UNSUPERVISED IMAGE SEGMENTATION USING ADAPTIVE FRAGMENTATION IN PARALLEL MRF-BASED WINDOWS FOLLOWED BY BAYESIAN CLUSTERING

The approach presented in this paper was intended for extending conven tional Markov random field (MRF) models to a more practical problem: t he unsupervised and adaptive segmentation of gray-level images. The '' unsupervised'' segmentation means that all the model parameters, inclu ding the number of image classes, are unknown and have to be estimated from the observed image. In addition, the ''adaptive'' segmentation m eans that both the region distribution and the image feature within a region are all location-dependent and their corresponding parameters m ust be estimated from location to location. We estimated local paramet ers independently from multiple small windows under the assumption tha t an observed image consists of objects with smooth surfaces, no textu re. Due to this assumption, the intensity of each region is a slowly v arying function plus noise, and the conventional homogeneous hidden MR F (HMRF) models are appropriate for these windows. In each window, we employed the EM algorithm for maximum-likelihood (ML) parameter estima tion, and then, the estimated parameters were used for ''maximizer of the posterior marginals'' (MPM) segmentation. To keep continuous segme nts between windows, a scheme for combining window Fragments was propo sed. The scheme comprises two parts: the programming of windows and th e Bayesian merging of window Fragments. Finally, a remerging procedure is used as post processing to remove the over-segmented small regions that possibly exist after the Bayesian merging. Since the final segme nts are obtained From merging, the number of image classes is automati cally determined. The use of multiple parallel windows makes our algor ithm to be suitable for parallel implementation. The experimental resu lts of real-world images showed that the surfaces (objects) consistent with our reasonable model assumptions were all correctly segmented as connected regions.