Nonlinear Eigenproblems in Image Processing and Computer Vision

This unique text/reference presents a fresh look at nonlinear processing through nonlinear eigenvalue analysis, highlighting how one-homogeneous convex functionals can induce nonlinear operators that can be analyzed within an eigenvalue framework.

Saved in:
Bibliographic Details
Main Author Gilboa, Guy
Format eBook
LanguageEnglish
Published Cham Springer Nature 2018
Springer International Publishing AG
Springer International Publishing
Edition1
SeriesAdvances in Computer Vision and Pattern Recognition
Subjects
Online AccessGet full text
ISBN9783319758473
3319758470
9783319758466
3319758462
ISSN2191-6586
2191-6594
DOI10.1007/978-3-319-75847-3

Cover

Table of Contents:
  • 3.4.2 ROF, TV-L1, and TV Flow -- References -- 4 Eigenfunctions of One-Homogeneous Functionals -- 4.1 Introduction -- 4.2 One-Homogeneous Functionals -- 4.3 Properties of Eigenfunction -- 4.4 Eigenfunctions of TV -- 4.4.1 Explicit TV Eigenfunctions in 1D -- 4.5 Pseudo-Eigenfunctions -- 4.5.1 Measure of Affinity of Nonlinear Eigenfunctions -- References -- 5 Spectral One-Homogeneous Framework -- 5.1 Preliminary Definitions and Settings -- 5.2 Spectral Representations -- 5.2.1 Scale Space Representation -- 5.3 Signal Processing Analogy -- 5.3.1 Nonlinear Ideal Filters -- 5.3.2 Spectral Response -- 5.4 Theoretical Analysis and Properties -- 5.4.1 Variational Representation -- 5.4.2 Scale Space Representation -- 5.4.3 Inverse Scale Space Representation -- 5.4.4 Definitions of the Power Spectrum -- 5.5 Analysis of the Spectral Decompositions -- 5.5.1 Basic Conditions on the Regularization -- 5.5.2 Connection Between Spectral Decompositions -- 5.5.3 Orthogonality of the Spectral Components -- 5.5.4 Nonlinear Eigendecompositions -- References -- 6 Applications Using Nonlinear Spectral Processing -- 6.1 Generalized Filters -- 6.1.1 Basic Image Manipulation -- 6.2 Simplification and Denoising -- 6.2.1 Denoising with Trained Filters -- 6.3 Multiscale and Spatially Varying Filtering Horesh-Gilboa -- 6.4 Face Fusion and Style Transfer -- References -- 7 Numerical Methods for Finding Eigenfunctions -- 7.1 Linear Methods -- 7.2 Hein-Buhler -- 7.3 Nossek-Gilboa -- 7.3.1 Flow Main Properties -- 7.3.2 Inverse Flow -- 7.3.3 Discrete Time Flow -- 7.3.4 Properties of the Discrete Flow -- 7.3.5 Normalized Flow -- 7.4 Aujol et al. Method -- References -- 8 Graph and Nonlocal Framework -- 8.1 Graph Total Variation Analysis -- 8.2 Graph P-Laplacian Operators -- 8.3 The Cheeger Cut -- 8.4 The Graph 1-Laplacian -- 8.5 The p-flow -- 8.5.1 Flow Main Properties
  • Intro -- Preface -- What are Nonlinear Eigenproblems and Why are They Important? -- Basic Intuition and Examples -- What is Covered in This Book? -- References -- Acknowledgements -- Contents -- 1 Mathematical Preliminaries -- 1.1 Reminder of Very Basic Operators and Definitions -- 1.1.1 Integration by Parts (Reminder) -- 1.1.2 Distributions (Reminder) -- 1.2 Some Standard Spaces -- 1.3 Euler-Lagrange -- 1.3.1 E-L of Some Functionals -- 1.3.2 Some Useful Examples -- 1.3.3 E-L of Common Fidelity Terms -- 1.3.4 Norms Without Derivatives -- 1.3.5 Seminorms with Derivatives -- 1.4 Convex Functionals -- 1.4.1 Convex Function and Functional -- 1.4.2 Why Convex Functions Are Good? -- 1.4.3 Subdifferential -- 1.4.4 Duality-Legendre-Fenchel Transform -- 1.5 One-Homogeneous Functionals -- 1.5.1 Definition and Basic Properties -- References -- 2 Variational Methods in Image Processing -- 2.1 Variation Modeling by Regularizing Functionals -- 2.1.1 Regularization Energies and Their Respective E-L -- 2.2 Nonlinear PDEs -- 2.2.1 Gaussian Scale Space -- 2.2.2 Perona-Malik Nonlinear Diffusion -- 2.2.3 Weickert's Anisotropic Diffusion -- 2.2.4 Steady-State Solution -- 2.2.5 Inverse Scale Space -- 2.3 Optical Flow and Registration -- 2.3.1 Background -- 2.3.2 Early Attempts for Solving the Optical Flow Problem -- 2.3.3 Modern Optical Flow Techniques -- 2.4 Segmentation and Clustering -- 2.4.1 The Goal of Segmentation -- 2.4.2 Mumford-Shah -- 2.4.3 Chan-Vese Model -- 2.4.4 Active Contours -- 2.5 Patch-Based and Nonlocal Models -- 2.5.1 Background -- 2.5.2 Graph Laplacian -- 2.5.3 A Nonlocal Mathematical Framework -- 2.5.4 Basic Models -- References -- 3 Total Variation and Its Properties -- 3.1 Strong and Weak Definitions -- 3.2 Co-area Formula -- 3.3 Definition of BV -- 3.4 Basic Concepts Related to TV -- 3.4.1 Isotropic and Anisotropic TV
  • 8.5.2 Numerical Scheme -- 8.5.3 Algorithm -- References -- 9 Beyond Convex Analysis-Decompositions with Nonlinear Flows -- 9.1 General Decomposition Based on Nonlinear Denoisers -- 9.1.1 A Spectral Transform -- 9.1.2 Inverse Transform, Spectrum, and Filtering -- 9.1.3 Determining the Decay Profiles -- 9.2 Blind Spectral Decomposition -- 9.3 Theoretical Analysis -- 9.3.1 Generalized Eigenvectors -- 9.3.2 Relation to Known Transforms -- References -- 10 Relations to Other Decomposition Methods -- 10.1 Decomposition into Eigenfunctions -- 10.2 Wavelets and Hard Thresholding -- 10.2.1 Haar Wavelets -- 10.3 Rayleigh Quotients and SVD Decomposition -- 10.4 Sparse Representation by Eigenfunctions -- 10.4.1 Total Variation Dictionaries -- 10.4.2 Dictionaries from One-Homogeneous Functionals -- References -- 11 Future Directions -- 11.1 Spectral Total Variation Local Time Signatures for Image Manipulation and Fusion -- 11.2 Spectral AATV (Adapted Anisotropic Total Variation) … -- 11.3 TV Spectral Hashing -- 11.4 Some Open Problems -- Reference -- A Numerical Schemes -- A.1 Derivative Operators -- A.2 Discretization of PDE's -- A.2.1 Discretized Differential Operators -- A.2.2 Evolutions -- A.2.3 CFL Condition -- A.3 Basic Numerics for Solving TV -- A.3.1 Explicit Method -- A.3.2 Lagged Diffusivity -- A.3.3 Chambolle's Projection Algorithm -- A.4 Modern Optimization Methods -- A.4.1 The Proximal Operator -- A.4.2 Examples of Proximal Functions -- A.4.3 ADMM -- A.4.4 FISTA -- A.4.5 Chambolle-Pock -- A.5 Nonlocal Models -- A.5.1 Basic Discretization -- A.5.2 Steepest Descent -- Appendix Glossary -- References -- Index