Image and Graphics Technologies and Applications: 18th Chinese Conference, IGTA 2023, Beijing, China, August 17–19, 2023, Revised Selected Papers (Communications in Computer and Information Science)

Preface
Organization
Contents
Image Processing and Enhancement Techniques
Underwater Image Enhancement and Restoration Techniques: A Comprehensive Review, Challenges, and Future Trends
	1 Introduction
	2 Underwater Image Enhancement and Restoration
		2.1 Challenges of Underwater Imaging
		2.2 Underwater Image Enhancement
		2.3 Underwater Image Restoration
	3 Underwater Image Datasets and Quality Metrics
	4 Experiments and Analysis
	5 Opportunities and Future Trends
	6 Conclusion
	References
A Self-supervised Learning Reconstruction Algorithm with an Encoder-Decoder Architecture for Diffuse Optical Tomography
	1 Introduction
	2 Method
		2.1 Light Propagation Model
		2.2 Deep Convolutional Encoder-Decoder Architecture
	3 Results
		3.1 Dataset
		3.2 Reconstruction Results
	4 Conclusions
	References
TSR-Net: A Two-Step Reconstruction Approach for Cherenkov-Excited Luminescence Scanned Tomography
	1 Introduction
	2 Methods
		2.1 Forward Problem
		2.2 Inverse Problem
		2.3 Two-Step Reconstruction Algorithm
	3 Results
		3.1 Reconstruction Depth Test
		3.2 Spatial Resolution Test
		3.3 Generalization Ability Test
	4 Discussion and Conclusions
	References
A Method for Enhancing the Quality of Compressed Videos Based on 2D Convolution and Aggregating Spatio-Temporal Information
	1 Introduction
	2 Related Work
		2.1 Unet
		2.2 Attention Mechanisms
	3 The Proposed Approach
		3.1 Method Proposal
		3.2 Framework
		3.3 Conv-block Module
	4 Experiments
		4.1 Experimental Setup
		4.2 Dataset
		4.3 Comparison to State-of-the-Arts
		4.4 Analysis and Discussions
	5 Conclusion
	References
Multimedia-Based Informal Learning in Museum Using Augmented Reality
	1 Introduction
	2 Related Work
		2.1 Cognitive Theory of Multimedia Learning in AR
		2.2 AR in Museum Learning
	3 Method
		3.1 Environment Scenario
		3.2 Experiment Design
		3.3 Experiment Procedure
		3.4 Participant
	4 Result
		4.1 Task Load
		4.2 Cognitive Load
		4.3 Acquired Knowledge
		4.4 User Preference
	5 Discussion
	6 Conclusions and Future Work
	References
Machine Vision and 3D Reconstruction
MAIM-VO: A Robust Visual Odometry with Mixed MLP for Weak Textured Environment
	1 Introduction
	2 Related Work
	3 Method
		3.1 System Architecture
		3.2 Mixer-WMLP Architecture
	4 Experiments
		4.1 Feature Matching
		4.2 Localization on TUM RGB-D Dataset
	5 Conclusions
	References
Visual SLAM Algorithm Based on Target Detection and Direct Geometric Constraints in Dynamic Environments
	1 Introduction
	2 Related Works
		2.1 Indirect Geometric Constraint Method
		2.2 Direct Geometric Constraint Method
	3 Method
		3.1 System Overview
		3.2 Dynamic Feature Detection Based on Target Detection and Direct Geometric Constraints
		3.3 Multi-frame Dynamic Target Tracking Based on Target Detection
		3.4 Saliency Map Reconstruction and Supervision Strategy
	4 Experiments
		4.1 Dynamic Target Detection and Multi-Frame Tracking Experiments
		4.2 Dynamic Feature Rejection Ablation Experiment
		4.3 Improved SLAM Algorithm Performance Comparison Experiment
	5 Conclusion
	References
3D Shape Similarity Measurement Based on Scale Invariant Functional Maps
	1 Introduction
		1.1 Related Works
		1.2 Contribution
	2 Fundamentals
		2.1 Laplace-Beltrami Operator
		2.2 Scaling Invariant Wave Kernel Signature (SIWKS)
		2.3 Scaling Invariant Heat Kernel Signature (SIHKS)
	3 Feature and Similarty Measurement
		3.1 AvgSI
		3.2 Invariance of AvgSI
		3.3 Scale Invariant BCICP
		3.4 Calculate Functional Maps
		3.5 3D Models Similarity Measurement
	4 Experiments and Analysis
		4.1 Comparison with Other Descriptors
		4.2 Robutness of the AvgSI
		4.3 The Similarity Results Based on the AvgSI
	5 Conclusion
	References
Visualization Research on Industry and Spatial Distribution of Industrial Heritage in New China
	1 Introduction
	2 Data and Sources for the Study
	3 Comprehensive Analysis of the Spatial Distribution of Industrial Heritage in New China
		3.1 Kernel Density Analysis
		3.2 Industry Distribution Analysis
	4 Regional Analysis of the New China’s Industrial Heritage
		4.1 Proportional Distribution of Numbers
		4.2 Distribution of Industry Types
	5 Conclusions and Insights
	References
Image/Video Big Data Analysis and Understanding
SCGTS: Semantic Content Guiding Teacher-Student Network for Group Activity Recognition
	1 Introduction
	2 Related Works
		2.1 Group Activity Recognition
		2.2 Large-Scale Pre-training Language Model
		2.3 Knowledge Distillation
	3 Semantic Content Guiding Teacher-Student Network
		3.1 The Teacher Network: BERT
		3.2 The Student Network: Detector-Free
		3.3 Semantic Guiding Knowledge Distillation
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Comparison with the State-of-the-Art Methods
		4.4 Ablation Studies
		4.5 Visual Analysis
	5 Conclusion
	References
Infrared Small Target Detection Based on Prior Weighed Sparse Decomposition
	1 Introduction
	2 Related Work
		2.1 Methods Based on Human Visual System
		2.2 Methods Based on Deep Learning
		2.3 Methods Based on Sparse Decomposition
	3 Method
		3.1 Construction of Image Tensor
		3.2 Sparse Decomposition in Tensor Space
		3.3 Representation of Tensor Sparsity
		3.4 Construction of Prior Weight Map
		3.5 Optimization Problem Construction and Solution
	4 Experiment and Results
		4.1 Dataset and Experiment Setup
		4.2 Evaluation Metrics
		4.3 Visual Comparison with Baselines
		4.4 Quantitative Evaluation
	5 Conclusion
	References
CPML: Category Probability Mask Learning for Fine-Grained Visual Classification
	1 Introduction
	2 Related Work
		2.1 Soft Attention
		2.2 Consistent Learning
	3 Proposed Method
		3.1 Category Probability Mask
		3.2 Category Consistency Loss
	4 Experiments
		4.1 Experimental Setup
		4.2 Results
	5 Conclusion
	References
DPFMN: Dual-Path Feature Match Network for RGB-D and RGB-T Salient Object Detection
	1 Introduction
	2 Related Works
		2.1 RGB-D Salient Object Detection
		2.2 RGB-T Salient Object Detection
	3 Method
		3.1 Feature Extraction and Cross-modal Matching
		3.2 Feature Diversity Enhancement and Dimension Reduction
		3.3 Global-Local Feature Match and Fusion
		3.4 Saliency Map Reconstruction and Supervision Strategy
	4 Experiments
		4.1 Datasets and Parameter Settings
		4.2 Evaluation Metrics
		4.3 Experimental Comparison and Analysis of Results
		4.4 Ablation Analysis
	5 Conclusion
	References
A Optical Flow-Based Fight Behavior Detection Method for Campus Scene
	1 Introduction
	2 Related Work
	3 Our Approach
		3.1 ROI Detection Based on Optical Flow
		3.2 Action Recognition Based on Swin Transformer
	4 Experiments
		4.1 Datasets and Settings
		4.2 Results
	5 Conclusion
	References
Attention-Guided Neural Network for Face Mask Detection
	1 Introduction
	2 Related Work
		2.1 YOLO Algorithm
		2.2 Attention Mechanism
	3 Proposed Method
		3.1 Backbone of AGNN
		3.2 Loss Function
	4 Experiment and Analysis
		4.1 Dataset
		4.2 Performance Metrics
		4.3 Experimental Result
		4.4 Visual Result
	5 Conclusion
	References
Double-Stream Network for Clothes-Changing Person Re-identification Based on Clothes Related Feature Suppression and Attention Mechanism
	1 Introduction
	2 Proposed Method
		2.1 Framework of Network
		2.2 Attention Module
		2.3 Clothes Related Feature Suppression Module
		2.4 Loss Function
	3 Experiment
		3.1 Experimental details
		3.2 Datasets
		3.3 Comparison with State-of-the-Art Methods
		3.4 Ablation Studies
		3.5 Hyperparameter Studies
		3.6 Visualization studies
	4 Conclusion
	References
Semantic Guided Attention for Weakly Supervised Group Activity Recognition
	1 Introduction
	2 Related Work
		2.1 Group Activity Recognition
		2.2 Weakly Supervised Group Activity Recognition
	3 Approach
		3.1 Motion Feature Extraction Module
		3.2 Semantic Attention Module (SA)
		3.3 Semantic-Context Interaction Module (SI)
		3.4 Context Aggregation Module (CA)
	4 Experiments
		4.1 Datasets
		4.2 Implementation Details
		4.3 Comparison with the State-of-the-Art Methods
		4.4 Ablation Studies
		4.5 Qualitative Analysis
	5 Conclusion
	References
Computer Graphics
Research on Contrast Calculation Method for Color Image
	1 Introduction
	2 Related Work
	3 Method
		3.1 Chrominance Information Acquisition Based on HSV Color Space
		3.2 Calculate Luminance Information
		3.3 Statistical Calculation of Contrast Value
	4 Experiment and Analysis
		4.1 Comparative Analysis of Natural Environment Images
		4.2 Comparative Analysis of Composite Images
	5 Conclusion
	References
A Position-Based Dynamics Simulation of Liver Deformation with Ellipsoidal Particles
	1 Introduction
	2 Methods
		2.1 Geometric Modeling of the Liver Based on Computed Tomography Images
		2.2 Physical Deformation Modeling Based on Ellipsoidal Particles
	3 Experimental and Results
	4 Discussions and Conclusions
	References
A Perceptually Uniform Gloss Space for Translucent Materials
	1 Introduction
	2 Related Works
		2.1 Perception and Modeling of Glossiness
		2.2 Perception and Modeling of Translucency
		2.3 Correlating Perceived Glossiness and Perceived Translucency
	3 Perceptually Uniform Space of Glossiness of Opaque Materials
		3.1 Stimulus
		3.2 Experiment 1
		3.3 Perceptually Uniform Space of Glossiness
	4 Perceptually Uniform Space of Glossiness of Translucent Materials
		4.1 Stimulus
		4.2 Experiment 2
		4.3 Perceived Glossiness vs BRDF and BSSRDF
	5 Conclusion and Discussion
	References
X-ray Computed Tomography Reconstruction Algorithm for Refractive Index Gradient
	1 Introduction
	2 Model and Methodology
		2.1 Optical Streamline Model
		2.2 Vector Reconstruction Algorithm for Refractive Index Gradient
		2.3 DPC Imaging with Inclined Phase and Analyzer Gratings
	3 Experimental Validation
		3.1 Experimental Setup
		3.2 Results
		3.3 Discussions
	4 Conclusion and Outlook
	References
Visualization and Visual Analysis
Global Temperature Prediction Models Based on ARIMA and LSTM
	1 Introduction
	2 Related Work
	3 Preliminary Data Analysis
		3.1 Data Pre-processing
		3.2 Data Analysis
	4 Data Source and Methodology
		4.1 Data Source and Data-processing
		4.2 ARIMA
		4.3 LSTM
	5 ARIMA VS LSTM: An Experimental Study of Global Temperature Prediction
		5.1 Global Temperature Prediction Based on ARIMA
		5.2 Global Temperature Prediction Based on LSTM
		5.3 Comparison of Both Global Temperature Prediction Models
	6 Conclusion
	References
Virtual Reality and Human-Computer Interaction
Real-Time Image Stitching with Transformers for Complex Traffic Environment
	1 Introduction
	2 Related work
	3 Method
		3.1 Image Matching TRLWAM
		3.2 Knowledge Distillation Model
		3.3 Image Stitching Model
	4 Experiments
		4.1 Image Stitching Experiment
		4.2 Image Stitching Results
	5 Conclusions
	References
Research on 3D Visual Perception Quality Metric Based on the Principle of Light Field Image Display
	1 Introduction
	2 The Associated Research
	3 A New Metric-Viewpoint Perception Density
		3.1 Light Field Images and Display Devices
		3.2 Optical Principle of Optical Field Display
		3.3 Calculation of Viewpoint Perception Density Metric
	4 3D Visual Perception Quality Evaluation of Light Field Images
		4.1 Preparation of Experimental Data
		4.2 Experiment Methods
		4.3 Experimental Result Processing
		4.4 Metric Performance Analysis
		4.5 Application of Metric
	5 Summary
	References
Nonlinearity Affection Analysis of Spectral Information Reconstruction by Trichromatic Imaging System
	1 Introduction
	2 Mapping Function in Spectrum Information Reconstruction
	3 Nonlinearity Affection of System Response
		3.1 Error of Reconstructed Spectral Information
		3.2 Color Difference of Reconstructed Spectral Information
	4 Comparison Between Linear and Nonlinear Data in Spectrum Reconstruction
		4.1 Differences of Linear and sRGB Data in Reconstruction Matrix
		4.2 Differences of Reconstructed Spectrum Between Linear and sRGB Nonlinear Data
		4.3 Error and Color Difference of Linear Data and sRGB Nonlinear Data
		4.4 Simulation Analysis of Error and Color Difference for Nonlinearity
	5 Conclusions
	References
Applications of Image and Graphics
Adversarial Reinforcement Learning for Steering Cars from Virtual to Real World
	1 Introduction
	2 Method
		2.1 Reinforcement Learning to Drive
		2.2 Domain-Adversarial Learning Across Reality Gap
		2.3 Domain-Adversarial Reinforcement Learning
	3 Experiment
		3.1 Dateset
		3.2 Settings
		3.3 Result and Analysis
	4 Conclusion
	References
Efficient Multimodal-Contribution-Aware N-pair Network for Focal Liver Lesions
	1 Introduction
	2 Related Works
		2.1 MRI Applied to the Diagnosis of FLLs
		2.2 Multimodal Medical Imaging Diagnosis
		2.3 Metric Learning
	3 Method
		3.1 Overview
		3.2 Efficient Multimodal Contribution Aware Block
		3.3 Multi-class N-pair Loss
		3.4 Final Loss
	4 Experiments
		4.1 Data Pre-processing
		4.2 Comparative Experiments
		4.3 Further Discussions
	5 Conclusion
	References
Lung Nodule Classification Based on SE-ResNet152 and Stratified Sampling
	1 Introduction
	2 Method
		2.1 Overall Architecture
		2.2 Se-ResNet152
		2.3 Stratified Sampling
		2.4 Transfer Learning
	3 Experiments and Result
		3.1 Dataset
		3.2 Experimental Setting
		3.3 Experiment Metrics
		3.4 Experiment Result
	4 Conclusion
	References
Fusing CNN and Transformer for Diabetic Retinopathy Image Grading
	1 Introduction
	2 Method
		2.1 EfficientNet
		2.2 Swin Transformer
		2.3 Overall Architecture
		2.4 Data Preprocessing
	3 Experiment Preparation
		3.1 Dataset
		3.2 Implementation Details
		3.3 Evaluation Metrics
	4 Experiment
		4.1 Comparison of Different EfficientNet Models
		4.2 Comparison of Different Swin Transformer Models
		4.3 Fusion Model Precision Results
		4.4 Comparison Results with Other Models
	5 Conclusion
	References
Cross-modal Domain Adaptive Instance Segmentation in SAR Images via Instance-aware Adaptation
	1 Introduction
	2 Related Work
		2.1 Domain Adaptive Object Detection
		2.2 Instance Segmentation in SAR Images
	3 Method
		3.1 Problem Setting
		3.2 Sample-level Image Transfer
		3.3 Feature-level Domain Adaptation
		3.4 Cross-domain Instance Segmentation
	4 Experiments
		4.1 Datasets and Evaluation Metrics
		4.2 Implementation Details
		4.3 Performance Comparison for SAR Ship Instance Segmentation
		4.4 Ablation Study on Sample-level Transfer
	5 Conclusion
	References
Application of Computer Vision Technology in Collaborative Control of the “Zhurong” Mars Rover
	1 Introduction
	2 Collaborative Control of the Rover
	3 Computer Vision Technology
		3.1 Lander Localization
		3.2 Rover Localization
		3.3 3D Reconstruction and Terrain Classification
		3.4 Path Planning and Simulation Validation
	4 On-Orbit Implementation and Applications
		4.1 Lander Localization and Terrain Analysis of the Landing Area
		4.2 Rover Localization and Error Evaluation
		4.3 3D Reconstruction and Terrain Classification
		4.4 Path Planning and Simulation Verification
	5 Conclusions
	References
ME-GraphSAGE: Minority Class Feature Enhanced GraphSAGE for Automatic Labeling of Coronary Arteries
	1 Introduction
	2 Methods
		2.1 Graph Generation
		2.2 Feature Extraction
		2.3 Minority Classes Feature Enhanced GraphSAGE
	3 Experiments and Results
		3.1 Dataset, Evaluation Metrics and Parameters
		3.2 Comparison with Existing Methods
		3.3 Ablation Study
		3.4 Bifurcation Features
	4 Conclusion
	References
LR-SARNET: A Lightweight and Robust Network for Multi-scale and Multi-scene SAR Ship Detection
	1 Introduction
	2 Methods
		2.1 Lightweight Backbone-CGNet
		2.2 Enhanced Feature Fusion Network-ENECK
		2.3 Multi-scale Object Detection Head—EESPHEAD
	3 Experiments
		3.1 Datasets and Setting
		3.2 Evaluation Index
		3.3 Experimental Results and Analysis
	4 Conclusion
	References
A Coverless Image Steganography Method Based on Feature Matrix Mapping
	1 Introduction
	2 Feature Matrix Mapping Rules
		2.1 Feature Matrix
		2.2 Feature Matrix Mapping
	3 CIS Scheme Based on Feature Matrix
		3.1 Information Hiding Phase
		3.2 Information Extraction Phase
	4 Experimental Results and Analysis
		4.1 Analysis of Robustness
		4.2 Analysis of Hidden Capacity
		4.3 Analysis of Imperceptibility
		4.4 Analysis of Security
	5 Summary
	References
Design and Implementation of a Digital Star Map Simulation Module
	1 Introduction
	2 Principle and Method of Star Map Simulation
		2.1 Acquisition and Preprocessing of SAO Catalog
		2.2 Coordinate System Transformation and Star simulation
	3 Simulation Result Analysis
	4 Conclusion
	References
Author Index