Enregistré dans:
Détails bibliographiques
Auteurs principaux: Modrzyk, Thibaut, Etxebeste, Ane, Bretin, Élie, Maxim, Voichita
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2603.24156
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915889744445440
author Modrzyk, Thibaut
Etxebeste, Ane
Bretin, Élie
Maxim, Voichita
author_facet Modrzyk, Thibaut
Etxebeste, Ane
Bretin, Élie
Maxim, Voichita
contents In this paper, we present a novel variational plug-and-play algorithm for Poisson inverse problems. Our approach minimizes an explicit functional which is the sum of a Kullback-Leibler data fidelity term and a regularization term based on a pre-trained neural network. By combining classical likelihood maximization methods with recent advances in gradient-based denoisers, we allow the use of pre-trained Gaussian denoisers without sacrificing convergence guarantees. The algorithm is formulated in the majorization-minimization framework, which guarantees convergence to a stationary point. Numerical experiments confirm state-of-the-art performance in deconvolution and tomography under moderate noise, and demonstrate clear superiority in high-noise conditions, making this method particularly valuable for nuclear medicine applications.
format Preprint
id arxiv_https___arxiv_org_abs_2603_24156
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A convergent Plug-and-Play Majorization-Minimization algorithm for Poisson inverse problems
Modrzyk, Thibaut
Etxebeste, Ane
Bretin, Élie
Maxim, Voichita
Computer Vision and Pattern Recognition
In this paper, we present a novel variational plug-and-play algorithm for Poisson inverse problems. Our approach minimizes an explicit functional which is the sum of a Kullback-Leibler data fidelity term and a regularization term based on a pre-trained neural network. By combining classical likelihood maximization methods with recent advances in gradient-based denoisers, we allow the use of pre-trained Gaussian denoisers without sacrificing convergence guarantees. The algorithm is formulated in the majorization-minimization framework, which guarantees convergence to a stationary point. Numerical experiments confirm state-of-the-art performance in deconvolution and tomography under moderate noise, and demonstrate clear superiority in high-noise conditions, making this method particularly valuable for nuclear medicine applications.
title A convergent Plug-and-Play Majorization-Minimization algorithm for Poisson inverse problems
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.24156