Image segmentation based on Bayesian network-Markov random field model and its application to in vivo plaque composition

Combining Bayesian network (BN) and Markov random field (MRF) models, this paper presents an effective supervised image segmentation algorithm. Representing information from different features, a Bayesian network generates the probability map for each pixel via the conditional PDF (probability density function) learned from a limited training data set. Considering the spatial relation and a priori knowledge of the image, MRF theory is used to generate a reasonable segmentation by minimizing the proposed energy functional. Applying this algorithm to multi-contrast MR image in vivo plaque composition measurement shows comparable results with expert manual segmentation