Color image segmentation based on three levels of texture statistical evaluation

Abstract

In this paper a new and efficient supervised method for color image segmentation is presented. This method improves a part of the automatic extraction problem. The basic technique consists in fusing information streaming from three different sources for the same image. The first source uses information coming from only one pixel, using the Mahalanobis distance. The second uses the multidimensional distribution of the three bands in a window centered in each pixel, using the Bhattacharyya distance. And the third employs cooccurrence matrices over the texture cube built around one pixel, using the Bhattacharyya distance again. The Dempster–Shafer theory of evidence is applied in order to fuse the information from the three sources which represent different orders of statistics. This method reveals the importance of applying context and textural properties for the segmentation process. The results show the potential of the method for real images starting from the three RGB bands only.