Image and Graphics Technologies and Applications: 15th Chinese Conference, IGTA 2020, Beijing, China, September 19, 2020, Revised Selected Papers

Preface
Organization
Contents
Image Processing and Enhancement Techniques
Single Image Super-Resolution Based on Generative Adversarial Networks
Abstract
1 Introduction
2 Related Work
3 Proposed Method
3.1 Network Structure
3.2 Densely Connected Dilated Convolution
3.3 Channel Attention Mechanism
4 Experiment
4.1 Experiment Preparation
4.2 Experiment Results
4.3 Ablation Experiment
5 Conclusion
References
A Striping Removal Method Based on Spectral Correlation in MODIS Data
Abstract
1 Introduction
2 The Proposed Destriping Algorithm
2.1 Auto Piecewise Moment Matching Method Based on Genetic Algorithm(GA)
2.2 Fourier Low-Pass Filtering of Residual Noise of High Correlation Band
3 Experiment Results
4 Conclusions
References
Multi-modal 3-D Medical Image Fusion Based on Tensor Robust Principal Component Analysis
Abstract
1 Introduction
2 Related Works
3 Proposed Method
3.1 Procedures of Proposed Method
3.2 3-D Weighted Local Laplacian Energy Rule
4 Experiments
4.1 Experimental Setting
4.2 Comparison on the Synthetic Data
4.3 Comparison on the Real Data
5 Conclusion
Acknowledgements
References
Accurate Estimation of Motion Blur Kernel Based on Genetic Algorithms
Abstract
1 Introduction
2 Related Works
2.1 Traditional Methods
2.2 Machine Learning Methods
3 Motion Blur Kernel Estimation Based on Genetic Algorithms
3.1 Individual Representation
3.2 Initialization of the First Generation Population
3.3 Genetic Operators
3.4 Fitness Function
3.5 Parameter Setting
4 Experiment and Analysis
4.1 Sample Data of Blur Kernel Estimation
4.2 The Rough Estimation Process of Motion Blur Kernel
4.3 Blur Kernel Estimation Results
5 Conclusion
Acknowledgment
References
Biometric Identification Techniques
Graph Embedding Discriminant Analysis and Semi-Supervised Extension for Face Recognition
Abstract
1 Introduction
2 Related Works
2.1 Marginal Fisher Analysis (MFA)
2.2 Graph Discriminant Embedding (GDE)
3 Methodology
3.1 The Proposed Method
3.2 Semi-Supervised Extension (SSE)
3.3 Algorithm
4 Experiments
4.1 Experiments on ORL Dataset
4.2 Experiments on AR Dataset
4.3 Experiments on FERET Dataset
4.4 Experiments Under Noise Condition
4.5 Experiments Under Blur Condition
5 Conclusion
Acknowledgments
References
Fast and Accurate Face Alignment Algorithm Based on Deep Knowledge Distillation
Abstract
1 Introduction
2 Related Work
3 Methodology
3.1 Deep Knowledge Distillation Method
3.2 Model Structure of Deep Knowledge Distillation
3.3 Teacher Model and Student Model Analysis
3.4 Reasoning Complexity
4 Experimental Results and Analysis
4.1 Experimental Preparation
4.2 Evaluation Metrics
4.3 Experimental Details
4.4 Evaluation Results on 300W
4.5 Evaluation Results on WFLW
5 Conclusion
References
Machine Vision and 3D Reconstruction
3D Human Body Reconstruction from a Single Image
Abstract
1 Introduction
2 Technical Methods
2.1 Overall Network Structure
2.2 Generative Network
2.3 Adversarial Network
2.4 3D Joint Angle Limitation
2.5 Loss of the Network Model
3 Experimental Results
4 Conclusion
Funding.
References
Image/Video Big Data Analysis and Understanding
Abnormal Crowd Behavior Detection Based on Movement Trajectory
Abstract
1 Introduction
2 Multiple Object Tracking Based on Detection
3 Abnormal Behaviour Detection of the Crowd
3.1 Abnormal Direction
3.2 Abnormal Crossing
3.3 Trajectory Traceback
4 Experiments and Analysis Results
4.1 Experimental Setup
4.2 Experimental Analysis
5 Conclusion
Acknowledgment
References
Full Convolutional Color Constancy with Attention
Abstract
1 Introduction
2 Related Work
3 Attention Structure Design
3.1 Channel Attention Mechanism
3.2 Spatial Attention Mechanism
3.3 Mixed Attention Mechanism
3.4 Network Structure
4 Experiments
5 Conclusion
References
Simplifying Sketches with Conditional GAN
Abstract
1 Introduction
2 Related Work
2.1 Sketch Simplification
2.2 Conditional GAN
3 Architecture
3.1 Overview
3.2 CGAN
3.3 Loss Function
4 Experiment
4.1 Platform
4.2 Dataset
4.3 Training Details
5 Evaluation
5.1 Comparison with the State of the Art
5.2 User Study
6 Conclusion
References
Improved Method of Target Tracking Based on SiamRPN
Abstract
1 Introduction
2 Baseline
2.1 SiamRPN
2.2 Region Proposal Network
3 Proposed Method
3.1 Convolutional Block Attention Module
3.2 Squeeze-and-Excitation Module
4 Experiment and Analysis
4.1 Datasets and Evaluation Criteria
4.2 Implementation Details
4.3 Experiment 1: Comparison of CAM Visualization Results of the Feature Extraction Networks
4.4 Experiment 2: Comparison with the Baseline Method SiamRPN on OTB2015
4.5 Experiment 3: Comparison with Other Tracking Methods on OTB2015
4.6 Experiment 4: Comparison with Other Tracking Methods on VOT2016
5 Conclusion
Acknowledgments
References
An Improved Target Tracking Method Based on DIMP
Abstract
1 Introduction
2 Baseline
2.1 Discriminative Learning Loss
2.2 Model Predictor
3 Proposed Method
3.1 Information Flow Through the Network
3.2 Projected Shortcut
4 Experiment and Analysis
4.1 Experiment 1: Selection of Model Training Rounds for Epoch
4.2 Experiment 2: Comparison Between Our Method and DIMP
4.3 Experiment 3: Robustness Analysis of Our Methodology and Some Mainstream Tracking Methods
4.4 Experiment 4: The Applicability of Our Method for Other Algorithms
5 Conclusion
Acknowledgments
References
Infrared Small Target Recognition with Improved Particle Filtering Based on Feature Fusion
Abstract
1 Introduction
2 Algorithm Framework
2.1 Preprocessing
2.2 Gray Feature Extraction
2.3 Motion Feature Extraction
2.4 Improved Particle Filter Algorithm Optimized by Krill Herd
3 Simulation Results and Analysis
4 Conclusions
References
Target Recognition Framework and Learning Mode Based on Parallel Images
Abstract
1 Introduction
2 Establishment of Artificial Image Data Set
2.1 Artificial Image Generation Method
2.2 Data Set Structure
3 Target Recognition Framework Based on Parallel Images
4 Insulator Recognition Experiment Based on PITR Framework
4.1 Experiment
4.2 Structure of Sample Classification Network
4.3 Results and Discussion
5 Conclusion
References
Crowd Anomaly Scattering Detection Based on Information Entropy
Abstract
1 Introduction
2 Multi-target Detection and Tracking
3 Scattered Detection Method of Crowd Abnormality
3.1 Movement Speed Factor of the Crowd
3.2 The Information Entropy of the Crowd
3.3 Evaluation Method
4 Experiments and Analysis Results
4.1 Experiments Setup
4.2 Experimental Analysis Based on Three Scenarios
5 Conclusion
Acknowledgments
References
Computer Graphics
View Consistent 3D Face Reconstruction Using Siamese Encoder-Decoders
1 Introduction
2 Related Work
2.1 3D Morphable Model
2.2 Learning-Based Methods
2.3 High-Fidelity Face Reconstruction
3 The Proposed Method
3.1 Siamese Network for Consistent Texture and Normal Map
3.2 Network Architecture
3.3 Loss Functions
4 Experiments
4.1 Implementation Details
4.2 Analysis on the Siamese Network
4.3 Analysis on the Normal Refinement
4.4 Comparison with the State-of-the-art Methods
5 Conclusion
References
An Angle-Based Smoothing Method for Triangular and Tetrahedral Meshes
Abstract
1 Introduction
1.1 Laplacian Smoothing
1.2 Variations of Laplacian Smoothing
1.3 Optimization-Based Smoothing
1.4 Combinations of Laplacian Smoothing and Optimization-Based Smoothing
1.5 Our Contributions
2 Related Work
2.1 Laplacian Smoothing
2.2 Smart Laplacian Smoothing
3 Another Way to Understand Laplacian Smoothing
4 Angle-Based Smoothing Method
4.1 Angle-Based Smoothing for Triangular Meshes
4.2 Angle-Based Smoothing for Tetrahedral Meshes
5 Examples and Analysis
5.1 Examples
5.2 Analysis
6 Conclusions and Future Work
References
Virtual Reality and Human-Computer Interaction
Rendering Method for Light-Field Near-Eye Displays Based on Micro-structures with Arbitrary Distribution
Abstract
1 Introduction
2 Method
2.1 Light Field Rendering
2.2 Light Field Rendering System Based on Cameras with Arbitrary Distribution
3 Experiments and Results
3.1 The Light-Field Near-Eye Displays Simulation Experiment
3.2 Near-Eye Displays Using Random Holes and a Pinhole Array
4 Conclusion
Acknowledgments
References
AUIF: An Adaptive User Interface Framework for Multiple Devices
1 Introduction
2 Related Work
3 Design of AUIF
3.1 Multi-device Collaboration Patterns
3.2 UI Components Allocation Algorithm
3.3 DUI Generation Framework
4 Prototype System
5 User Study
5.1 Design of the Experiment
5.2 Result
6 Conclusions
References
Applications of Image and Graphics
Control and on-Board Calibration Method for in-Situ Detection Using the Visible and Near-Infrared Imaging Spectrometer on the Yutu-2 Rover
Abstract
1 Introduction
2 Control Method and Engineering Applications
2.1 Control Method
2.2 Engineering Applications
3 Error Analysis and on-Board Calibration
3.1 Error Analysis
3.2 On-Board Calibration
4 Conclusion
Acknowledgments
References
Deep Attention Network for Remote Sensing Scene Classification
Abstract
1 Introduction
2 The Proposed Method
2.1 Overview
2.2 Channel Attention
2.3 Spatial Attention
2.4 Branch Fusion
3 Experimental Results
4 Conclusion
References
Thin Cloud Removal Using Cirrus Spectral Property for Remote Sensing Images
Abstract
1 Introduction
2 Methodology
2.1 Cirrus Spectral Property
2.2 Removal of Thin Cloud Effect
3 Experimental Results and Analysis
4 Conclusion
References
A Multi-line Image Difference Technique to Background Suppression Based on Geometric Jitter Correction
Abstract
1 Introduction
2 A Multi-line Image Difference Technique to Background Suppression Based on Geometric Jitter Correction
2.1 Geometric Jitter Correction Based on Adaptive Correlation
2.1.1 Geometric Correction Preprocessing
2.1.2 Geometric Jitter Correction
2.2 A Multi-line Image Difference Technique to Background Suppression Based on Geometric Jitter Correction
3 Simulation and Experiment
3.1 Experiment on the Effectiveness of Geometric Registration Strategy
3.2 Image Jitter Correction Effect Verification Experiment
3.3 Resampling Algorithm Simulation
3.4 Differential Detection Effect Verification Experiment
4 Conclusion
References
Other Research Works and Surveys Related to the Applications of Image and Graphics Technology
Image Recognition Method of Defective Button Battery Base on Improved MobileNetV1
Abstract
1 Introduction
2 Improved MobileNetV1
2.1 Basic Network
2.2 Double-Layer Asymmetric Convolution
2.3 Activation Function Replacement
3 Experiment Procedure
3.1 Button Battery Data Set
3.1.1 Data Collection and Data Procession
3.1.2 Tfrecord Data Format
3.2 Experimental Comparison and Analysis
3.2.1 Experimental Comparison of Different Activation Functions
3.2.2 Improve MobileNetV1 Performance
4 Conclusion
References
Author Index