EM ALGORITHM FOR IMAGE SEGMENTATION INITIALIZED BY A TREE STRUCTURE SCHEME

In this correspondence, the objective is to segment vector images, whi ch are modeled as multivariate finite mixtures. The underlying images are characterized by Markov random fields (MRF's), and the applied seg mentation procedure is based on the (EM) technique. We propose an init ialization procedure that does not require any prior information and y et provides excellent initial estimates for the EM method. The perform ance of the overall segmentation is demonstrated by segmentation of si mulated one-dimensional (1-D) and multidimensional magnetic resonance (MR) brain images.