دانلود کتاب Pattern Recognition and Computer Vision: 4th Chinese Conference, PRCV 2021, Beijing, China, October 29 – November 1, 2021, Proceedings, Part III ... Vision, Pattern Recognition, and Graphics)

فهرست مطالب :

Preface
Organization
Contents – Part III
Low-Level Vision and Image Processing
SaliencyBERT: Recurrent Attention Network for Target-Oriented Multimodal Sentiment Classification
1 Introduction
2 Related Work
3 Proposed Model
3.1 Task Definition
3.2 Recurrent Attention Network
4 Experiments
4.1 Experiment Settings
4.2 Results and Analysis
5 Conclusion
References
Latency-Constrained Spatial-Temporal Aggregated Architecture Search for Video Deraining
1 Introduction
2 The Proposed Method
2.1 Spatial-Temporal Aggregated Architecture
2.2 Architecture Search
3 Experimental Results
3.1 Experiment Preparation
3.2 Running Time Evaluation
3.3 Quantitative Comparison
3.4 Qualitative Comparison
3.5 Ablation Study
4 Conclusions
References
Semantic-Driven Context Aggregation Network for Underwater Image Enhancement
1 Introduction
2 Method
2.1 The Overall Architecture
2.2 Semantic Feature Extractor
2.3 Multi-scale Feature Transformation Module
2.4 Context Aggregation Enhancement Network and Loss Function
3 Experiments
3.1 Experimental Setup
3.2 Comparison with the State-of-the-Arts
3.3 Ablation Study
3.4 Application on Salient Object Detection
4 Conclusion
References
A Multi-resolution Medical Image Fusion Network with Iterative Back-Projection
1 Introduction
2 Proposed Approach
2.1 Overall Framework
2.2 Network Architecture
2.3 Loss Function
3 Experiments
3.1 Dataset and Training Details
3.2 Results and Analysis of IBPNet
4 Conclusions
References
Multi-level Discriminator and Wavelet Loss for Image Inpainting with Large Missing Area
1 Introduction
2 Related Work
2.1 Image Inpainting
2.2 Adversarial Training
3 Our Approach
3.1 Multi-level Discriminator
3.2 Wavelet Loss
4 Experiments
4.1 Experimental Settings
4.2 Performance Evaluation
4.3 Ablation Study
5 Conclusion
References
3D2Unet: 3D Deformable Unet for Low-Light Video Enhancement
1 Introduction
2 Related Work
2.1 Low-Light Image Enhancement
2.2 Low-Light Video Enhancement
3 Method
3.1 Problem Formulation
3.2 Overview of the Pipeline
3.3 RSTAB Module
3.4 Unet Architecture with Global Projection
4 Experiment
4.1 Experimental Setting
4.2 Comparison with State-of-the-art Methods
4.3 Ablation Study
5 Conclusion
References
Single Image Specular Highlight Removal on Natural Scenes
1 Introduction
2 Related Works
3 Proposed Method
3.1 Scene Illumination Evaluation
3.2 Smooth Feature Extraction
3.3 Coefficient Estimation and Highlight Removal
4 Experiments
4.1 Quantitative Comparison on Laboratory Images
4.2 Visual Effect Comparison on Natural Scene Images
4.3 Discussion of Important Parameters
5 Conclusion
References
Document Image Binarization Using Visibility Detection and Point Cloud Segmentation
1 Introduction
2 TPO(Target-Point Occlusion)
2.1 Point Cloud Transformation
2.2 Convex Hull
3 Algorithm
3.1 Binarization
3.2 Unshadow Binarization
4 Experiment
5 Conclusion
References
LF-MAGNet: Learning Mutual Attention Guidance of Sub-Aperture Images for Light Field Image Super-Resolution
1 Introduction
2 Related Work
2.1 Light Field Image Super-Resolution
2.2 Visual Attention Mechanism
3 Proposed Method
3.1 Shallow Feature Extraction
3.2 Mutual Attention Guidance
3.3 LF Image Reconstruction
4 Experiment
4.1 Dataset and Implementation Details
4.2 Ablation Studies
4.3 Comparisons with the State-of-The-Arts
5 Conclusion
References
Infrared Small Target Detection Based on Weighted Variation Coefficient Local Contrast Measure
1 Introduction
2 The Proposed Algorithm
2.1 Variation Coefficient Local Contrast Measure
2.2 Weighted Variation Coefficient Local Contrast Measure
2.3 Target Detection
3 Experimental Results
3.1 Enhancement Performance Comparison
3.2 Detection Performance Comparison
4 Conclusion
References
Scale-Aware Distillation Network for Lightweight Image Super-Resolution
1 Introduction
2 Related Work
3 Proposed Method
3.1 Network Architecture
3.2 Scale-Aware Distillation Block
3.3 Comparisons with Other Information Distillation Methods
4 Experiments
4.1 Implementation Details
4.2 Ablation Study
4.3 Comparisons with the State-of-the-Arts
5 Conclusion
References
Deep Multi-Illumination Fusion for Low-Light Image Enhancement
1 Introduction
2 Deep Multi-Illumination Fusion
2.1 Network Structure
2.2 Loss Function
3 Experimental Results
3.1 Implementation Details
3.2 Performance Evaluation
3.3 Ablation Analysis
3.4 Object Instance Segmentation
4 Conclusion
References
Relational Attention with Textual Enhanced Transformer for Image Captioning
1 Introduction
2 Related Work
2.1 Relationship Exploration
2.2 Transformer Architecture
3 The Proposed Approach
3.1 Relation Module
3.2 Attention Module
3.3 Decoder Module
3.4 Training and Objectives
4 Experiments
4.1 Datasets and Evaluation Metrics
4.2 Implementation Details
4.3 Ablation Study
4.4 Comparison with State-of-the-Art
5 Conclusion
References
Non-local Network Routing for Perceptual Image Super-Resolution
1 Introduction
2 Related Work
3 Methodology
3.1 Non-local Network Routing
3.2 Model Learning
4 Experiments
4.1 Evaluation Dataset and Metric
4.2 Implementation Details
4.3 Derived Architecture
4.4 Comparison with State-of-the-Art Methods
5 Conclusion
References
Multi-focus Image Fusion with Cooperative Image Multiscale Decomposition
1 Introduction
2 Cooperative Image Multiscale Decomposition Based MGF
2.1 Mutually Guided Filter
2.2 Cooperative Image Multiscale Decomposition
3 Image Fusion with CIMD
3.1 Base Layers Fusion
3.2 Detailed Layers Fusion
3.3 Reconstruction
4 Experiment
4.1 Experiment Setup
4.2 Comparison to Classical Fusion Method
5 Conclusions
References
An Enhanced Multi-frequency Learned Image Compression Method
1 Introduction
2 Related Works
3 Proposed Method
3.1 Formulation of Multi-frequency Learned Compression Models
3.2 Channel Attention Scheme
3.3 Decoder-Side Enhancement
4 Experiment Results
4.1 Parameter Description
4.2 Results Evaluation
5 Conclusion
References
Noise Map Guided Inpainting Network for Low-Light Image Enhancement
1 Introduction
2 Related Works
2.1 Low-Light Image Enhancement
2.2 Image Inpainting
3 Method
3.1 Stage I: Decomposition
3.2 Stage II: Restoration
4 Experiment
4.1 Implementation Details
4.2 Results and Analysis
4.3 Ablation Study
5 Conclusion
References
FIE-GAN: Illumination Enhancement Network for Face Recognition
1 Introduction
2 Related Works
3 Methodology
3.1 Architecture
3.2 Loss Function
3.3 Deployment in Face Recognition
4 Experiments
4.1 Experimental Settings
4.2 Visual Perception Results
4.3 Face Recognition Results
4.4 Ablation Study
5 Conclusion
References
Illumination-Aware Image Quality Assessment for Enhanced Low-Light Image
1 Introduction
2 Illumination-Aware Quality Assessment of Enhanced Low-Light Image
2.1 Intrinsic Decomposition Module
2.2 CNN-based Feature Extraction Module
2.3 Learnable Perceptual Distance Measurement
3 Experiments
3.1 Basic Evaluations
3.2 Evaluations of LIE-IQA with Related Methods
3.3 LIE-IQA for Low-Light Enhancement
4 Conclusion
References
Smooth Coupled Tucker Decomposition for Hyperspectral Image Super-Resolution
1 Introduction
2 Tensor Notations and Preliminaries
3 Problem Formulation
3.1 Problem Description and Degradation Model
3.2 MAP Formulation
3.3 Smooth Coupled Tucker Decomposition Model
3.4 Optimization
4 Experimental Results
4.1 Experimental Settings and Implementation Issues
4.2 Experimental Results
4.3 Choice of Model Order
5 Conclusion
1. References
Self-Supervised Video Super-Resolution by Spatial Constraint and Temporal Fusion
1 Introduction
2 Related Work
2.1 SISR
2.2 VSR
3 Methodology
3.1 Overview
3.2 Internal-Data Based VSR
3.3 Spatio-Temporal VSR
4 Experiments
4.1 Protocols
4.2 Real-World Testing
4.3 Ablation Study
5 Conclusion
References
ODE-Inspired Image Denoiser: An End-to-End Dynamical Denoising Network
1 Introduction
2 Related Work
2.1 Image Denoising with CNN
2.2 Neural ODEs V.S. Residual Learning
3 Proposed Method
3.1 Network Architecture
3.2 Problem Formulation
3.3 OI-Block
4 Experiments
4.1 Ablation Study
4.2 Synthetic Noisy Images
4.3 Real Noisy Images
5 Conclusion
References
Image Outpainting with Depth Assistance
1 Introduction
2 Related Work
3 Our Model
3.1 Framework Design
3.2 Training
4 Experiment
4.1 Dataset
4.2 Quantitative Evaluation
4.3 Qualitative Evaluation
4.4 Comparison of Depth Feature Extraction Solutions
5 Conclusion
References
Light-Weight Multi-channel Aggregation Network for Image Super-Resolution
1 Introduction
2 Proposed Method
2.1 Network Architecture
2.2 MCA: Multi-Channel Aggregation Module
2.3 DA: Dilated Attention Module
3 Experiments
3.1 Experimental Settings
3.2 Comparisons with the State-Of-The-Arts
3.3 Ablation Study
4 Conclusion
References
Slow Video Detection Based on Spatial-Temporal Feature Representation
1 Introduction
2 Background and Related Work
3 Proposed Method
3.1 Temporal Feature Representation
3.2 Spatial Feature Representation
3.3 Detecting Slow Videos
4 Experiments
4.1 Dataset
4.2 Experimental Results
5 Conclusion
References
Biomedical Image Processing and Analysis
The NL-SC Net for Skin Lesion Segmentation
1 Introduction
2 Related Works
2.1 Traditional Computer Vision Methods
2.2 CNN-Based Methods
3 The Proposed Method
3.1 Network Architecture
3.2 Loss Function
4 Experimentation and Results
4.1 Experimentation Details
4.2 Ablation Analysis
4.3 Comparison with Other Methods
5 Conclusion
References
Two-Stage COVID-19 Lung Segmentation from CT Images by Integrating Rib Outlining and Contour Refinement
1 Introduction
2 Methodology
2.1 Rib Outline Detection
2.2 Coupling the Rib Outline and the Otsu Mask
2.3 Contour Refinement
3 Results
3.1 Dataset
3.2 Metrics
3.3 Comparison with Other Methods
3.4 Comparison with Deep-Learning Method
3.5 Running Time
4 Conclusion
References
Deep Semantic Edge for Cell Counting and Localization in Time-Lapse Microscopy Images
1 Introduction
2 Related Work
3 The Proposed Method
3.1 ROI Extracting and Cell Counting
3.2 Semantic Edge Detection
3.3 Ellipse Fitting and Verification
4 Experiment
4.1 Dataset
4.2 Results on ROI Extracting and Cell Number Predicting
4.3 Comparison of Edge Detectors and Ellipse Fitting Methods
4.4 Ablation Study
5 Conclusion
References
A Guided Attention 4D Convolutional Neural Network for Modeling Spatio-Temporal Patterns of Functional Brain Networks
1 Introduction
2 Methods
2.1 Data Acquisition, Preprocessing and Training Label Identification
2.2 Model Architecture of GA-4DCNN
2.3 Model Training Scheme, Evaluation, and Validation
3 Results
3.1 Spatio-Temporal Pattern Modeling of DMN via GA-4DCNN
3.2 Group-Averaged Spatio-Temporal Pattern of DMN via GA-4DCNN
3.3 Evaluation of Different Model Structures and Parameters
4 Conclusion
References
Tiny-FASNet: A Tiny Face Anti-spoofing Method Based on Tiny Module
1 Introduction
2 Related Work
3 Approach
3.1 Architecture
3.2 Central Difference Convolution
3.3 Simplified Streaming Module
3.4 Tiny Module and Tiny Bottlenecks
4 Experiments
4.1 Dataset
4.2 Metrics
4.3 Result Analysis
4.4 Ablation Experiment
5 Conclusion
References
Attention-Based Node-Edge Graph Convolutional Networks for Identification of Autism Spectrum Disorder Using Multi-Modal MRI Data
1 Introduction
2 Materials and Methods
2.1 Data Acquisition and Preprocessing
2.2 Construction of Node and Edge Feature Maps
2.3 Model Architecture of ANEGCN
2.4 Model Training
2.5 Model Generalizability and Interpretability
3 Results
3.1 Classification Accuracy Between ASD and TD
3.2 Superiority of Using Multi-modal Data for Classification
3.3 Node and Edge Features Contributing to ASD/TD Classification
4 Conclusion
References
Segmentation of Intracellular Structures in Fluorescence Microscopy Images by Fusing Low-Level Features
1 Introduction
2 Methodology
2.1 Feature Blocking
2.2 Low-Level Feature Fusion
2.3 Loss Function
3 Experiments
3.1 Experimental Setup
3.2 Performance Evaluation
3.3 Ablation Study
4 Conclusions
References
Interactive Attention Sampling Network for Clinical Skin Disease Image Classification
1 Introduction
2 Related Work
2.1 Skin-Disease Recognition
2.2 Visual Attention
3 Methodology
3.1 Class Activation Map Revisit
3.2 Local Peaks Search
3.3 Learning Interactive Attention
3.4 Attention-Based Sampling
3.5 Network Optimization
4 Experiments
4.1 Dataset
4.2 Experimental Settings
4.3 Ablation Study
4.4 Quantitative Results
4.5 Qualitative Results
5 Conclusion
References
Cross-modality Attention Method for Medical Image Enhancement
1 Introduction
2 Related Work
3 Method
3.1 The Architecture of CycleGAN
3.2 Cross-modality Attention Method (CMAM)
4 Experiments and Results
4.1 Evaluation Dataset and Implementation Details
4.2 Results and Analysis
5 Conclusion
References
Multi-modal Face Anti-spoofing Based on a Single Image
1 Introduction
2 Related Work
2.1 Face Anti-spoofing
2.2 Generative Adversarial Network
3 The Proposed Method
3.1 Modeling
3.2 Framework
3.3 Optimization
4 Experiments
4.1 Datasets and Settings
4.2 Implementation Details
4.3 Performance
4.4 Visualization
5 Conclusion
References
Non-significant Information Enhancement Based Attention Network for Face Anti-spoofing
1 Introduction
2 Related Work
3 Methodology
3.1 Non-significant Information Enhancement Based Attention Module
3.2 Multi-scale Refinement Strategy
3.3 Loss Function
4 Experiments
4.1 Databases and Metrics
4.2 Implementation Details
4.3 Ablation Study
4.4 Intra Testing
4.5 Inter Testing
4.6 Analysis and Visualization.
5 Conclusion
References
Early Diagnosis of Alzheimer\'s Disease Using 3D Residual Attention Network Based on Hippocampal Multi-indices Feature Fusion
1 Introduction
2 Methods
2.1 Multi-indices Feature Fusion Framework.
2.2 3D Residual Attention Network
2.3 Data Preprocessing and Feature Extraction
2.4 Implementation
3 Results
3.1 Demographic Characteristics and Neuropsychological Assessment of the Groups
3.2 The Performance of the Classification Model with Different Hippocampal Features
3.3 Comparison of Different Methods
4 Discussion
5 Conclusion
References
HPCReg-Net: Unsupervised U-Net Integrating Dilated Convolution and Residual Attention for Hippocampus Registration
1 Introduction
2 Method
2.1 The Proposed HPCReg-Net for Hippocampus Registration
2.2 Loss Functions
3 Experiments
3.1 Data and Pre-processing
3.2 Baseline Methods
3.3 Results
4 Discussion and Conclusion
References
Characterization Multimodal Connectivity of Brain Network by Hypergraph GAN for Alzheimer\'s Disease Analysis
1 Introduction
2 Method
2.1 Data and Pre-processing
2.2 Hypergraph and Optimal Hypergraph Homomorphism Algorithm
2.3 Generator and Interactive Hyperedge Neurons Module
2.4 Discriminator and Loss Function
3 Experiments
4 Conclution
References
Multimodal Representations Learning and Adversarial Hypergraph Fusion for Early Alzheimer\'s Disease Prediction
1 Introduction
2 Method
2.1 Distribution-Based GraphGAN
2.2 CNN-Based GraphAE
2.3 Adversarial Hypergraph Fusion
2.4 Training Strategy
3 Experiments
3.1 Data
3.2 Experimental Settings
3.3 Results
4 Conclusion
References
Model-Based Gait Recognition Using Graph Network with Pose Sequences
1 Introduction
2 Related Work
3 Skeleton-Based Gait Recognition
3.1 Notation and Graph Convolutional Networks (GCNs)
3.2 Network and Implementation Details
4 Experiments
4.1 Dataset and Training Details
4.2 Performance of Networks
4.3 Comparison with Other Model-Based Methods
4.4 Result on CASIA-B Dataset and Comparison with Appearance-Based Method
4.5 Spatio-Temporal Study
5 Conclusion
References
Multi-directional Attention Network for Segmentation of Pediatric Echocardiographic
1 Introduction
2 Methodology
2.1 Overview
2.2 Multi-direction Attention Module
2.3 Layered Feature Fusion Model
3 Experiments
3.1 Experimental Setup
3.2 Results and Analysis
4 Conclusion
References
Deep-Based Super-Angular Resolution for Diffusion Imaging
1 Introduction
2 Methods
2.1 Q-Space Sampling
2.2 Mapping Network of Architecture
3 Experiments
3.1 Dataset Description and Implementation Details
3.2 Effect of Sampling Scheme
3.3 Performance on HARDI Data
3.4 Performance on Fiber Reconstruction from HARDI Data
4 Conclusion
References
A Multiple Encoders Network for Stroke Lesion Segmentation
1 Introduction
2 Related Work
3 Proposed Method
4 Experiments
4.1 Experimental Setup
4.2 Mixing Loss Function
4.3 Overall Accuracy Evaluation
4.4 The Performance on Different Levels
4.5 Loss Validity
5 Conclusion
References
Nodule Synthesis and Selection for Augmenting Chest X-ray Nodule Detection
1 Introduction
2 Related Work
2.1 Nodule Detection
2.2 Image Synthesis
3 Method
3.1 Nodule Synthesis
3.2 Nodule Selection
4 Experimental Results
4.1 Dataset and Experimental Setup
4.2 Nodule Synthesis Assessment
4.3 Data Augmentation Efficacy Assessment
5 Conclusion and Discussion
References
Dual-Task Mutual Learning for Semi-supervised Medical Image Segmentation
1 Introduction
2 Related Work
2.1 Semi-supervised Medical Image Segmentation
2.2 Signed Distance Maps
3 Method
3.1 Dual-Task Networks
3.2 Mutual Learning for Semi-supervised Segmentation
4 Experiments
4.1 Dataset and Implementation Details
4.2 Ablation Analysis
4.3 Quantitative Evaluation and Comparison
5 Conclusion
References
DPACN: Dual Prior-Guided Astrous Convolutional Network for Adhesive Pulmonary Nodules Segmentation on CT Sequence
1 Introduction
2 Dual Prior Astrous Convolutional Network
2.1 Segmentation Net
2.2 Visual Prior Module
2.3 Location Prior Module
2.4 Concatenate Strategy
2.5 Loss Function
3 Experiment and Results
3.1 Dataset and Implementation
3.2 Qualitative Evaluation
3.3 Quantitative Comparison
3.4 Ablation Study
4 Conclusion
References
Face Anti-spoofing Based on Cooperative Pose Analysis
1 Introduction
2 Related Work
3 Methodology
3.1 System Design
3.2 STN for Automatic Face Alignment
3.3 Pose Aware Quadrilateral
3.4 PAQ Similarity Measurement
4 Experiment
4.1 PAQ Discriminability Test
4.2 CPA System Test
5 Conclusion
References
A Dark and Bright Channel Prior Guided Deep Network for Retinal Image Quality Assessment
1 Introduction
2 Proposed Method
3 Experimental Results
3.1 Datasets
3.2 Experimental Setting
3.3 Comparison with the State-of-the-arts
3.4 Ablation Study
4 Conclusion
References
Continual Representation Learning via Auto-Weighted Latent Embeddings on Person ReID
1 Introduction
2 Related Work
2.1 Continual Learning
2.2 Person Re-Identification
3 Methodology
3.1 Latent Embeddings with Autoencoders
3.2 Stability and Plasticity
3.3 Auto-Weighted Embedding
3.4 Matching Representation
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Results and Comparisons
5 Conclusions
References
Intracranial Hematoma Classification Based on the Pyramid Hierarchical Bilinear Pooling
1 Introduction
2 Methods
2.1 Hierarchical Bilinear Pooling
2.2 Pyramid Hierarchical Bilinear Model
3 Experiments
3.1 Dataset and Implementation Details
3.2 Ablation Experiments
3.3 Comparison with Other Network
3.4 Visualization of Training Process
4 Conclusions
References
Multi-branch Multi-task 3D-CNN for Alzheimer\'s Disease Detection
1 Introduction
2 Materials and Methods
2.1 Materials and Preprocessing
2.2 Method Overview
2.3 Multi-branch Multi-task Learning Network
2.4 Multi-branch Multi-task Learning Network with Auxiliary Losses
3 Experiments and Results
3.1 Experimental Setting
3.2 Comparison with the Other Methods
3.3 Ablation Study
4 Conclusion
References
Correction to: A Multiple Encoders Network for Stroke Lesion Segmentation
Correction to: Chapter “A Multiple Encoders Network for Stroke Lesion Segmentation” in: H. Ma et al. (Eds.): Pattern Recognition and Computer Vision, LNCS 13021, https://doi.org/10.1007/978-3-030-88010-1_44
Author Index