BAYESIAN TEXTURE MODEL SELECTION BY HARMONIC MEAN - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2012

BAYESIAN TEXTURE MODEL SELECTION BY HARMONIC MEAN

Résumé

The paper presents a model selection method for texture images, more specifically, it finds the most adequate model for the pixels' in-teraction. This approach relies on a Bayesian framework, that prob-abilizes all the quantities and determines the joint a posteriori law for the models and the parameters. In order to compute the a pos-teriori model probabilities, the model parameters are marginalized by means of sampling, performed independently for each model in a within-model sampling strategy using a Metropolis-Hastings (M-H) algorithm. The resulting chains are used to compute the evidence of each model by an harmonic averaging of the likelihoods computed for the aforementioned sampled values. The work presented in the following represents a complex formalism based on state of the art methods for parameter estimation, model selection techniques and sampling algorithms, the novelty being the design of such an ap-proach for a texture model selection problem. An image processing application of this kind raises serious difficulties regarding the large amount of data, the data correlation and the highly non-linear de-pendencies of the data with respect to the parameters. Despite these challenges, our method successfully solves the problem of texture model selection and parameter estimation.
Fichier principal
Vignette du fichier
vacar.pdf (347.23 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-00781487 , version 1 (28-01-2013)

Identifiants

  • HAL Id : hal-00781487 , version 1

Citer

Cornelia Vacar, Jean-François Giovannelli, Ana-Maria Roman. BAYESIAN TEXTURE MODEL SELECTION BY HARMONIC MEAN. International Conference on Image Processing, Oct 2012, Orlando, United States. pp.2003-2006. ⟨hal-00781487⟩
122 Consultations
163 Téléchargements

Partager

Gmail Facebook X LinkedIn More