Color Image Segmentation Based on a New Geometric Active Contour Model

Abstract

A novel geometric active contour model for color images is proposed in this paper. It combines the new edge detector for color images together with the Chan-Vese minimal variance criterion and geodesic active contour without re-initialization. The new edge detector for color images is more precise than the original edge detector and the Chan-Vese minimal variance criterion can promote the evolution velocity. Experimental results on real color images have shown that the new model can shorten the evolution time compared with geodesic active contours model and can extract contours of objects in images more precisely than Chan-Vese model.