Face segmentation algorithm based on ASM

Face segmentation is a primary problem in facial recognition. In this paper, a novel face segmentation algorithm is proposed based on the traditional ASM. For face contour is similar to an elliptical shape from outside character, we deal with the images under polar coordinates and mark the landmarks according to a certain regulation manually. A global shape model and each feature point local texture model are built respectively. For each of the landmarks that describe the shape, at each resolution level take into account during the segmentation local Log-Gabor wavelet character extracted from filtered versions of images and gray level character information. After initialize the shape model, we begin the search stage using the two type models and then use a binary mask image to obtain face segmentation. Experimental result shows that our method has a better performance compared with the traditional segmentation algorithm such as ASM and Snakes.