Computer Vision -- ACCV 2010 Workshops
ACCV 2010 International Workshops. Queenstown, New Zealand, November 8-9, 2010. Revised Selected Papers, Part I
Published on 15. September 2011
XXX, 464 pages
978-3-642-22822-3 (ISBN)
Description
The two-volume set LNCS 6468-6469 contains the carefully selected and reviewed papers presented at the eight workshops that were held in conjunction with the 10th Asian Conference on Computer Vision, in Queenstown, New Zealand, in November 2010.
From a total of 167 submissions to all workshops, 89 papers were selected for publication. The contributions are grouped together according to the main workshops topics, which were: computational photography and aesthetics; computer vision in vehicle technology: from Earth to Mars; electronic cultural heritage; subspace based methods; video event categorization, tagging and retrieval; visual surveillance; application of computer vision for mixed and augmented reality.
More details
Other editions
Additional editions
Reinhard Koch | Fay Huang
Computer Vision -- ACCV 2010 Workshops
ACCV 2010 International Workshops. Queenstown, New Zealand, November 8-9, 2010. Revised Selected Papers, Part I
Book
09/2011
Springer
€53.49
Shipment within 7-9 days
Content
- Title
- Preface
- Table of Contents - Part I
- Workshop on Visual Surveillance
- Second-Order Polynomial Models for Background Subtraction
- Introduction
- Models and Solutions
- Modeling of Local Photometric Distortions
- Bayesian Polynomial Fitting for Background Subtraction
- Experimental Results
- Conclusions
- Adaptive Background Modeling for Paused Object Regions
- Introduction
- Framework
- Probabilistic Model and Predictive Model
- Probabilistic Model Base on GMM
- Predictive Model Based on Exponential Smoothing
- Foreground Detection Based on MRF
- Update of Model Parameters
- Experimental Results
- Evaluation of Implicit Model Update
- Accuracy of Paused Object Detection
- Evaluation of Robustness against Illumination Changes
- Conclusion
- Determining Spatial Motion Directly from Normal Flow Field: A Comprehensive Treatment
- Introduction
- The Apparent Flow Direction (AFD) Constraint
- Preliminaries
- The Special Case: Pure Translation
- The Special Case: Pure Rotation
- Solving the System of Linear Inequalities for the Two Special Cases
- The Case of General Motion
- The Apparent Flow Magnitude (AFM) Constraint
- Putting the Two Constraints Together
- Experimental Results on Benchmark Data
- Conclusion and Future Work
- Background Subtraction for PTZ Cameras Performing a Guard Tour and Application to Cameras with Very Low Frame Rate
- Introduction
- Related Work
- Background Subtraction by Keypoint Density Estimation
- Texture Descriptors for Background Subtraction
- Weighting the SURF Descriptor
- Evaluating the Quality of the Texture Descriptor
- Background Update
- Experimental Results
- Conclusion
- Bayesian Loop for Synergistic Change Detection and Tracking
- Introduction and Related Work
- Models and Assumptions
- Cognitive Feedback
- Reasoning Probabilistically on Change Maps
- Experimental Results
- Conclusions
- Real Time Motion Changes for New Event Detection and Recognition
- Introduction
- Activity Area
- Change Detection
- Statistical Data Distribution Modeling
- Recognition
- Experiments
- Conclusions
- Improving Detector of Viola and Jones through SVM
- Introduction
- Preliminaries
- Basic Cascading Classifier
- Adaboost Learning
- Improving Cascade through SVM
- Improving Cascading Structure
- Optimization of ``H" Stage
- Training Process of Improving Cascading Structure
- Experiment Results
- Conclusion
- Multi-camera People Localization and Height Estimation Using Multiple Birth-and-Death Dynamics
- Introduction
- Related Work
- Proposed Method
- Multi-plane Projection
- Feature Extraction
- Marked Point Process Model
- Optimization by Multiple Birth-and-Death Dynamics
- Experiments
- Conclusion
- Unsupervised Video Surveillance
- Introduction
- The Proposed Method for Behavior Analysis
- Batch Training Phase
- Behavior Models Update
- Application to a Video-Surveillance Scenario
- Training and Model Selection via Loose Annotation
- Experiments on Data Pruning
- Model Evolution
- Discussion
- Multicamera Video Summarization from Optimal Reconstruction
- Introduction
- Related Work
- Approach
- Scene Decomposition
- Feature Clustering and Region Linking
- Learning Occurrence Models
- Single Region Activity Reconstruction
- Key Patch Selection
- Extension to Multiple Regions
- Experiments
- Single Region
- Multiple Regions
- Conclusion
- Noisy Motion Vector Elimination by Bi-directional Vector-Based Zero Comparison
- Introduction
- Noisy Vector Elimination
- Zero Comparison
- Vector-Based Zero Comparison
- Bi-directional Vector-Based Zero Comparison
- Experiments
- Conclusion
- Spatio-Temporal Optimization for Foreground/Background Segmentation
- Introduction
- Contribution
- Spatio-temporal Optimization
- Fore- and Background Models
- 3d Reconstruction by Probabilistic Fusion
- Integrating Iterations
- Evaluation
- Conclusion
- Error Decreasing of Background Subtraction Process by Modeling the Foreground
- Introduction
- Related Works
- The Proposed Approach
- First Segmentation Strategy: Targeted TPR or TNR
- Second Segmentation Strategy: Best TNR/TPR Ratio
- Extension to More General Data Distribution
- Validation on Synthetic Sequences
- Results on Real Sequences
- Classification Improvement
- Targeting Rate
- Conclusion
- Object Flow: Learning Object Displacement
- Introduction
- Object Flow
- Problem Formulation and Learning
- Flow Estimation
- Experimental Results
- Object Flow for Pedestrians
- Object Flow for Different Objects
- Quantitative Comparison
- Conclusions
- HOG-Based Descriptors on Rotation Invariant Human Detection
- Introduction
- HOG: Histogram of Oriented Gradients
- Review of Rotation Invariant Features
- Various Features on Rotation Invariance
- Effect of Shape of Detection Window on Presence/Absence of Human Classification
- Rotation Invariant Classification of the Presence of Human
- Discussion and Conclusion
- Fast and Accurate Pedestrian Detection Using a Cascade of Multiple Features
- Introduction
- Related Work
- CoHOG Descriptors
- Joint Haar-Like Features with AdaBoost
- The Proposed Pedestrian Detection System
- CoHOG Descriptors with PCA
- Cascading Haar and CoHOG Descriptors
- Experiments
- Experimental Setup
- CoHOG with PCA
- AdaBoost with CoHOG
- Conclusions
- Frontal Face Generation from Multiple Low-Resolution Non-frontal Faces for Face Recognition
- Introduction
- Frontal Face Generation from Multiple Non-frontal Faces
- Face Patch Correspondence
- Patch-Wise Face Image Transformation
- Synthesis of Frontal Face
- Experiment
- Frontal Face Generation
- Face Recognition
- Result and Discussion
- Conclusion
- Probabilistic Index Histogram for Robust Object Tracking
- Introduction
- Related Work
- Index Histogram
- Palette Indexing
- Probabilistic Indexing Histogram
- Spatial Distance
- Cross Bin-Ratio Dissimilarity
- Bayesian State Inference for Object Tracking
- Experiments
- Conclusions
- Mobile Surveillance by 3D-Outlier Analysis
- Introduction
- Related Work
- Algorithm Overview
- Information Gathering by S+M
- Motion Clustering
- Maintaining the Object Centered Representation
- Online Rigid Object Representation
- Initialization of the Object Centered Representation
- Update
- Re-mapping of Re-appeared Point Features
- Experimental Results
- Conclusions
- Person Re-identification Based on Global Color Context
- Introduction
- Related Work
- Global Color Context
- Color Descriptors
- Color Word Assignment
- Global Color Context
- Experimental Results and Analysis
- Conclusions
- Visual Object Tracking via One-Class SVM
- Introduction
- Object Tracking Based on OC-SVM
- Tracking Sample Set Construction
- Examples Selection for OC-SVM
- Object Tracking Based on OC-SVM
- Experiments and Results
- Conclusions and Future Works
- Attenuated Sequential Importance Resampling (A-SIR) Algorithm for Object Tracking
- Introduction
- Related Work
- Particle Filter
- Non-linear Resampling Algorithm
- Multi-Part Histogram (MPH) Based Measurement
- Object Feature Descriptor
- Color Measurement Model
- Experiments and Results
- The Error Metric the Performance Evaluation
- Conclusion
- Vehicle Class Recognition Using Multiple Video Cameras
- Introduction
- Vehicle Recognition Literature
- System Overview
- Structure from Motion Using Cross Ratio Invariance
- Epipolar Geometry for Our Problem
- Cross Ratio Invariance
- Histogram Method for Cross Ratio Invariance
- N frames and Scale Factor
- Recognition Using 3D Information: Reconstruction Error
- Bayesian Recognition
- Recognition Experiments
- Recognition Using Pairwise Video Clips
- Conclusion
- Efficient Head Tracking Using an Integral Histogram Constructing Based on Sparse Matrix Technology
- Introduction
- Efficient Orientation Histogram Extraction
- Proposed Head Tracking
- Head Detecting with a Circular Shift Orientation Histogram Matching
- Head Tracking Using the Orientation Histogram Matching-Based Proposal
- Experimental Results
- Evaluation of the Proposed Head Tracking
- Running Time Analysis
- Conclusions
- Workshop on Video Event Categorization, Taggingand Retrieval (VECTaR)
- Analyzing Diving: A Dataset for Judging Action Quality
- Introduction
- Related Work
- The FINA09 Diving Dataset
- How Diving Is Scored
- Background Subtraction and Tracking
- Robust Registration
- Background Subtraction
- Foreground Object Tracking
- Representation
- Describing Pose with Gradient Orientation Histograms
- Leveraging Background Subtraction
- Classifying Time Series
- Bag of Poses
- Dynamic Time Alignment Kernel (DTAK)
- Classification Experiments and Results
- Conclusion
- Appearance-Based Smile Intensity Estimation by Cascaded Support Vector Machines
- Introduction
- Smile Detection and Intensity Estimation
- Feature Extraction
- Detection and Intensity Estimation
- Data Preparation
- Experiments
- Performance Evaluation by LIH
- Performance Evaluation by CS-LBP
- Performance Evaluation by LIH+CS-LBP
- Performance of Cascaded SVM Smile Detector
- Smile Intensity Estimation
- Demonstration System
- Conclusion
- Detecting Frequent Patterns in Video Using Partly Locality Sensitive Hashing
- Introduction
- Overview of the Proposed Method
- Partly Locality Sensitive Hashing
- Combining Locality Sensitive and Insensitive Hash Functions
- Sparse Sampling Using PLSH
- Detection of Frequent Patterns
- Detection of Frequent Patterns Using Data Density
- Classification of Frequent Patterns
- Experimental Results
- Evaluation of Sampling Methods in Data Density Estimation
- Detection Rate of Frequent Patterns
- Computational Time vs. Amount of Data
- Conclusions
- Foot Contact Detection for Sprint Training
- Introduction
- Domain-Specific Constraints
- Identifying Foot Contacts
- Properties of a Contact Event
- Recognising the Toe-off Event
- Background Subtraction
- Static Foreground Accumulation
- Identifying Candidate Toe-off Events
- Toe-off Selection
- Evaluation and Results
- Video Capture System
- Results
- Related Work
- Conclusions
- Interpreting Dynamic Meanings by Integrating Gesture and Posture Recognition System
- Introduction
- Proposed System
- Pre-processing
- Feature Extraction and Classification
- Feature Extraction
- Classification
- Integration
- Particle Filter System
- Lexicon and Regular Language
- Interpretation and Inference
- Experimental Results
- Conclusion and Future Work
- Learning from Mistakes: Object Movement Classification by the Boosted Features
- Introduction
- Related Works
- Overview of the Proposed Method
- Rejection of the Stable Changes without Object Movements via Learning by Mis-detections
- Classification Framework
- Feature Set of Weak Classifiers
- Experiments
- Discussion
- Conclusion
- Modeling Multi-Object Activities in Phase Space
- Introduction
- Related Work and Contributions
- Modeling Multi-Object Activities
- Advantages
- Application to Activity Modeling
- Generalization
- MOPA Experimental Results
- Conclusions and Future Work
- Sparse Motion Segmentation Using Multiple Six-Point Consistencies
- Introduction
- Mathematical Background
- The 6-Point Matching Constraint
- Estimation of s
- Matching Score
- A Motion Segmentation Algorithm
- Experimental Results
- Conclusion
- Systematic Evaluation of Spatio-Temporal Features on Comparative Video Challenges
- Introduction
- Spatio-Temporal Features
- Experimental Setup
- Video Data-Set and Features
- Detector Evaluation
- Descriptor Evaluation
- Results
- Detector Evaluation
- Descriptor Evaluation
- Conclusion
- Two-Probabilistic Latent Semantic Model for Image Annotation and Retrieval
- Introduction
- Literature Reviews and Related Works
- Two-Probabilistic Latent Analysis Model
- Learning Parameters Using EM Algorithm
- Annotating an Unlabeled Image
- Querying by Words
- Performance Measurement
- Simulation Results
- Dataset and Simulation Condition
- Image Annotation: mAP Performance and Processing Time
- Text-Based Image Retrieval: mAP and Processing Time
- Conclusions
- Using Conditional Random Field for Crowd Behavior Analysis
- Introduction
- The Methodology
- Pre-processing
- Creating Block-Clips
- Mixture Model
- Conditional Random Field and Crowd Behavior Detection
- Experiments and Discussion
- Conclusion
- Workshop on Gaze Sensing and Interactions
- Understanding Interactions and Guiding Visual Surveillance by Tracking Attention
- Introduction
- Coarse Gaze Estimation
- Measuring Attention Using Coarse Gaze
- Using Gaze Estimates for Camera Control
- Understanding Human Interactions via Attention
- Conclusions
- Algorithm for Discriminating Aggregate Gaze Points: Comparison with Salient Regions-Of-Interest
- Introduction
- Background
- Classification Framework
- Extracting Similarity Scores
- Computing the Classification Threshold
- Estimating Classifier Performance via Cross-Validation
- Empirical Evaluation
- Results
- Discussion
- Conclusion
- Gaze Estimation Using Regression Analysis and AAMs Parameters Selected Based on Information Criterion
- Introduction
- Proposed Method
- Facial Area Search
- Active Appearance Models
- Regression Analysis and Model Selection Method
- Experiment
- Experimental Conditions
- Results
- Conclusion
- Estimating Human Body and Head Orientation Change to Detect Visual Attention Direction
- Introduction
- System Overview
- Estimating Body Orientation
- Head Orientation Change Estimation
- Experimental Results
- Conclusions
- Can Saliency Map Models Predict Human Egocentric Visual Attention?
- Introduction
- Related Work
- Procedure for Computing Saliency Maps for Videos
- Experiment
- Experimental Procedure
- Results
- Discussion
- Conclusion and Future Work
- An Empirical Framework to Control Human Attention by Robot
- Introduction
- Robot Behavior and Architecture
- Robot Behaviors
- Hardware Configuration
- Software Configuration
- Experiments
- Experiment 1: To Attracts Human Attention
- Experiment 2: Eye-Contact Experiment
- Experiment 3
- Discussion and Conclusion
- Improvement and Evaluation of Real-Time Tone Mapping for High Dynamic Range Images Using Gaze Information
- Introduction
- Reinhard's Global Tone Mapping Operator
- Gazing Area Based Tone Mapping Operator
- Proposed Method
- Experiments and Results
- Experimental Conditions
- Experiment 1: Checking Suitable Processing Area on Previous Operator
- Experiment 2: Checking Processing Area for Proposed Method
- Preference Comparison of Algorithm Performances
- Discussion
- Evaluation of the Impetuses of Scan Path in Real Scene Searching
- Introduction
- The Impetus of Scan Path Generation
- Database
- Saliency Map
- Task Guidance
- Oculomotor Bias
- Scan Path Generation Flow
- The Performances of the Sources on Scan Path Generation
- Single and Combined Sources
- Whole Scan Path
- Conclusions and Future Work
- Author Index
