Une approche bayésienne non paramétrique pour l'apprentissage d'un dictionnaire de taille adaptative

Abstract : Dictionary learning for sparse representation is well known in solving inverse problems in image processing. In general, the number of dictionary atoms is fixed in advance. We propose a dictionary learning approach that automatically learns a dictionary of adapted size thanks to a Bayesian non parametric approach : the Indian Buffet Process prior. The noise level is also accurately estimated so that nearly no parameter tuning is needed. The denoising comparative results show the relevance of the proposed method.
Liste complète des métadonnées

Cited literature [16 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01249821
Contributor : Pierre Chainais <>
Submitted on : Sunday, January 3, 2016 - 5:43:57 PM
Last modification on : Saturday, March 23, 2019 - 1:24:39 AM
Document(s) archivé(s) le : Thursday, November 10, 2016 - 7:55:15 PM

File

Dang_Chainais_Gretsi2015.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01249821, version 1

Collections

Citation

Hong-Phuong Dang, Pierre Chainais. Une approche bayésienne non paramétrique pour l'apprentissage d'un dictionnaire de taille adaptative. GRETSI 2015, Sep 2015, Lyon, France. 2015, Actes du GRETSI 2015. 〈http://www.gretsi.fr/colloque2015/〉. 〈hal-01249821〉

Share

Metrics

Record views

224

Files downloads

141