Statistical Model Based Image Segmentation Using Region Growing, Contour Relaxation And Classification

A three-step segmentation method is introduced that is based on a layered statistical model both for region shapes as well as for the texture signals inside the regions. The result of an initial region growing phase is refined by a contour relaxation algorithm which, yields smooth region outlines. Finally, the remaining 'false' contours are detected and suppressed using a contour classifier.