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 IV
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 IV

Ron Kimmel Reinhard Klette Akihiro Sugimoto(Herausgeber*in)

Springer (Verlag)

Erschienen am 28. Februar 2011

XXXVII, 722 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-642-19282-1 (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 3 of ACCV 2010
Fast Computation of a Visual Hull
Introduction
The Proposed Method
Computation of Intersections
1D Line Homography for 3D Ray Back-Projection
Experimental Results
Conclusion
References
Active Learning with the Furthest Nearest Neighbor Criterion for Facial Age Estimation
Introduction
An Overview of 2DLDA
Incremental 2DLDA and Active Learning
I2DLDA
Pool-Based Active Learning
Experimental Results
Conclusion
References
Real-Time Human Detection Using Relational Depth Similarity Features
Introduction
Local Features Based on Depth Information
Relational Depth Similarity Features
Varied Rectangular Region Sizes
Faster Depth Histogram Calculations by Integral Histograms
Human Detection Using Depth Information
Construction of Classifiers by Real Adaboost
Raster Scanning in 3D Space
Classification Adjusted for Occlusions
Mean-Shift Clustering in 3D Space
Evaluation of the Proposed Method by Classification Experiments
Database
Feature Evaluation Experiments
Evaluation Experiments Adjusted for Occlusions
Features Selected by Training
Real-Time Human Detection
Conclusions
References
Human Tracking by Multiple Kernel Boosting with Locality Affinity Constraints
Introduction
Multiple Kernel Boosting
Standard MKL
Multiple Kernel Boosting
Locality Affinity Constraints
Main Tracking Framework
Experiments
Experimental Settings
Results
Discussions
Conclusion
References
A Temporal Latent Topic Model for Facial Expression Recognition
Introduction
Feature Extraction and Indication
TLTM for Facial Expression Recognition
Latent Dirichlet Allocation
Temporal Latent Topic Model
Applying TLTM to Facial Expression Recognition
Experiments
Dataset and Parameter Settings
Experimental Results
Some Recognition Examples
Conclusions and Future Work
References
From Local Features to Global Shape Constraints: Heterogeneous Matching Scheme for Recognizing Objects under Serious Background Clutter
Introduction
Related Work
Proposed Method
Voting Scheme in Object Recognition
Image Patches
Shape Context
Reference Point
Final Scoring
Experiments
Comparison Datasets
Temple Dataset
Conclusion
References
3D Structure Refinement of Nonrigid Surfaces through Efficient Image Alignment
Introduction
Warp between the Template and the Input Image
Notations and Assumptions
2D-2D Warp Function
Appearance-Based Non-rigid 3D Recovery
Direct Lucas-Kanade Method for Non-rigid 3D Recovery
Efficient Image Alignment for Non-rigid 3D Recovery
Fusing Features and Appearance
Experimental Results
Synthetic Data
Real Images
Conclusion
References
Local Empirical Templates and Density Ratios for People Counting
Introduction
Related Works
Proposed Method
Offline Learning and Local Empirical Template Extraction
Online Counting
Experimental Results and Performance Evaluation
Evaluation on PETS2009 Benchmark
Evaluation on In-House Collection
Concluding Remarks and Future Works
References
Curved Reflection Symmetry Detection with Self-validation
Introduction
Our Approach
Symmetry Particles Discovering
Generating Pairwise Consistency Graph
Multiple Hypothesis Path Sampling and Evaluation
Experimental Results and Comparison
Our Results
Quantitative Evaluation and Comparison with P63leeSym
Conclusions
References
An HMM-SVM-Based Automatic Image Annotation Approach
Introduction
Related Work
Proposed Approach
Image Sub-blocking and Feature Extraction
Hidden Markov Model for Low-Level Annotation
High-Level Concept Classifier
Summary of the Training Procedure
Experiments
Dataset
Comparison with Other Methods
Effectiveness of the Proposed Approach
Conclusion
References
Video Deblurring and Super-Resolution Technique for Multiple Moving Objects
Introduction
Related Work
Algorithm Overview
Motion Deblurring for Multiple Moving Objects
Segmentation of Blurry Image Sequence
Blur Kernel Estimation Using Optical Flow
Motion Deblurring for Each Segment
Super-Resolution Technique for Multiple Depth
Estimating Candidate Planes Based on Feature Point Tracking
Pixel-Based Segmentation by Minimization of Re-projection Errors
Adaptive SR by MAP Estimation
Experiments
Evaluation of the Method Using Real-Data
Handheld Video Data Scene
Multiple Moving Objects Captured by Static Camera
Conclusion
References
Sparse Source Separation of Non-instantaneous Spatially Varying Single Path Mixtures
Introduction
Problem Definition
Mixture Alignment
1D/N-D Signals
Separation Process
Assumptions
Mixing System Estimation
Signal Separation
Generalizing to the mn Scenario
Results
Conclusions
References
Improving Gaussian Process Classification with Outlier Detection, with Applications in Image Classification
Introduction
Multi-class Gaussian Process Classification
Outlier Detection in Gaussian Process Classification
A Well-Rounded Classifier
Experiments
Handwritten Digits Recognition
Localization of Mobile Robots
Conclusion
References
Robust Tracking Based on Pixel-Wise Spatial Pyramid and Biased Fusion
Introduction
Overview of the Method
Our Online Learning Tracker
Sequential Inference Model
Pixel-Wise Spatial Pyramid
Update Scheme
Biased Multiplicative Formula
Implementation and Experiments
Quantitative Evaluation
Performance of the Individual Tracking Module
Qualitative Evaluation
Conclusion and Future Work
References
Compressive Evaluation in Human Motion Tracking
Introduction
Human Body Template
Compressive Sensing
Signal Sparse Representation
L1 Minimisation Recovery
Compressive Annealed Particle Filtering
Restricted Isometry Property and Pairwise Distance Preservation
Multilevel Wavelet Likelihood Evaluation on Compressive Measurements
Experiments
Conclusion and Future Work
References
Reconstructing Mass-Conserved Water Surfaces Using Shape from Shading and Optical Flow
Introduction
Related Work
Surface Geometry Reconstruction
Fluid Tracking
Reconstructing 3D Mass-Conserved Water
Conservation of Mass
Recovering the Water Surface Using Shape from Shading
Combined Shape from Shading and Optical Flow
Experiment
Conclusion
References
Earth Mover's Morphing: Topology-Free Shape Morphing Using Cluster-Based EMD Flows
Introduction
Related Work
Earth Mover's Morphing
Construction of Floating-Bin Histogram
Acquisition of EMD Flow
Cluster-Based Morphing
Graph-Cut Denoising
Experiments
Simple Shapes
Real Shapes
Shapes with Large Deformation
Split Disks
Discussion
Conclusion
References
Object Detection Using Local Difference Patterns
Introduction
Probabilistic Background Model Considering Illumination Fluctuation in Localized Region
Design of Local Difference Pattern
Construction of Local Difference Pattern
Probabilistic Background Model Based on LDP
Object Detection Using LDP
Experimental Result
Preliminary Experiment for Adjusting Parameters
Object Detection Accuracy
Discussion
Conclusion
References
Randomised Manifold Forests for Principal Angle-Based Face Recognition
Introduction
Related Work
Combining Features and Kernels for KPA
Base Classifier Design
Combining Features and Kernels
Random Manifold Forests
Random Manifold Forests for Randomised Kernel Learning
Constrained Random Manifold Forests
Experimental Results
Performance of KPA and CMSM with Different Features and Kernels
Multiple Classifier System for KPA
Random Manifold Forests
Constrained Random Manifold Forests
Conclusions
References
Estimating Meteorological Visibility Using Cameras: A Probabilistic Model-Driven Approach
Introduction
Vision through the Atmosphere
The Model-Driven Approach
Contrast of a Distant Target
Probabilistic Modeling
Expectation of the Mean Contrast
Model Inversion and Error Estimation
Experimental Evaluation
Methodology
Location of Lambertian Surfaces
Results
Conclusion
References
Optimizing Visual Vocabularies Using Soft Assignment Entropies
Introduction
Ground Truth Correspondence Data
Modelling Mean Average Precision from Vocabulary Statistics
Optimizing the Vocabulary with Respect to Entropy
Entropy Model
Optimizing Entropy
Experimental Validation
Dataset and Evaluation
Vocabularies with Hard Assignment
Vocabularies with Soft Assignment
Effects of Margin
Image Retrieval
Conclusions
References
Totally-Corrective Multi-class Boosting
Introduction
Multi-class Boosting
AdaBoost.MO
SAMME
Totally Corrective Multi-class Boosting
MultiBoostTC1
MultiBoostTC2
Experiments
UCI Datasets
MNIST Handwritten Digit Dataset
Conclusion
References
Pyramid Center-Symmetric Local Binary/Trinary Patterns for Effective Pedestrian Detection
Introduction
Preliminaries
The LBP and LTP Features
The CS-LBP/LTP Patterns
The Pyramid CS-LBP/LTP Features and Pyramid Uniform CS-LBP/LTP Features
Pedestrian Detection Based on Pyramid CS-LBP/LTP Features
Experiments
Experiment Setup
Performance of the Pyramid CS-LBP/LTP Feature Based Detector
Detection Results with Features Combined with Pyramid CS-LBP and PHOG
Conclusions
References
Reducing Ambiguity in Object Recognition Using Relational Information
Introduction
Overall Structure of the Proposed Method
Object Recognition with ARG
ARG-Based Object Representation
Inexact Graph Matching
Graph-Based Approach in a Single Image
Graph-Based Approach in a Stereo Image
Experiments and Performance Evaluation
Image Database
Experiments and Analysis
Conclusion
References
Posing to the Camera: Automatic Viewpoint Selection for Human Actions
Introduction
Why Some Views Are Better than Others?
Evaluating Viewpoint Quality
Measures of Action Visibility
Combining Visibility Measures
Applications and Experiments
Automatic Camera Selection
Action Recognition
Conclusion
References
Orthogonality Based Stopping Condition for Iterative Image Deconvolution Methods
Introduction
The Necessity of Stopping Condition for Iterative Methods
Existing Stopping Conditions
Orthogonality Based Stopping Condition
Results
Comparative Results
Conclusion
References
Probabilistic 3D Object Recognition Based on Multiple Interpretations Generation
Introduction
Multiple Interpretations Generation
Probability Computation
Likelihood Computation
Unlikelihood Computation
Pose Verification and Refinement
Experimental Results
Conclusion
References
Planar Affine Rectification from Change of Scale
Introduction
The Method
The Choice of Parametrization
Extension to Multiple Independent Sets
Experiments
Toy Example
Text Rectification
Darts
Segmentation of Multiple Planes with Repeated Pattern
Textures
Scale Change from Local Patches
Sensitivity to Noise
Image to Image Homography
Conclusions
References
Sensor Measurements and Image Registration Fusion to Retrieve Variations of Satellite Attitude
Introduction
Motion Model
Data Fusion
Bayesian Formulation
Prior Model p( "026A30C a,b)
Star Tracker Term p(b "026A30C s)
Image Data Term p(I "026A30C )
Algorithm
Experimental Results
Conclusion
References
Image Segmentation Fusion Using General Ensemble Clustering Methods
Introduction
Framework for Segmentation Combination by General Ensemble Clustering Methods
Pre-processing of the Image Segmentation Ensemble
Ensemble Clustering Methods
Post-processing
Experiments
Datasets
Combination by Ensemble Clustering vs. Supervised Learning
Evaluation of Segmentations
Results
Conclusion
References
Real Time Myocardial Strain Analysis of Tagged MR Cines Using Element Space Non-rigid Registration
Introduction
Methods
Finite Element Model
Optimization
Pattern Detection
Cardiac Motion Tracking
CUDA Implementation
Experiments
Results
Discussion and Future Work
References
Extending AMCW Lidar Depth-of-Field Using a Coded Aperture
Introduction
Background Theory and System Design
AMCW Lidar
The Range-Imager
Image Formation
Methodology
The Coded Aperture
The Point Spread Function
Restoration Method
Results and Discussion
Conclusion
References
Surface Extraction from Iso-disparity Contours
Introduction
Background
Stereo Disparity
Iso-disparity Surfaces
Extracting Planes from Iso-disparity Contours
Formulation of Iso-disparity Contours
Determining Planes along Iso-disparity Contours
Determining Planes along Disparity Gradients
Determining Surface Orientation Boundaries
Special Cases and Exceptions
Implementation Overview
Extracting and Segmenting Iso-disparity Contours
Iso-disparity Contour Grouping
Experimental Results
Planar Surface Extraction
Surface Segmentation during Navigation
Ground Surface Extraction
Conclusion
References
Image De-fencing Revisited
Introduction
Related Work
Lattice Detection
Image Completion
Near Regular Texture Segmentation
Clustering for the Foreground Segmentation
Online Learning-Based Lattice Detection
Multi-view and Symmetry Augmented Inpainting
Multi-view Inpainting
Symmetry-Augmented Inpainting
Experimental Results
Lattice Detection
Comparison with Liu et al.
Multi-view Inpainting Result
Conclusion
References
Feature-Assisted Dense Spatio-temporal Reconstruction from Binocular Sequences
Introduction
Related Work
Method
The Spatio-temporal Reconstruction Problem
Feature-Assisted Framework
Spatio-temporal Consistency
Matching Feature Points Using Spatio-temporal Consistency
Feature-Assisted Dense Correspondences
Handling Tracking Failure and Reoccurrence of Lost Points
Implementation Details
Experimental Results
Simulated Cloth Fluttering in the Wind
Flag Sequence
Facial Skin Deformation
Conclusion
References
Improved Spatial Pyramid Matching for Image Classification
Introduction
Related Methods
Proposed Method: Multi-scale SPM
Feature Generation in Different Scales
Calculate Dictionary
Compile Pyramid Histogram
Kernel Computation and SVM Classification
Experimental Results
Test Dataset
Performance Metric
Conclusion and Future Scope
References
Dense Multi-frame Optic Flow for Non-rigid Objects Using Subspace Constraints
Introduction
Low-Rank Non-rigid Subspace Constraints
Linear Basis Shape Model
2D Low-Rank Trajectory Basis
Dense Optic Flow with Subspace Constraints
Horn-Schunk Approach with Subspace Constraints
Minimisation
Experimental Results
Comparison with State of Art Variational Optic Flow Method
Conclusions and Future Work
References
Fast Recovery of Weakly Textured Surfaces from Monocular Image Sequences
Introduction
Mathematical Formulation
Surface Model
Optimization Formulation
Optimization
Implementation Details
Bundle Adjustment
Results
Conclusion
References
Ghost-Free High Dynamic Range Imaging
Introduction
Previous Works
Proposed Algorithm
Reference Image Selection and Global Image Alignment
Joint Pdf Estimation
Ghost Region Estimation
Camera Response Function Estimation and Radiance Value Refinement
Generalized Weighted Filtering Method
Experimental Results
Conclusion
References
Pedestrian Recognition with a Learned Metric
Introduction
Metric Learning
Nearest Neighbor with Rejection
Experiments
Methodology
Image Representation
Results
Conclusions
References
A Color to Grayscale Conversion Considering Local and Global Contrast
Introduction
Review of Previous Methods
Fixed Mapping
Variable Mapping by Energy Minimization
Details of the Proposed Method
Energy Formulation
Fast Approximate Solution
Experimental Results
Conclusions
References
Affordance Mining: Forming Perception through Action
Introduction
Association Rule Mining
Robotic Platform and Training Data Collection
Action Space Clustering
System Overview
Mining Percept-Action Associations
Feature Configurations
Percept-Action Transaction Database
Mining P-A Association Rules
Affordance Based Representation
Learning Action-Type Specific P-A Mappings
Responding to Novel Data
Evaluation
Discussion
References
Spatiotemporal Contour Grouping Using Abstract Part Models
Introduction
Related Work
Overview of the Approach
Detecting Consistent Cycles
Path Initialization
Path Extension
Training the Classifiers
Temporal Coherence of Consistent Cycles
Retrieving Consistent Cycle Trajectories
Probability Density Estimation
Results
Conclusions
References
Efficient Multi-structure Robust Fitting with Incremental Top-k Lists Comparison
Introduction
Data Similarity by Comparing Top-k Lists
Top-k Lists from Residual Sorting Information
The Spearman Footrule Distance
Measuring Similarity between Data
Guided Sampling with Hypothesis Filtering
Guided Sampling
Incremental Top-k Lists Comparison
Simultaneous Hypothesis Filtering
Multi-structure Fitting
Experiments
Multiple 2D Line and Circle Fitting
Homography Estimation
Fundamental Matrix Estimation
Conclusions
References
Flexible Online Calibration for a Mobile Projector-Camera System
Introduction
Related Work
Problem Formulation
Geometry of the Projector-Camera System
Fiducial-Less Calibration
Nonlinear Least Squares Optimization
Sequential Least Squares Optimization
Adaptation to Discontinuous System Change
Imposing an Upper Bound on the Information Matrix
Adaptive Control of Window Size
Implementation Details
Experimental Results
Summary
References
3D Object Recognition Based on Canonical Angles between Shape Subspaces
Introduction
Calculation Procedure of Shape Subspace
Factorization of an Image Sequence
Generation Based on the Coordinates of Multiple Points
The Proposed Method
Matching Feature Points Using Orthogonal Projection Matrices
Similarity between Shape Subspaces
3D Object Recognition Based on the Proposed Similarity Measure
Experimental Results
Experiment I: Matching Feature Points Using Synthetic Data
Experiment II: Face Recognition
Conclusions
References
An Unsupervised Framework for Action Recognition Using Actemes
Introduction
Related Work
The Method
Experimental Results
Conclusion and Future Work
References
Segmentation of Brain Tumors in Multi-parametric MR Images via Robust Statistic Information Propagation
Introduction
Methods
Graph Theoretic Approach for Semi-supervised Segmentation
Robust Edge Weight Measurement
Label Initialization Using Reliable Statistical Information
Experimental Results
Evaluation on UCNIA Simulated MR Images
Evaluation on Real MR Images
Conclusion
References
Face Recognition with Decision Tree-Based Local Binary Patterns
Introduction
Background Information
Local Binary Patterns
Face Recognition Pipeline
Our Approach: Decision Tree Local Binary Patterns
Tree Learning Details
Related Work
Experiments
Effect of Tree Depth and Neighborhood Size
Results on FERET
Results on CAS-PEAL-R1
Discussion
Conclusions and Future Work
References
Occlusion Handling with l1-Regularized Sparse Reconstruction
Introduction
Proposed Algorithm
Sparse Reconstruction Based Appearance Model
Particle Filter Tracking with Occlusion Reasoning
Update the Template
Experiments
Track Multi-object
Conclusions
References
An Approximation Algorithm for Computing Minimum-Length Polygons in 3D Images
Introduction
Basics
Algorithms
The Algorithm for Computing an Approximate MLP
Algorithms for Finding Arcs with Minimal Number of Cubes
Correctness and Time Complexity
Experimental Results
Concluding Remarks
References
Classifier Acceleration by Imitation
Introduction
Related Works
Classifier Molding by Linear Regression Tree
Data Generation for Classifier Molding
Classification Based Character Segmentation
CCS Problem
CCS Solution
Experiments
Classifier Comparison
Classifier Molding
Classification Based Character Segmentation
Conclusion
References
Recognizing Continuous Grammatical Marker Facial Gestures in Sign Language Video
Introduction
Recognizing Continuous Facial Gestures in Sign Language
Challenges
Layered Conditional Random Field Model
Experiments and Results
Conclusion
References
Invariant Feature Set Generation with the Linear Manifold Self-organizing Map
Introduction
Local Linear Manifold Learning for Invariant Feature Representation at Each Neuron
The Gaussian Likelihood Function
Stochastic Optimization
The Local Linear Manifold Learning Rules
Topographically Ordered Invariant Feature Set Generation under a Self-organizing Framework
Experiments
Data and Related Work
Invariant Feature Set Generation
Recognition Results
Conclusions and Perspectives
References
A Multi-level Supporting Scheme for Face Recognition under Partial Occlusions and Disguise
Introduction
Multi-level Supporting Scheme
Fisherface Subspace Classifiers
Matching Likelihood Ratio
Class Candidates Selection
Multi-level Supporting
Experimental Results
Synthetic Occlusions
Facial Expression Changes and Real Occlusions
Conclusions
References
Foreground and Shadow Segmentation Based on a Homography-Correspondence Pair
Introduction
Segmentation Based on Homography-Correspondence Pairs
Problem Setting
Asymmetric Treatment of Homography Constraint
Symmetric Approach Based Homography-Correspondence Pair
Problem Formulation
Implementation
Seed Generation
Data Term
Smoothness Term
Experiments
Data Set and Parameters
Benchmark
Results
Discussions
Conclusions
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken

Computer Vision - ACCV 2010

Beschreibung

Weitere Details

Weitere Ausgaben

Andere Ausgaben

Inhalt

Systemvoraussetzungen