textures
Solution by Iterative Convex Optimization
Fast Algorithm Based on Augmented Lagrangian Multiplier
Additional Constraints and Implementation Details
Experiments and Applications
Range of Convergence of TILT
Robustness of TILT
Shape from Low-Rank Texture Detection
Rectifying Different Categories of Low-Rank Textures
References
Translation-Symmetry-Based Perceptual Grouping with Applications to Urban Scenes
Introduction
Translation-Symmetry-Based Perceptual Grouping
Low-Level Feature Aggregation
Grouping of Low-Level Features
Translation-Symmetry-Based Grouping
Quantitative Evaluation
Application
Super-Resolution from a Single View
Frontal View Facades Estimation from a Single View
Conclusion
References
Towards Hypothesis Testing and Lossy Minimum Description Length: A Unified Segmentation Framework
Introduction
Segmentation by Hypothesis Testing
Statistical Model of Image Segmentation
Multi-normal Population Hypothesis Test
Segmentation by Hypothesis Testing
Segmentation by Hypothesis Testing and Lossy Data Coding: A Unified Approach
Noise Modified Likelihood Ratio
Lossy Minimum Description Length Segmentation: A Hypothesis Testing Perspective
Experiments
Feature Construction
Verification
Conclusion
References
A Convex Image Segmentation: Extending Graph Cuts and Closed-Form Matting
Introduction
Previous Works
Graph Cuts
Closed-Form Matting
A Convex Segmentation
Specializations
Remarks
Optimization
Reformulation
A Primal-Dual Interior Point Method
A Newton Step
Applications
Matting
Segmentation
Conclusion
References
Linear Solvability in the Viewing Graph
Introduction
The Viewing Graph
The Solution Set of the Viewing Graph
The Viewing Graph System
Characterisation of the Extended and Non-redundant Solution Sets
The Linear Minimal Solution Superset of a Nonlinear Viewing Graph
Topology and Solvability
An Inductively Constructible Set of One Solution Linear Viewing Graphs
Conclusions and Discussion
References
Inference Scene Labeling by Incorporating Object Detection with Explicit Shape Model
Introduction
Representation
Bottom-Up Candidates Generation
Candidacy Graph Construction
Probabilistic Formulation
Inference by Composite Cluster Sampling
Experiments
Summary
References
Saliency Density Maximization for Object Detection and Localization
Introduction
Related Work
Saliency Map
From Saliency Map to Salient Object
The Proposed Method
Existing Schemes
Our New Formulation
Our Algorithm
Experimental Results
Database
Comparison MSD with Exhaustive Search
Comparison MSD with MSR
Evaluation on Different Saliency Maps
Parameter Evaluation
Time Complexity
Conclusion
References
Modified Hybrid Bronchoscope Tracking Based on Sequential Monte Carlo Sampler: Dynamic Phantom Validation
Introduction
SMC Sampler-Based Bronchoscope Tracking
SMC Sampler
Definitions of Bronchoscopic Camera Motion
SMC Sampler for Camera Motion Estimation
Experimental Results
Discussion
Conclusions and Future Work
References
Affine Warp Propagation for Fast Simultaneous Modelling and Tracking of Articulated Objects
Introduction
Related Work
Image Alignment via Affine Warps
Affine Warp Propagation
Non-probabilistic Graphical Model Definition
Message Definition
Message Correction
Learning Spatial Relations and Weights
Experiments
Discussion
References
kPose: A New Representation For Action Recognition
Introduction
Feature for Pose Description
Pose Weighted Distribution Model
Intra- and Inter-class Distance
Pose-Weighted Distribution Model Learning
KPose Based Action Representation
Action Classification
Experiment
Conclusion
References
Identifying Surprising Events in Videos Using Bayesian Topic Models
Introduction
Motivation
Related Work
Our Approach
Activity Representation
Objects as Space Time Tubes
Trajectories
Modeling of Typical Activities Using LDA
Surprise Detection
Bayesian Surprise and the LDA Model
Experimental Results
Dataset
Trajectory Classification
Surprising Events
Conclusions
References
Face Detection with Effective Feature Extraction
Introduction
Simple Edge Descriptors
HoG and LBP
Rectangular Features Based on HoG and LBP
Joint Rectangular Features
Experiments
Frontal Face Detection
Discussion
Conclusion
References
Multiple Order Graph Matching
Introduction
Multiple Order Tensor Formulation
Tensor Power Iteration
Symmetric Higher-Order Power Iteration
Symmetric Multiple Order Tensor Power Iteration
Implementation Issues
Potential Details
Sampling
Experiments
Synthetic Data
Real Images
Conclusion
References
Abstraction and Generalization of 3D Structure for Recognition in Large Intra-Class Variation
Introduction
Motivation and Related Work
Dataset
Creating COMPAS
Representation
Features and Measures
Classifier
Recognition
Experiment 1
Experiment 2
Future Direction and Conclusion
References
Exploiting Self-similarities for Single Frame Super-Resolution
Introduction
Proposed Algorithm
Exploiting Self-similarities to Generate Example Pairs
Exploiting Group Self-similarities to Construct High-Resolution Images
Experimental Results
Concluding Remarks
References
On Feature Combination and Multiple Kernel Learning for Object Tracking
Introduction
Related Work
Algorithms
The Multiple Kernel Learning
Spatial Pyramid Histogram Feature
Online Updating
Our Tracking Framework
Experimental Results
Conclusions
References
Correspondence-Free Multi Camera Calibration by Observing a Simple Reference Plane
Introduction
Single Camera Calibration Using a Reference Plane
Zhang's Method P490bib:zhang2000-calib
Corresponding Points between Two Cameras
Multi Camera Calibration
Observation of a Reference Plane
Initial Estimation of the Relative Poses of Cameras and Its Optimization
Proposed Method
Experiments Using Real Images
Stability Evaluation with a Polka-Dot Pattern
Building a Simple Shape Measurement System
Conclusion
References
Over-Segmentation Based Background Modeling and Foreground Detection with Shadow Removal by Using Hierarchical MRFs
Introduction
Related Work
Over-Segmentation Based Background Modeling
Markov Random Fields for Classification
Energy Minimization
Hierarchical Belief Propagation Optimization over MRFs
Experimental Results
Conclusion
References
MRF-Based Background Initialisation for Improved Foreground Detection in Cluttered Surveillance Videos
Introduction
Foreground Segmentation Framework
Proposed Background Initialisation Algorithm
Similarity Criteria for Lables
Markov Random Field (MRF) Framework
Node Initialisation
Calculation of the Energy Potential
Modified Background Model for Foreground Segmentation
Experiments
Main Findings
References
Adaptive eLBP for Background Subtraction
Introduction
From LBP to Adaptive LBP
Background Subtraction Based on Adaptive LBP
Experimental Results
Qualitative Evaluation
Quantitative Evaluation
Computation Cost
Conclusion
References
Continuous Surface-Point Distributions for 3D Object Pose Estimation and Recognition
Introduction
Related Work
Object Model
Inference
Learning
Modeling a Point-Cloud Reconstruction
Learning from Multiple Views
Evaluation
Cluttered-Scene Pose Estimation
Object Detection and Recognition
Conclusion
References
Efficient Structured Support Vector Regression
Introduction
Our Approach
Problem Description
Our Max-Margin Formulation
Parameter Learning and Inference Details
Relationship to Existing Work
Experiments
Binary Synthetic Images
Middlebury Stereo Datasets
Conclusion
References
Cage-Based Tracking for Performance Animation
Introduction
Related Works
Problem Statement and Contributions
Background and Notations
Pipeline Overview
Linear Least Squares Cage Fitting
Harmonic Subspace Deformation
Laplacian-Based Subspace
Linear Subspace Constraints
Results and Evaluation
Discussion
Conclusion and Future Works
References
Modeling Dynamic Scenes Recorded with Freely Moving Cameras
Introduction
Reconstruction Procedure
Results
Conclusions
References
Learning Image Structures for Optimizing Disparity Estimation
Introduction
Previous Approaches
Overview of Our Approach
Details of Solution
Triangulation
Structures and Clusters
Disparity Calculation
Results
Conclusion
References
Image Reconstruction for High-Sensitivity Imaging by Using Combined Long/Short Exposure Type Single-Chip Image Sensor
Introduction
Use of Single-Chip Image Sensor for Combined Long/Short Exposure Type Image Sensing
Arrangement of Long-Term and Short-Term Exposed Pixels
Arrangement of RGB Color Filters
Reconstruction of Color Images
Cost Function
Minimization of the Cost Function
Experimental Results
Simulation
Development of Dedicated Image Sensor
Discussion
Concluding Remarks
References
On the Use of Implicit Shape Models for Recognition of Object Categories in 3D Data
Introduction
3D Implicit Shape Model
Codebook
Codebook Size
Sharing Codewords among Categories
Codeword Activation Strategy
Votes Weighting Strategy
Experimental Results
Discussion
Conclusions
References
Phase Registration of a Single Quasi-Periodic Signal Using Self Dynamic Time Warping
Introduction
Related Work
Phase Registration
Problem Statement
Pseudo Period Estimation
Self Dynamic Time Warping
Phase Sequence Optimization
Experiments
Simulation Data
Real Data
Conclusion
References
Latent Gaussian Mixture Regression for Human Pose Estimation
Introduction
Gaussian Mixture Regression
Latent GMR Body Pose Estimation
Locality Preserving Projections
Learning
Inference
Experiments
Data Sets
Comparison
Conclusions and Future Work
References
Top-Down Cues for Event Recognition
Introduction
Overall Architecture
Top-Down Attention Histogram for Event Representation
STIPs: Video Representation
Top-Down Attention Histogram
Experiment
Data Sets
Classifier
HOHA
TRECVID
Conclusion
References
Robust Photometric Stereo via Low-Rank Matrix Completion and Recovery
Introduction
Photometric Stereo as Low-Rank Matrix Recovery with Sparse Errors
Efficient Solution via Convex Programming
Convex Relaxation and Modification
Fast Algorithm Using Augmented Lagrange Multiplier
Experiments
Quantitative Evaluation with Synthetic Images
Qualitative Evaluation with Real Images
Discussion and Future Work
References
Robust Auxiliary Particle Filter with an Adaptive Appearance Model for Visual Tracking
Introduction
Related Works
Receding Horizon Auxiliary Particle Filter
Particle Filter for Appearance Based Visual Tracking
Receding Horizon Auxiliary Particle Filter
Adaptive Appearance Learning and Robust Observation Function
Experiments
Comparison with IVT
Comparison with Geometric PF with OIF
Conclusion
References
Sustained Observability for Salient Motion Detection
Introduction
State-Space Representation
Dynamic Textures
Observability of a State Space Model
Observability from Outputs and Salient Motion
Quantitative Measure for Observability
Sustained Observability
Experiments
Quantitative Evaluation
Comparison
Conclusion
References
Markerless and Efficient 26-DOF Hand Pose Recovery
Introduction
Related Work
Methodology
Observation Model
Hand Model
Hypothesis Evaluation
Particle Swarm Optimization
Exploiting Parallelism
Experimental Evaluation
Quantitative Evaluation Based on Synthetic Data
Experiments with Real World Images
Discussion
References
Stick It! Articulated Tracking Using Spatial Rigid Object Priors
Introduction
Articulated Tracking
Related Work
Projected Spatial Priors
The KKB Tracker
Spatial Object Interaction Prior
Two Dimensional Object Information
Visual Measurements
Experimental Results
Discussion
References
A Method for Text Localization and Recognition in Real-World Images
Introduction
Problem Description
Text Localization and Recognition
MSER Detection
Character and Non-character Classification
Text Line Hypothesis Formation
Geometric Normalization
Character Recognition
The Typographic Model
The Language Model
Experiments
Chars74K Dataset
ICDAR 2003 Dataset
Conclusions
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken