Bayesian entropy estimation applied to non-gaussian robust image segmentation

Abstract

We introduce a new approach for robust image segmentation combining two strategies within a Bayesian framework. The first one is to use a Markov random field (MRF) which allows to introduce prior information with the purpose of image edges preservation. The second strategy comes from the fact that the probability density function (pdf) of the likelihood function is non-Gaussian or unknown, so it should be approximated by an estimated version, which is obtained by using the classical non-parametric or kernel density estimation. This lead us to the definition of a new maximum a posteriori (MAP) estimator based on the minimization of the entropy of the estimated pdf of the likelihood function and the MRF at the same time, named MAP entropy estimator (MAPEE). Some experiments were made for different kind of images degraded with impulsive noise (salt & pepper) and the segmentation results are very satisfactory and promising.