IMAGE SEGMENTATION FROM CONSENSUS INFORMATION

A new approach toward image segmentation is proposed. A set of slightl y different segmentations is derived from the same input and the final result is based on the consensus among them. The perturbations are in troduced by exploiting the probabilistic component of a region adjacen cy graph (RAG) pyramid-based segmentation. From the set of initial seg mentations the cooccurrence probability field is obtained in which glo bal information about the delineated regions becomes locally available . The final segmentation is based on this field and is obtained with t he same hierarchical, RAG pyramid technique. No user set parameters or context-dependent thresholds are required. (C) 1997 Academic Press.