Fully Bayesian Image Segmentation - an Engineering Perspective

Abstract : Developments in Markov Chain Monte Carlo procedures have made it possible to perform fully Bayesian image segmentation. By this we mean that all the parameters are treated identically, be they the segmentation labels, the class parameters or the Markov Random Field prior parameters. We perform the analysis by sampling from the posterior distribution of all the parameters. Sampling from the MRF parameters has traditionally been considered if not intractable then at least computationally prohibitive. In the statistics literature there are descriptions of experiments showing that the MRF parameters may be sampled by approximating the partition function. These experiments are all, however, on \lq toy' problems -- for the typical size of image encountered in engineering applications phase transition behaviour of the models becomes a major limiting factor in the estimation of the partition function. Nevertheless, we show that with some care, fully Bayesian segmentation can be performed on realistic sized images. We also compare the fully Bayesian approach with the approximate pseudolikelihood method.
Type de document :
RR-3017, INRIA. 1996
Liste complète des métadonnées

Contributeur : Rapport de Recherche Inria <>
Soumis le : mercredi 24 mai 2006 - 13:29:29
Dernière modification le : samedi 27 janvier 2018 - 01:31:31
Document(s) archivé(s) le : dimanche 4 avril 2010 - 23:54:23



  • HAL Id : inria-00073677, version 1



Robin Morris, Xavier Descombes, Josiane Zerubia. Fully Bayesian Image Segmentation - an Engineering Perspective. RR-3017, INRIA. 1996. 〈inria-00073677〉



Consultations de la notice


Téléchargements de fichiers