Digital Image Enhancement and Reconstruction

Description

Digital Image Enhancement and Reconstruction: Techniques and Applications explores different concepts and techniques used for the enhancement as well as reconstruction of low-quality images. Most real-life applications require good quality images to gain maximum performance, however, the quality of the images captured in real-world scenarios is often very unsatisfactory. Most commonly, images are noisy, blurry, hazy, tiny, and hence need to pass through image enhancement and/or reconstruction algorithms before they can be processed by image analysis applications. This book comprehensively explores application-specific enhancement and reconstruction techniques including satellite image enhancement, face hallucination, low-resolution face recognition, medical image enhancement and reconstruction, reconstruction of underwater images, text image enhancement, biometrics, etc. Chapters will present a detailed discussion of the challenges faced in handling each particular kind of image, analysis of the best available solutions, and an exploration of applications and future directions. The book provides readers with a deep dive into denoising, dehazing, super-resolution, and use of soft computing across a range of engineering applications.Presents comprehensive coverage of digital image enhancement and reconstruction techniquesExplores applications across range of fields, including intelligent surveillance systems, human-computer interaction, healthcare, agriculture, biometrics, modellingExplores different challenges and issues related to the implementation of various techniques for different types of images, including denoising, dehazing, super-resolution, and use of soft computing Contents List of contributors Preface Acknowledgments 1 Video enhancement and super-resolution 1.1 Introduction 1.2 Recent related work 1.2.1 Development status of video IE 1.2.2 Application status and trend of Machine Vision (MV) in IP 1.3 Analysis of low-quality video IE model based on DL algorithm 1.3.1 Demand Analysis (DA) of resolution-oriented video IE 1.3.2 Analysis of the application of DL algorithm to video IP 1.3.3 Analysis of video image super-resolution and deblurring model based on high-order gated AM and improved GAN 1.3.4 Experimental analysis 1.4 Results and discussion 1.4.1 Accuracy analysis of video image recognition based on different algorithms 1.4.2 Analysis of data transmission performance of different algorithms 1.4.3 Comparative analysis of acceleration efficiency of various NN algorithms 1.5 Conclusion Acknowledgment References 2 On estimating uncertainty of fingerprint enhancement models 2.1 Introduction Research contributions 2.2 Related work 2.2.1 Fingerprint enhancement 2.2.1.1 Classical image processing techniques for enhancement 2.2.1.2 Learning-based enhancement models 2.2.2 Uncertainty estimation 2.2.2.1 Uncertainty estimation through single network deterministic techniques 2.2.2.2 Uncertainty estimation through ensemble techniques 2.2.2.3 Uncertainty estimation through test-time augmentation techniques 2.2.2.4 Uncertainty estimation through Bayesian techniques 2.3 Model uncertainty estimation 2.3.1 Bayesian neural networks 2.3.2 Approximating inference via Monte Carlo dropout 2.3.3 Estimating model uncertainty 2.4 Data uncertainty estimation 2.4.1 Estimating data uncertainty for regression loss-based fingerprint enhancement models 2.4.2 Estimating data uncertainty for cross-entropy loss-based fingerprint enhancement models 2.5 Experimental evaluation 2.5.1 Databases 2.5.2 Evaluation metrics 2.5.3 Effect of estimating model uncertainty 2.6 Results and analysis 2.6.1 Effect of estimating data uncertainty 2.6.2 Comparison of model and data uncertainty 2.6.2.1 Comparison of model complexity 2.6.2.2 Comparison of inference time 2.6.3 Generalization on fingerprint ROI segmentation 2.6.3.1 Predicted uncertainty 2.6.3.2 Segmentation performance 2.7 Conclusion References 3 Hardware and software based methods for underwater image enhancement and restoration 3.1 Introduction 3.2 Literature survey 3.2.1 Hardware-based methods 3.2.2 Software-based methods 3.2.2.1 Image restoration methods 3.2.2.2 Color correction methods 3.2.2.3 DCP-based methods 3.2.2.4 Fusion-based methods 3.2.2.5 Integrated methods 3.2.2.6 CNN-based methods 3.2.2.7 GAN-based methods 3.3 Research gaps 3.4 Conclusion and future scope References 4 Denoising and enhancement of medical images by statistically modeling wavelet coefficients 4.1 Introduction 4.2 Literature survey 4.3 Background and basic principles 4.3.1 Homomorphic filter 4.3.2 Haar wavelet 4.3.3 MAD estimator 4.3.4 MMSE estimator 4.3.5 Gaussian scale mixture (GSM) model 4.4 Methodology 4.5 Simulation results 4.5.1 Dataset and experimental settings 4.5.2 Performance and quality measurement indexes 4.5.3 Result discussion 4.6 Conclusion References 5 Medical image denoising using convolutional neural networks 5.1 Introduction 5.2 Different medical imaging modalities 5.2.1 Magnetic resonance imaging 5.2.2 Computed tomography 5.2.3 Positron emission tomography 5.2.4 Ultrasound 5.3 Convolutional neural networks 5.4 Review on existing CNN denoisers 5.4.1 CNNs for MR images 5.4.2 CNNs for CT images 5.4.3 CNNs for ultrasound images 5.4.4 CNNs for positron emission tomography (PET) images 5.5 Result and discussion 5.6 Challenges of CNN's denoisers 5.7 Conclusion References 6 Multimodal learning of social image representation 6.1 Introduction 6.2 Representation learning methods of social media images 6.2.1 Multimodal learning of social image representation 6.2.2 Learning network for multimodal representation 6.2.3 Multiview-based representation learning 6.2.4 Applications 6.3 Conclusion References 7 Underwater image enhancement: past, present, and future List of abbreviations 7.1 Introduction 7.2 Underwater environment 7.2.1 Underwater imaging model 7.2.2 Underwater scattering effect 7.3 Underwater image enhancement methods 7.3.1 Supplementary information and hardware-based UI enhancement methods 7.3.2 Nonphysical model-based UI enhancement methods 7.3.3 Physical model-based UI enhancement methods 7.3.4 Data-driven (deep-learning-based) UI enhancement methods 7.4 Underwater image datasets 7.5 Underwater image quality assessment 7.6 Challenges and future recommendations 7.7 Conclusion References 8 A comparative analysis of image restoration techniques 8.1 Introduction 8.2 Reasons for degradation in image 8.2.1 Blurring model 8.2.2 Noise model 8.3 Image restoration techniques 8.3.1 Direct inverse filtering 8.3.2 Wiener filter algorithm 8.3.3 Richardson–Lucy filter algorithm [18] 8.3.4 Regularization filter algorithm 8.3.5 Model selection criterion 8.3.5.1 Akaike information criterion (AIC) [23] 8.3.5.2 Improved AIC for the regularized choice of parameter 8.3.5.3 The Bayesian approach 8.3.6 Discrete wavelet transformation (DWT) 8.3.6.1 Haar wavelet 8.3.6.2 Methodology 8.3.7 Iterative denoising and backward projections 8.3.7.1 Problem formulation 8.3.7.2 Plug-and-play method 8.3.7.3 Iterative denoising and backward projection approach 8.3.8 Fast and adaptive boosting techniques for variational based image restoration 8.3.8.1 Variational modeling 8.3.8.2 Description of boosting techniques 8.3.8.3 Adaptive boosting techniques 8.3.8.4 ADMM method used for image restoration model 8.3.9 Image restoration using DWT in a tile-based manner 8.3.9.1 Detailed design 8.3.9.2 Noise removal algorithm 8.3.10 Hybrid sparsity learning 8.3.10.1 Prior prelearning from the training dataset 8.3.10.2 From structured analysis sparsity coding (SASC) to structured analysis sparsity learning (SASL) 8.4 Performance analysis of image restoration 8.5 Performance analysis 8.6 Conclusion References 9 Comprehensive survey of face super-resolution techniques 9.1 Introduction 9.2 Face image degradation model 9.3 Classification of face hallucination methods 9.3.1 Patch-based method 9.3.1.1 Position patch-based methods 9.3.1.2 Neighbor patch-based methods 9.3.2 Bayesian inference 9.3.2.1 Markov random field 9.3.3 Regularization methods 9.3.3.1 Sparse representation 9.3.3.2 Lp norm 9.3.3.3 Locality-constrained 9.3.3.4 Tikhonov regularization 9.3.3.5 Total variation regularization 9.3.4 Subspace learning 9.3.4.1 Linear subspace learning 9.3.4.2 Nonlinear manifold-based learning 9.3.5 Deep learning 9.4 Assessment criteria of face super-resolution algorithm 9.4.1 Mean opinion score 9.4.2 Mean square error 9.4.3 Peak signal-to-noise ratio 9.4.4 Structural similarity index measure 9.5 Issues and challenges 9.6 Conclusion References 10 Fusion-based backlit image enhancement and analysis of results using contrast measure and SSIM 10.1 Introduction 10.2 Basic HVS characteristics 10.2.1 Contrast enhancement at the local level 10.2.2 Average luminance level adaptation 10.2.3 Consistency in color 10.2.4 Weber's law 10.2.5 Fusion 10.3 Methodology 10.3.1 Contrast stretch 10.3.2 Exposure and gradient maps 10.3.3 Filtering 10.3.4 Fusion 10.4 Result discussion 10.5 Summary References 11 Recent techniques for hyperspectral image enhancement 11.1 Introduction 11.2 The major objective of hyperspectral image enhancement 11.3 Recent techniques in hyperspectral image enhancement by compressing the image 11.4 Advantages of hyperspectral imaging over multispectral imaging 11.5 Application of hyperspectral image 11.6 Conclusion References 12 Classification of COVID-19 and non-COVID-19 lung computed tomography images using machine learning 12.1 Introduction 12.2 Methodology 12.3 Results 12.4 Discussions 12.5 Limitations and future improvements 12.6 Conclusion Acknowledgments References 13 Brain tumor image segmentation using K-means and fuzzy C-means clustering 13.1 Introduction 13.2 Brain tumor extraction using image segmentation 13.2.1 Preprocessing 13.2.2 Brain tumor segmentation 13.2.3 Tumor contouring 13.3 Review on K-means clustering 13.4 Review on fuzzy C-means clustering 13.5 Performance analysis and assessment 13.6 Observations and discussions 13.7 Conclusions References 14 Multimodality medical image fusion in shearlet domain 14.1 Introduction 14.1.1 Frequency domain technique 14.1.2 Spatial domain techniques 14.2 Multimodality medical image fusion 14.2.1 Principal component analysis 14.2.2 Pyramid technique 14.2.3 Discrete wavelet transform (DWT) 14.2.4 Artificial neural networks (ANN) 14.3 Proposed methodology 14.4 Results and discussion 14.5 Conclusions References 15 IIITM Faces: an Indian face image database 15.1 Introduction 15.2 Related work 15.3 Methodology 15.3.1 Camera setup 15.3.2 Overview of IIITM Face dataset 15.3.3 Attribute descriptions 15.3.4 Problem formulation 15.3.5 Evaluation metrics 15.4 Experimental setup 15.4.1 Implementation details 15.5 Results 15.5.1 Application areas 15.6 Conclusions Acknowledgments References Index