LEADER 10952nam 2200505 450 001 996464421103316 005 20211211052238.0 010 $a3-030-75549-5 035 $a(CKB)4100000011912211 035 $a(MiAaPQ)EBC6578620 035 $a(Au-PeEL)EBL6578620 035 $a(OCoLC)1249507995 035 $a(PPN)255289510 035 $a(EXLCZ)994100000011912211 100 $a20211211d2021 uy 0 101 0 $aeng 135 $aurcnu|||||||| 181 $ctxt$2rdacontent 182 $cc$2rdamedia 183 $acr$2rdacarrier 200 00$aScale space and variational methods in computer vision $e8th international conference, SSVM 2021, Virtual Event, May 16-20, 2021, proceedings /$fAbderrahim Elmoataz [and four others], editors 210 1$aCham, Switzerland :$cSpringer,$d[2021] 210 4$d©2021 215 $a1 online resource (584 pages) 225 1 $aLecture Notes in Computer Science ;$v12679 311 $a3-030-75548-7 320 $aIncludes bibliographical references and index. 327 $aIntro -- Preface -- Organization -- Contents -- Scale Space and Partial Differential Equations Methods -- Scale-Covariant and Scale-Invariant Gaussian Derivative Networks -- 1 Introduction -- 2 Relations to Previous Work -- 3 Gaussian Derivative Networks -- 3.1 Provable Scale Covariance -- 4 Experiments with a Single-Scale-Channel Network -- 4.1 Discrete Implementation -- 5 Experiments with a Multi-Scale-Channel Network -- 5.1 Scale Selection Properties -- 6 Summary and Discussion -- References -- Quantisation Scale-Spaces -- 1 Introduction -- 2 Quantisation Scale-Spaces -- 2.1 Scale-Space Properties -- 3 Relations to Sparsification Scale-Spaces -- 4 Applications to Quantisation and Compression -- 4.1 Uncommitted and Committed Quantisation -- 4.2 Inpainting-Based Compression -- 5 Conclusions -- References -- Equivariant Deep Learning via Morphological and Linear Scale Space PDEs on the Space of Positions and Orientations -- 1 Introduction -- 2 Design of PDE-Based Equivariant Neural Network -- 2.1 The Lifting Layer: Extending the Image Domain from Rd to Md -- 2.2 PDE Layers by Linear and Morphological Scale Spaces on Md -- 2.3 PDE-Based Deep Learning by G-CNNs on M2 -- 3 Linear and Morphological Kernel Implementation -- 3.1 Analytic Approximations of -Scale-Space Kernels on M2 -- 3.2 Analytic Approximations of -Dilation/Erosion Kernels on M2 -- 4 Experimental Observations and Analysis -- 5 Conclusion -- References -- Nonlinear Spectral Processing of Shapes via Zero-Homogeneous Flows -- 1 Introduction -- 2 Background -- 2.1 Differential Operators on Manifolds -- 2.2 Vectorial Total Variation -- 2.3 Laplacian-Based Flows -- 2.4 TV Mesh Processing -- 3 Proposed Methods -- 3.1 Naive Method: Unpaired Coordinate Spectral TV -- 3.2 Method 1 (M1): Shape Spectral TV -- 3.3 Method 2 (M2): Conformalized P-Laplace -- 3.4 Method 3 (M3): Directional Shape TV. 327 $a4 Discussion and Conclusion -- References -- Total-Variation Mode Decomposition -- 1 Introduction -- 2 Preliminary -- 2.1 Dynamic Mode Decomposition (DMD) -- 2.2 Total Variation Spectral Decomposition -- 3 DMD of the TV Flow -- 3.1 Closed Form Solution -- 3.2 Flow Transitions and a Fast TV-flow Algorithm -- 3.3 Rescaled-DMD -- 3.4 Analysis of the Rescaled-DMD -- 4 Results and Conclusion -- References -- Fast Morphological Dilation and Erosion for Grey Scale Images Using the Fourier Transform -- 1 Introduction -- 2 Morphological Dilation and Erosion -- 3 Greyscale Dilation Discretisation with Convolution -- 3.1 Analytic Motivation of the Algorithm -- 3.2 Details on the Algorithm -- 4 Experiments -- 5 Conclusion -- References -- Diffusion, Pre-smoothing and Gradient Descent -- 1 Introduction -- 2 Integrability Analysis of Diffusion with Pre-smoothing -- 3 Alternatives -- 4 Experiments -- 5 Summary -- References -- Local Culprits of Shape Complexity -- 1 Introduction -- 2 Method -- 3 Illustrative Experiments -- 4 Summary and Future Work -- References -- Extension of Mathematical Morphology in Riemannian Spaces -- 1 Introduction -- 2 Background on MM and HLO Formulas -- 2.1 Basic Morphological Operators, Scale-Spaces and PDEs -- 2.2 HLO Formulas and Hamilton-Jacobi Equations -- 3 HLO Formulas in Riemannian Manifolds -- 3.1 Viscosity Solutions in Riemannian Manifolds -- 3.2 Example and Some Properties -- 4 Conclusion -- References -- Flow, Motion and Registration -- Multiscale Registration -- 1 Introduction -- 2 Mathematical Modelling -- 2.1 Depiction of the Model -- 2.2 Theoretical Results -- 3 Numerical Resolution -- 4 Numerical Experiments -- 5 Conclusion -- References -- Challenges for Optical Flow Estimates in Elastography -- 1 Introduction and Motivation -- 2 Displacement Field Estimation via Optical Flow -- 2.1 Speckle Tracking. 327 $a2.2 Boundary Conditions -- 2.3 Homogeneous Background Information -- 3 The Elastographic Optical Flow Method -- 4 Numerical Results -- 4.1 Simulated Data -- 4.2 Experimental Data -- 5 Summary -- References -- An Anisotropic Selection Scheme for Variational Optical Flow Methods with Order-Adaptive Regularisation -- 1 Introduction -- 2 Variational Optical Flow -- 3 Regularisation with Isotropic Selection Scheme -- 4 Regularisation with Anisotropic Selection Scheme -- 5 Minimisation -- 6 Evaluation -- 6.1 Stand-Alone Approach -- 6.2 Refinement Approach -- 7 Conclusion -- References -- Low-Rank Registration of Images Captured Under Unknown, Varying Lighting -- 1 Introduction -- 2 Related Works -- 3 Low-Rank Registration of Photometric Stereo Images -- 3.1 Displacement Parameterization -- 3.2 Low-Rank Formulation and Convex Relaxation -- 4 Augmented Lagrangian Framework -- 4.1 Splitting -- 4.2 Updating the Corrected Observation Matrix A -- 4.3 Updating the Error Vector e -- 4.4 Updating the Displacement Coefficients -- 5 Implementation -- 5.1 Multi-scale Scheme -- 5.2 Using a Sparse Subset of Pixels -- 5.3 Empirical Evaluation -- 6 Conclusions and Perspectives -- References -- Towards Efficient Time Stepping for Numerical Shape Correspondence -- 1 Introduction -- 2 Modelling of the Shape Correspondence Framework -- 3 Spatial Discretisation -- 4 Time Discretisation -- 5 Experiments -- 6 Conclusion and Further Work -- References -- First Order Locally Orderless Registration -- 1 Introduction -- 2 Related Work -- 3 Background on Locally Orderless Image Information -- 3.1 Notations -- 3.2 Lebesgue Integration and Histograms -- 3.3 LOI and LOR Framework -- 4 Extension of LOI and LOR to Higher Information -- 4.1 First Order Locally Orderless Registration (FLOR) -- 4.2 First Order Deformation Model -- 4.3 Registration Objectives and Deformations. 327 $a4.4 Implementation -- 5 Experiments and Results -- 5.1 The Similarity Properties -- 5.2 Non-rigid Registration -- 5.3 Discussion and Limitations -- 6 Conclusion -- References -- Optimization Theory and Methods in Imaging -- First-Order Geometric Multilevel Optimization for Discrete Tomography -- 1 Introduction -- 2 Multilevel Optimization in Euclidean Space -- 3 Geometric Approach -- 4 Experiments -- 5 Conclusion -- References -- Bregman Proximal Gradient Algorithms for Deep Matrix Factorization -- 1 Introduction -- 2 Related Work -- 3 Bregman Proximal Minimization -- 3.1 Bregman Proximal Gradient -- 4 BPG for Deep Matrix Factorization -- 4.1 Smooth Adaptable Property for DLNN -- 4.2 Closed Form Updates for BPG -- 4.3 Global Convergence of BPG for Regularized DLNN -- 4.4 Discussion of BPG Variants -- 5 Experiments -- 6 Conclusion -- References -- Hessian Initialization Strategies for -BFGS Solving Non-linear Inverse Problems -- 1 Introduction -- 2 Proposed Hessian Initialization Strategy and Algorithm -- 3 Numerical Experiments and Results -- 3.1 Quadratic Problem -- 3.2 Image Registration -- 4 Conclusion -- References -- Inverse Scale Space Iterations for Non-convex Variational Problems Using Functional Lifting -- 1 Motivation and Introduction -- 2 Sublabel-Accurate Lifting Approach -- 3 Equivalency of the Lifted Bregman Iteration -- 4 Numerical Discussion and Results -- References -- A Scaled and Adaptive FISTA Algorithm for Signal-Dependent Sparse Image Super-Resolution Problems -- 1 Introduction -- 2 Scaled and Adaptive FISTA -- 3 Application to Sparse Weighted Models for Super-Resolution Microscopy -- 3.1 Sparse Image Super-Resolution with Poisson Data -- 3.2 Solving (8) Efficiently: IRL1 Algorithm with S-FISTA Solver -- 4 Numerical Results -- 4.1 Weighted-2-1 Minimization -- 4.2 Weighted-CEL0 Minimization -- 5 Conclusions -- References. 327 $aConvergence Properties of a Randomized Primal-Dual Algorithm with Applications to Parallel MRI -- 1 Introduction -- 2 Algorithm -- 3 Main Result -- 4 Sketch of the Proof -- 5 Proof of Convergence -- 5.1 Proof of Proposition1 i) -- 5.2 Proof of Proposition1 ii)-iii) -- 5.3 Proof of Proposition 1 iv) -- 5.4 Proof of Proposition2 -- 6 Relation to Other Work -- 6.1 Chambolle et al. (2018) -- 6.2 Combettes and Pesquet (2014) -- 6.3 Alacaoglu et al. (2019) -- 7 Numerical Examples -- 8 Conclusions -- References -- Machine Learning in Imaging -- Wasserstein Generative Models for Patch-Based Texture Synthesis -- 1 Introduction -- 2 Gradient of Optimal Transport -- 3 Image Optimization -- 3.1 Mono-Scale Texture Synthesis Algorithm -- 3.2 Multi-scale Texture Synthesis -- 3.3 Experiments -- 4 Training a Convolutional Generative Network -- 4.1 Proposed Algorithm for Semi-discrete Formulation -- 4.2 Experimental Results and Discussions -- 4.3 Evaluation of Texture Synthesis Methods -- References -- Sketched Learning for Image Denoising -- 1 Introduction -- 2 Model Estimation and Denoising with EPLL -- 2.1 Denoising with EPLL -- 2.2 EM -- 3 Compressive GMM Learning from Large Image Patches Database with Sketches -- 3.1 Compressive Mixture Estimation -- 3.2 Design of Sketching Operator: Randomly Sampling the Characteristic Function -- 3.3 Extension to Low Rank Covariances -- 3.4 An Algorithm for Patch Prior Learning from Sketch: LR-COMP (Low Rank Continuous Orthogonal Matching Pursuit) -- 4 Results and Analysis -- 4.1 Experiments with Synthetic Data -- 4.2 Results with Real Images -- 5 Conclusion -- References -- Translating Numerical Concepts for PDEs into Neural Architectures -- 1 Introduction -- 2 Networks from Algorithms for Evolution Equations -- 2.1 Generalised Nonlinear Diffusion -- 2.2 Residual Networks -- 2.3 Expressing Explicit Schemes as Residual Networks. 327 $a2.4 Criteria for Well-Posed and Stable Residual Networks. 410 0$aLecture notes in computer science ;$v12679. 606 $aComputer vision$vCongresses 615 0$aComputer vision 676 $a006.37 702 $aElmoataz$b Abderrahim 801 0$bMiAaPQ 801 1$bMiAaPQ 801 2$bMiAaPQ 906 $aBOOK 912 $a996464421103316 996 $aScale Space and Variational Methods in Computer Vision$9773149 997 $aUNISA