Segmentation d’image médicale par champs de Markov

Abstract

Segmentation (edge detection textures) images is one of the key problems in image processing. Among the various models and approaches developed, some of the statistical methods commonly used are based on the model by Hidden Markov fields (CMC). This success is mainly due to the model's ability to take into account the spatial dependencies of random variables, even when in large numbers, which can exceed one million. In this model the hidden field is assumed Markov and must be estimated from the observed field. Such treatment is possible by analysis of markovianité conditionally. This model was then generalized to Markov fields couples (CMCouples), where it is assumed directly markovianité the pair (X, Y) that offer the same treatment options as CMC and provide a better noise model which allows in particular, to better take into account the existence of textures.