A MARKOV RANDOM-FIELD MODEL-BASED APPROACH TO UNSUPERVISED TEXTURE SEGMENTATION USING LOCAL AND GLOBAL SPATIAL STATISTICS

Many studies have proven that statistical model-based texture segmenta tion algorithms yield good results provided that the model parameters and the number of regions be known a priori. In this correspondence, w e present an unsupervised texture segmentation method that does not re quire knowledge about the different texture regions, their parameters, or the number of available texture classes. The proposed algorithm re lies on the analysis of local and global second and higher order spati al statistics of the original images. The segmentation map is modeled using an augmented-state Markov random field, including an outlier cla ss that enables dynamic creation of new regions during the optimizatio n process. A Bayesian estimate of this map is computed using a determi nistic relaxation algorithm. Results on real-world textured images are presented.