Fast variational Bayesian signal recovery in the presence of Poisson-Gaussian noise

Abstract : This paper presents a new method for solving linear inverse problems where the observations are corrupted with a mixed Poisson-Gaussian noise. To generate a reliable solution, a regularized approach is often adopted in the literature. In this context, the optimal selection of the regularization parameters is of crucial importance in terms of estimation performance. The variational Bayesian-based approach we propose in this work allows us to automatically estimate the original signal and the associated regularization parameter from the observed data. A majorization-minimization technique is employed to circumvent the difficulties raised by the intricate form of the Poisson-Gaussian likelihood. Experimental results show that the proposed method is fast and achieves state-of-the art performance in comparison with approaches where the regularization parameters are manually adjusted.
Type de document :
Communication dans un congrès
IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Mar 2016, Shangaie, China. Proceedings of the 39th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2016), p. 3964-3968, 2016, 〈10.1109/icassp.2016.7472421 〉
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01370499
Contributeur : Emilie Chouzenoux <>
Soumis le : jeudi 22 septembre 2016 - 16:38:18
Dernière modification le : lundi 29 janvier 2018 - 16:55:06

Identifiants

Citation

Yosra Marnissi, Yuling Zheng, Jean-Christophe Pesquet. Fast variational Bayesian signal recovery in the presence of Poisson-Gaussian noise. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Mar 2016, Shangaie, China. Proceedings of the 39th IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2016), p. 3964-3968, 2016, 〈10.1109/icassp.2016.7472421 〉. 〈hal-01370499〉

Partager

Métriques

Consultations de la notice

124