Computer Analysis of Images and Patterns. 19th International Conference, CAIP 2021 Virtual Event, September 28–30, 2021 Proceedings

Preface
Organization
Contents – Part I
Contents – Part II
3D Vision
Simultaneous Bi-directional Structured Light Encoding for Practical Uncalibrated Profilometry
	1 Introduction
	2 Related Work
	3 Mathematical Investigation
		3.1 Background: Sinusoidal Phase Shifting Method
		3.2 Amplitude of Superposition
		3.3 Combined Patterns
		3.4 Mathematical Solution to the Problem
	4 Application to Real World
		4.1 Swapping Step
	5 Evaluation
	6 Conclusions
	References
Joint Global ICP for Improved Automatic Alignment of Full Turn Object Scans
	1 Introduction
	2 Related Work
	3 Background: Rigid Point Cloud Alignment
		3.1 Orthogonal Procrustes Problem
		3.2 Iterative Closest Point (ICP)
		3.3 Full Turn Registration: Pulli's Approach
	4 Joint Rigid Point Cloud Alignment
	5 Outlier Rejection
	6 Evaluation
		6.1 Stopping Criterion
	7 Conclusion
	References
Fast Projector-Driven Structured Light Matching in Sub-pixel Accuracy Using Bilinear Interpolation Assumption
	1 Introduction
	2 Related Work
	3 Fast Projector Driven Matching (FPDM)
		3.1 Matching Integer Pixel Quads
		3.2 Topological Consistency Check (TCC)
	4 Bilinear Sub-pixel Matching
		4.1 Sub-pixel Position in Unit Patch
		4.2 Mapping to Convex Quad
	5 Results
	6 Conclusions
	References
Pyramidal Layered Scene Inference with Image Outpainting for Monocular View Synthesis
	1 Introduction
	2 Proposed Method
		2.1 Outpainting
		2.2 Pyramidal Network Architecture
	3 Results
	4 Conclusions
	References
Out of the Box: Embodied Navigation in the Real World
	1 Introduction
	2 Related Work
	3 Real-World Navigation with Habitat
		3.1 Baseline Architecture
		3.2 Training in Simulation
		3.3 LoCoNav: Adapting for Real World
	4 Experiments
	5 Conclusion
	References
Toward a Novel LSB-based Collusion-Secure Fingerprinting Schema for 3D Video
	1 Introduction
	2 Related Work
		2.1 Overview on the Existing 3D Video Watermarking Techniques
		2.2 The Tracing Traitor: A Brief Review
	3 The General Tracing System
		3.1 The Copyright Registration Step
		3.2 The Collusion Attacks
		3.3 The Copyright Identification Step
	4 The Proposed Traitor Tracing Framework
		4.1 The Proposed Copyright Registration Step
		4.2 The Proposed Copyright Identification Scheme
	5 Experimental Results
		5.1 The Watermarking Results
		5.2 The Tracing Results
	6 Conclusions and Future Work
	References
A Combinatorial Coordinate System for the Vertices in the Octagonal C4C8(R) Grid
	1 Introduction
	2 Related Work
		2.1 The 2-Valued Labelling by Ashrafi and Loghman ch7AshrafiLsps2008
		2.2 The 4-Valued Coordinate System by Taghizadeh et al. ch7TaghizadehAGsps2012
		2.3 The 2-Valued Coordinate System by Taghizadeh et al. ch7TaghizadehAGsps2012
		2.4 The 3-Valued Coordinate System by Siddiqui et al. ch7SiddiquiNRIsps2016 and Naeem et al. ch7NaeemSGGsps2018
		2.5 The 3-Valued Coordinate System by Heydari and Taeri ch7HeydariTspsRsps2007
	3 The Combinatorial Coordinate System
		3.1 Definition
		3.2 Connection with the Cartesian Coordinates
		3.3 Neighbors
	4 Conversion to/from Existing Coordinate Systems
		4.1 The 2-Valued Labelling by Ashrafi and Loghman ch7AshrafiLsps2008
		4.2 The 4-Valued Coordinate System by Taghizadeh et al. ch7TaghizadehAGsps2012
		4.3 The 3-Valued Coordinate System by Siddiqui et al. ch7SiddiquiNRIsps2016 and Naeem et al. ch7NaeemSGGsps2018
		4.4 The 3-Valued Coordinate System by Heydari and Taeri ch7HeydariTspsRsps2007
	5 Discussion
	References
Bilingual Speech Recognition by Estimating Speaker Geometry from Video Data
	1 Introduction
	2 3D Speaker Geometry Estimation
	3 Methodology
		3.1 Object Detection
		3.2 Speech Recognition System
	4 Results
	5 Conclusions and Future Work
	References
Cost-Efficient Color Correction Approach on Uncontrolled Lighting Conditions
	1 Introduction
	2 Previous Work
	3 Methodology
		3.1 Data
		3.2 Color Correction
	4 Results
	5 Conclusions
	References
HPA-Net: Hierarchical and Parallel Aggregation Network for Context Learning in Stereo Matching
	1 Introduction
	2 Related Work
		2.1 Deep Neural Networks for Stereo Matching
		2.2 Multi-scale Information
	3 Proposed Method
		3.1 Network Architecture
		3.2 Hierarchical Aggregation (HA) Network
		3.3 Parallel Aggregation (PA) Network
		3.4 Output Module and Loss Function
	4 Experimental Results
		4.1 Datasets and Evaluation Metrics
		4.2 Implementation Details
		4.3 Ablation Studies
		4.4 KITTI Datasets Results
	5 Conclusion
	References
MTStereo 2.0: Accurate Stereo Depth Estimation via Max-Tree Matching
	1 Introduction
	2 MTStereo 2.0
		2.1 Steps Performed by MTStereo 2.0
	3 Experiments
		3.1 Evaluation
		3.2 Results and Comparison
	4 Conclusions
	References
Biomedical Image and Pattern Analysis
H-OCS: A Hybrid Optic Cup Segmentation of Retinal Images
	1 Introduction
	2 Proposed Method
		2.1 Region of Interest Extraction
		2.2 Network Architecture
		2.3 Loss Function
		2.4 Transfer Learning
		2.5 Postprocessing
	3 Experimental Analysis
		3.1 Datasets
		3.2 Implementation Details
		3.3 Effectiveness of TL and IA
		3.4 Effectiveness of Loss Functions
		3.5 Comparison with Other Approaches
	4 Conclusion
	References
Retinal Vessel Segmentation Using Blending-Based Conditional Generative Adversarial Networks
	1 Introduction
	2 Proposed Method
		2.1 Datasets
		2.2 Blending and Enhancement-Based Strategy
		2.3 GAN Synthesization
		2.4 CNN-Based Segmentation
	3 Experimental Evaluation
		3.1 Qualitative Result
		3.2 Quantitative Result
	4 Conclusion
	References
U-Shaped Densely Connected Convolutions for Left Ventricle Segmentation from CMR Images
	1 Introduction
	2 Related Works
	3 Proposed Method
		3.1 Dataset
		3.2 Preprocessing
		3.3 Architecture
		3.4 Post-processing
		3.5 Evaluation Metrics
	4 Experiments and Results
	5 Conclusion
	References
Deep Learning Approaches for Head and Operculum Segmentation in Zebrafish Microscopy Images
	1 Introduction
	2 Methodology
		2.1 Image Acquisition and Dataset Description
		2.2 Two Deep Learning Strategies
		2.3 U-Net Implementation
		2.4 Experimental Protocol
		2.5 Model Training with Different Loss Functions
	3 Results and Discussion
		3.1 Robustness to Image Acquisition with Another Microscope
	4 Conclusions
	References
Shape Analysis Approach Towards Assessment of Cleft Lip Repair Outcome
	1 Introduction
	2 Method
		2.1 Dataset and Tools
		2.2 Feature Description and Detection
		2.3 Symmetrical Axis Detection and Measurement
		2.4 Conversion of Similarity Measure to a Numeric Score
	3 Experimental Results
		3.1 Image Segmentation
		3.2 Evaluation of Aesthetic Assessment
	4 Conclusion
	References
MMEC: Multi-Modal Ensemble Classifier for Protein Secondary Structure Prediction
	1 Introduction
	2 Methodology
		2.1 Convolutional Neural Networks
		2.2 BERT
		2.3 Inception Recurrent Networks
		2.4 Multi-Modal Ensemble Classifier
		2.5 Genetic Algorithm
	3 Datasets and Evaluation Metric
		3.1 Datasets
		3.2 Evaluation Metric
	4 Experimental Evaluation
		4.1 CB6133
		4.2 CB513
	5 Conclusions and Future Work
	References
Patch-Level Nuclear Pleomorphism Scoring Using Convolutional Neural Networks
	1 Introduction
		1.1 Related Work
	2 Data and Materials
	3 Methods
		3.1 Model Training
	4 Experimental Set-Up
	5 Results
		5.1 Validation Results
		5.2 Test Results
	6 Discussion
	7 Conclusion
	References
Automatic Myelofibrosis Grading from Silver-Stained Images
	1 Introduction
	2 Related Work and Open Issues
	3 Materials and Methods
		3.1 Data Set
		3.2 Image Classification
	4 Experimental Evaluation
		4.1 Experimental Set-Up
		4.2 Results
	5 Conclusions
	References
A Deep Learning-Based Pipeline for Celiac Disease Diagnosis Using Histopathological Images
	1 Introduction
	2 Materials and Methods
		2.1 Data
		2.2 Methodology
	3 Results
	4 Discussion
	5 Conclusion
	References
HEp-2 Cell Image Recognition with Transferable Cross-Dataset Synthetic Samples
	1 Introduction
	2 Related Work
	3 Datasets
	4 Proposed Method
	5 Evaluation and Discussion
	6 Conclusion
	References
Clinically Guided Trainable Soft Attention for Early Detection of Oral Cancer
	1 Introduction
	2 Related Work
	3 Materials
	4 Method
		4.1 Attention Network
		4.2 Guided Attention
		4.3 Technical Details
	5 Results
	6 Discussion and Conclusion
	References
Small and Large Bile Ducts Intrahepatic Cholangiocarcinoma Classification: A Preliminary Feature-Based Study
	1 Introduction
	2 Proposed Classification Algorithm
		2.1 Tumor Segmentation
		2.2 Feature Extraction
		2.3 Feature Selection
	3 Performance Evaluation
		3.1 Patients Dataset
		3.2 Classification Procedure and Results
	4 Conclusions
	References
A Review on Breast Cancer Brain Metastasis: Automated MRI Image Analysis for the Prediction of Primary Cancer Using Radiomics
	1 Introduction
	2 Literature Review
		2.1 Image Preprocessing and Brain Metastasis Segmentation
		2.2 Prediction of Brain Metastasis: Breast Cancer Origin
	3 Discussion
	References
An Adaptive Semi-automated Integrated System for Multiple Sclerosis Lesion Segmentation in Longitudinal MRI Scans Based on a Convolutional Neural Network
	1 Introduction
	2 Materials and Methods
		2.1 Acquisition of Brain MRI Images
		2.2 Image Pre-processing
		2.3 CNN Model Architecture (U-Net)
		2.4 Image Post-processing
		2.5 Proposed CNN System
		2.6 Software Implementation
		2.7 Evaluation Metrics
	3 Experimental Results
		3.1 Manual and Semi-automated Lesions Segmentation on Brain MRI Images
		3.2 Evaluation of the Proposed System
	4 Discussion
	5 Concluding Remarks
	References
A Three-Dimensional Reconstruction Integrated System for Brain Multiple Sclerosis Lesions
	1 Introduction
	2 Material and Methods
		2.1 Acquisition of Brain MRI Images
		2.2 MRI MS Lesion Preprocessing and Semi-automated Segmentation
		2.3 Image Normalization and Contour Lesion Generation
		2.4 3D Lesion Volume Estimation
		2.5 3D Reconstruction
		2.6 Evaluation Metrics
	3 Results
	4 Discussion
	5 Conclusions and Future Trends
	References
Rule Extraction in the Assessment of Brain MRI Lesions in Multiple Sclerosis: Preliminary Findings
	1 Introduction
	2 Materials and Methods
		2.1 Study Group and MRI Acquisition
		2.2 Intensity Normalization and Manual Lesion Delineation
		2.3 Texture Feature Extraction
		2.4 Classification Analysis
		2.5 Rule Extraction
	3 Results
	4 Discussion
	5 Conclusion
	References
Invariant Moments, Textural and Deep Features for Diagnostic MR and CT Image Retrieval
	1 Introduction
	2 Related Works
	3 Our Image Retrieval System
		3.1 Feature Extraction
		3.2 Feature Selection
		3.3 Image Ranking
	4 Experimental Evaluation
		4.1 Data Sets
		4.2 Experimental Setup
	5 Results
	6 Conclusions
	References
Toward Multiwavelet Haar-Schauder Entropy for Biomedical Signal Reconstruction
	1 Introduction
	2 HSCH Multiwavelet Modeling and Associated Entropy
	3 Experimentations
		3.1 Biosignal Reconstruction
		3.2 Multiwavelet Entropy
	4 Conclusion
	References
Machine Learning
Handling Missing Observations with an RNN-based Prediction-Update Cycle
	1 Introduction and Related Work
	2 RNN-based Prediction-Update-Cycle
	3 Data Generation and Evaluation
	4 Conclusion
	References
eGAN: Unsupervised Approach to Class Imbalance Using Transfer Learning
	1 Introduction
	2 Related Work
	3 Methodology and Experimental Design
		3.1 Addressing Class Imbalance with eGAN
		3.2 Selection of Pre-trained Image Classification Weights
		3.3 Dataset
	4 Results and Discussion
		4.1 Imbalance Ratios
	5 Conclusion
	References
Progressive Contextual Excitation for Smart Farming Application
	1 Introduction
	2 Related Work
	3 Method
		3.1 Feature Extraction
		3.2 Progressive Contextual Excitation
	4 Experiments
		4.1 Dataset
		4.2 Implementation Details
		4.3 Comparison with Other Approaches
		4.4 Visualization
	5 Conclusion
	References
Fine-Grained Image Classification for Pollen Grain Microscope Images
	1 Introduction
	2 Related Works
	3 Method and Materials
		3.1 CutOcclusion: Training Data Augmentation
		3.2 Pipeline
		3.3 Test Time Augmentation.
	4 Dataset
	5 Experiments
	6 Results and Discussion
	7 Conclusions
	8 Future Works
	References
Adaptive Style Transfer Using SISR
	1 Introduction
	2 Related Work
	3 Proposed Scheme
		3.1 SISR Network
		3.2 Style Transfer Network
		3.3 Loss Function
	4 Experiments
		4.1 Experimental Setup
		4.2 Results
		4.3 Conclusion
	References
Object-Centric Anomaly Detection Using Memory Augmentation
	1 Introduction
	2 Related Work
		2.1 Reconstruction, Prediction and Hybrid Approaches
		2.2 Memory Augmentation
		2.3 Object-Centric Anomaly Detection
	3 ObjMemAE Method
		3.1 Object Detection
		3.2 Autoencoder: Encoder - Memory Module - Decoder
		3.3 Anomaly Score
	4 Experiments
		4.1 Dataset
		4.2 Evaluation Metric
		4.3 Implementation Details
	5 Results and Discussion
		5.1 Quantitative Results
		5.2 Ablation Study
		5.3 Qualitative Results
	6 Conclusion
	References
Document Language Classification: Hierarchical Model with Deep Learning Approach
	1 Introduction
		1.1 Literature Review
	2 Proposed Approach
		2.1 Preprocessing
		2.2 Convolutional Neural Network Model
	3 Experimental Results
		3.1 Dataset
		3.2 Performance of the Proposed Approach
	4 Conclusion
	References
Parsing Digitized Vietnamese Paper Documents
	1 Introduction
	2 Related Work
		2.1 Existing Datasets
		2.2 Parsing Digitized Paper Documents Problem
	3 UIT-DODV Dataset
		3.1 Dataset Collection
		3.2 Category Selection
		3.3 Dataset Description
	4 Computational Model
		4.1 Object Detector
		4.2 Loss Function
	5 Experimental Results and Discussion
		5.1 Experimental Setting
		5.2 Analysis Results
	6 Conclusion and Future Work
	References
EnGraf-Net: Multiple Granularity Branch Network with Fine-Coarse Graft Grained for Classification Task
	1 Introduction
		1.1 Related Work
	2 Methodology
		2.1 Network Architecture
		2.2 Loss Functions
	3 Experiments
		3.1 Results
		3.2 Visualization Analysis
	4 Conclusions
	References
When Deep Learners Change Their Mind: Learning Dynamics for Active Learning
	1 Introduction
	2 Related Work
	3 Active Learning for Image Classification
		3.1 Label-Dispersion Acquisition Function
		3.2 Informativeness Analysis
	4 Experimental Results
		4.1 Experimental Setup
		4.2 Results
	5 Conclusion
	References
Learning to Navigate in the Gaussian Mixture Surface
	1 Introduction
	2 Related Works
	3 Gaussian Mixture Models
	4 Gaussian Mixture Centers Loss
		4.1 Scheme
	5 Experiments
		5.1 Datasets and Implementation Details
		5.2 Ablation Experiments
		5.3 Visualization Analysis
	6 Conclusions
	References
A Deep Hybrid Approach for Hate Speech Analysis
	1 Introduction
	2 Materials and Methods
		2.1 Dataset Description and Construction
		2.2 Proposed Model Architecture and Experimental Setup
	3 Results and Discussion
	4 Conclusion
	References
On Improving Generalization of CNN-Based Image Classification with Delineation Maps Using the CORF Push-Pull Inhibition Operator
	1 Introduction
	2 Related Works
	3 Methods
		3.1 Overview
		3.2 CORF Operator with Push-Pull Inhibition
		3.3 AlexNet
		3.4 Image Perturbations
	4 Experiments and Results
		4.1 Data Set
		4.2 Experiments
	5 Discussion
	6 Conclusion
	References
Fast Hand Detection in Collaborative Learning Environments
	1 Introduction
	2 Methodology
		2.1 Hand Detection Method
		2.2 Optimal Data Augmentation
	3 Results
		3.1 Results for Optimal Data Augmentation
		3.2 Hand Detection Results
	4 Conclusion
	References
Assessing the Role of Boundary-Level Objectives in Indoor Semantic Segmentation
	1 Introduction
	2 Related Work
	3 Method
		3.1 Boundary Loss
		3.2 Active Boundary Loss
	4 Experiments
		4.1 Dataset
		4.2 Implementation Details and Evaluation Protocol
		4.3 Quantitative Evaluation
	5 Conclusion
	References
Skin Lesion Classification Using Convolutional Neural Networks Based on Multi-Features Extraction
	1 Introduction
	2 Related Works
	3 Methodology and Materials
		3.1 PH2 Dataset
		3.2 Convolutional Neural Networks
		3.3 Classifiers
		3.4 Evaluation Metrics
	4 Experimental Results and Discussion
	5 Conclusion
	References
Recursively Refined R-CNN: Instance Segmentation with Self-RoI Rebalancing
	1 Introduction
	2 Related Works
	3 Recursively Refined R-CNN
	4 Experiments
		4.1 Dataset and Evaluation Metrics
		4.2 Implementation Details
		4.3 Analysis of R3-CNN
		4.4 Ablation Study on the Evaluation Phase
		4.5 Ablation Study on the Training Phase
		4.6 Extensions on R3-CNN
	5 Conclusions
	References
Layer-Wise Relevance Propagation Based Sample Condensation for Kernel Machines
	1 Introduction
	2 Kernel Machines
	3 Sample Condensation Method
		3.1 LRP for Relevance Measure of Kernel Machine
		3.2 Relevance-Based Sample Condensation
	4 Experimental Validation
		4.1 Data Sets
		4.2 Results
	5 Conclusion
	References
Author Index