Image segmentation by iterative region-contour control minimizing a convergence criterion

This paper focuses on the problem of image segmentation. It addresses the inherent deficiencies occurring when extracting low-level features (primitives) and when dealing with the complexity of real scenes. To solve such a problem we propose a general framework which integrates several types of primitives formed progressively from different complementary transformations of the image in an iterative and cooperative way. The main idea consists in initializing the method by several over segmentations. Then, the formation of primitives with increasingly tolerant criteria is iterated until a convergence between the results is reached.