Methods for Solving Regularized Inverse Problems: From Non-Euclidean Fidelities to Computational Imaging Applications

Many branches of science and engineering are concerned with the problem of recording signals from physical phenomena. However, an acquisition system does not always directly provide the high-quality signal representations that a given application requires... Continuer

Many branches of science and engineering are concerned with the problem of recording signals from physical phenomena. However, an acquisition system does not always directly provide the high-quality signal representations that a given application requires. Signal processing and the study of inverse problems offer a set of powerful tools to recover a good signal quality from altered raw measurements.
After an overview of the field, this thesis presents three contributions. The first contribution focuses on recovering a key structural property of a sparse signal, its support. It discusses guarantees associated to a convex optimization method with atypical fidelity, e.g., using a non-Euclidean norm. The second part introduces a method for learning a convolutional dictionary, used as a multimodal imaging prior.
This constitutes a practical way of sharing information between several imaging modalities, such as depth and light intensity. The last contribution revolves around the design of two multispectral compressive imaging strategies using spectrally filtered sensors. The first scheme relies on a generalized inpainting formulation in the multispectral volume, while the second system leverages the principles of compressed sensing from coded optical convolutions. This last chapter studies and compares both sensing models and discusses implementation challenges and tradeoffs.


Livre broché - En anglais 28,80 €
Info Les commandes en ligne se font via notre partenaire i6doc.

Spécifications


Éditeur
Presses universitaires de Louvain
Auteur
Kévin Degraux,
Collection
Thèses de l'École polytechnique de Louvain
Langue
anglais
Catégorie (éditeur)
Sciences appliquées > Ingénierie mathématique
Catégorie (éditeur)
Sciences appliquées
BISAC Subject Heading
TEC000000 TECHNOLOGY & ENGINEERING > TEC009000 TECHNOLOGY & ENGINEERING / Engineering (General)
Code publique Onix
06 Professional and scholarly
CLIL (Version 2013 )
3069 TECHNIQUES ET SCIENCES APPLIQUEES
Date de première publication du titre
19 octobre 2017
Type d'ouvrage
Thèse
Avec
Bibliographie, Lexique

Livre broché


Details de produit
1 Couverture pelliculée
Date de publication
19 octobre 2017
ISBN-13
978-2-87558-605-6
Ampleur
Nombre de pages de contenu principal : 226
Dépôt Légal
D/2017/9964/42 Louvain-la-Neuve, Belgique
Code interne
95771
Format
16 x 24 x cm
Poids
367 grammes
Type de packaging
Aucun emballage extérieur
Prix
28,80 €
ONIX XML
Version 2.1, Version 3

Google Livres Aperçu


Publier un commentaire sur cet ouvrage

Si vous avez une question, utilisez plutôt notre formulaire de contact

Sommaire


Nomenclature xxi
1 Introduction 1
1.1 Motivations . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Outline and contributions . . . . . . . . . . . . . . . . . . 4
2 Preliminaries 7
2.1 Regularized inverse problems . . . . . . . . . . . . . . . . 7
2.1.1 Forward model . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Low complexity priors . . . . . . . . . . . . . . . . 15
2.1.3 Sensing model and embedding . . . . . . . . . . . 25
2.2 Recovery methods . . . . . . . . . . . . . . . . . . . . . . 32
2.2.1 General optimization formulation . . . . . . . . . 33
2.2.2 Non-convex recovery methods . . . . . . . . . . . 35
2.2.3 Convex recovery methods . . . . . . . . . . . . . . 41
2.2.4 Algorithms for convex optimization . . . . . . . . 46
2.2.5 Dictionary Learning . . . . . . . . . . . . . . . . . 51
3 Sparse Support Recovery with Convex Fidelity Constraint 57
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.1.1 Sparse regularization with convex fidelity constraint 58
3.1.2 Dual Certificates . . . . . . . . . . . . . . . . . . . 60
3.1.3 Main result for sparse support recovery . . . . . . 62
3.1.4 Relation to PriorWorks . . . . . . . . . . . . . . . 64
x Table of contents
3.2 Preliminaries and main result . . . . . . . . . . . . . . . . 65
3.2.1 Noiseless support stability . . . . . . . . . . . . . . 65
3.2.2 Model subspace and restricted injectivity conditions 66
3.2.3 Formal statement of the main result . . . . . . . . 71
3.3 Proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.3.1 Proofs of the lemmas and subdifferential decomposability
. . . . . . . . . . . . . . . . . . . . . . . 74
3.3.2 Proof of Theorem 5 . . . . . . . . . . . . . . . . . . 83
3.4 Numerical experiments . . . . . . . . . . . . . . . . . . . . 91
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4 Online Convolutional Dictionary Learning for
Multimodal Imaging 95
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.1.1 Main Contributions . . . . . . . . . . . . . . . . . . 97
4.1.2 RelatedWork . . . . . . . . . . . . . . . . . . . . . 99
4.2 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . 100
4.2.1 Problem Formulation . . . . . . . . . . . . . . . . 100
4.2.2 Online Convolutional Dictionary Learning
Algorithm . . . . . . . . . . . . . . . . . . . . . . . 104
4.2.3 Dictionary update . . . . . . . . . . . . . . . . . . 105
4.2.4 Implementation details . . . . . . . . . . . . . . . 107
4.3 Experimental Evaluation . . . . . . . . . . . . . . . . . . . 110
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5 Multispectral Compressive Imaging Strategies using
Fabry-Pérot Filtered Sensors 119
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 120
5.1.1 Main Contributions . . . . . . . . . . . . . . . . . . 121
5.1.2 RelatedWork . . . . . . . . . . . . . . . . . . . . . 122
5.2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.2.1 Fabry-Pérot Filtered Sensors . . . . . . . . . . . . 125
5.2.2 Forward model and analysis prior . . . . . . . . . 127
Table of contents xi
5.2.3 Recovery Method . . . . . . . . . . . . . . . . . . . 128
5.3 Multispectral Compressive Imaging by Generalized Inpainting
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.3.1 Image Formation Model . . . . . . . . . . . . . . . 132
5.3.2 Simulations . . . . . . . . . . . . . . . . . . . . . . 135
5.3.3 Experiments . . . . . . . . . . . . . . . . . . . . . . 138
5.4 Multispectral Compressive Imaging by Out-of-Focus
Random Convolution . . . . . . . . . . . . . . . . . . . . . 140
5.4.1 Image Formation Model . . . . . . . . . . . . . . . 140
5.4.2 Non-idealities and practical considerations . . . . 145
5.4.3 Sensing matrix implementation . . . . . . . . . . . 151
5.4.4 Simulations . . . . . . . . . . . . . . . . . . . . . . 153
5.5 Final Comparison . . . . . . . . . . . . . . . . . . . . . . . 155
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
6 Conclusions 163
6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
6.2 Perspectives and open questions . . . . . . . . . . . . . . 166
References 173
Appendix A Elements of Convex Optimization 195


Extrait