MARKOV RANDOM-FIELD MODELS FOR UNSUPERVISED SEGMENTATION OF TEXTURED COLOR IMAGES

We present an unsupervised segmentation algorithm which uses Markov ra ndom field models for color textures. These models characterize a text ure in terms of spatial interaction within each color plane and intera ction between different color planes. The models are used by a segment ation algorithm based on agglomerative hierarchical clustering, At the heart of agglomerative clustering is a stepwise optimal merging proce ss that at each iteration maximizes a global performance functional ba sed on the conditional pseudolikelihood of the image. A test for stopp ing the clustering is applied based on rapid changes in the pseudolike lihood, We provide experimental results that illustrate the advantages of using color texture models and that demonstrate the performance of the segmentation algorithm on color images of natural scenes. Most of the processing during segmentation is local making the algorithm amen able to high performance parallel implementation.