Extending active shape models with color information for facial features localization

Face segmentation detects the locations of faces in images. In this paper we present a new approach to extend active shape models with color information for facial features localization. Fisher transformed profile models are incorporated into cost function. As fisher transform enlarges the divergence factor between the color vectors inside an object and those outside it, the use of Fisher transform operation leads to a modified image gradient. When optimization is operated based on Fisher transformed profile models, the contour evolution for each feature becomes robust to the interference of noise edges. The experiments results show that our method leads to considerable-improvement over earlier methods.