Mpetition: Unifying Snakes, Egion Growin and Bayes/MDL for Image Segmentation

Song Chun Zhu,Alan Yuille

1996-01-01

Abstract:We present a novel statistical and variational approach to image segmentation based on a new algorithm named region compefifion. This algorithm is derived by minimizing a generalized BayesiMDL criterion using the variational principle. The algorithm is guaranteed to converge to a local minimum and combines aspects of snakesiballoons and region growing. Indeed the classic snakesiballoons and region growing algorithms can be directly derived from our approach. We provide theoretical analysis of region competition including accuracy of boundary location, criteria for initial conditions, and the relationship to edge detection using filters. It is straightforward to generalize the algorithm to multiband segmentation and we demonstrate it on gray level images, color images and texture images. The novel color model allows us to eliminate intensity gradients and shadows, thereby obtaining segmentation based on the albedos of objects. It also helps detect highlight regions. Index Terms-Image segmentation, region growing, snakes, minimum description length, Bayes statistics, uncertainty principle, color model.