Color image segmentation using the Dempster-Shafer evidence theory for the fusion of uncertain information sources

The evidence theory aims to represent and handle uncertain information. An important property of this theory is its ability to merge different data sources in order to improve the quality of the information. In this paper, a color image segmentation approach based on the Dempster-Shafer’s theory is presented. The three image components (Red, Green and Blue) are considered as uncertain information sources. An automatic thresholding approach is utilized for finding all major homogeneous regions in each images component at first stage. The evidence theory is then used for the fusion of information coming from the three information sources for the same image. The fusion process does not start from a single frame of discernment, as done in most previously reported works, but starts from first defining three independent frames of discernment associated with the three images to be fused, and then combining them for forming a new frame of discernment. The strategy used to define the mass distributions in the combined framework is discussed in detail. The proposed segmentation algorithm has been applied to textured and biomedical cell image in order to illustrate the methodology. The obtained results show the robustness of the method.