Relevance of the Dempster-Shafer Evidence Theory for Image Segmentation

This paper describes a new color image segmentation method based on data fusion techniques. The used methodology modeling in the Dempster-Shafer evidence theory is in general successful, for representing the information extracted from image as measures of belief. The proposed method addresses the information modelization problem and the color image segmentation within the context of Dempster-Shafer theory. The mass functions are computed from the probability that a pixel belong to a region. The mass functions are then combined with the Dempster rules of combination, and the maximum of mass function is used for decision-making. The computation of conflict between images, the modelization of both uncertainty and imprecision, the possible introduction of a priori information, witch are powerful aspects of the evidence theory and witch have a great influence on the final decision, are exploited in color image segmentation. We present quantitative and comparative results concerning color medical images.