Multiple face contour detection based on geometric active contours

An approach based on a multiphase geometric active contour model is presented for multiple face contour detection in grayscale still images. The proposed method first adopts the multiphase Chan-Vese model for image segmentation. Bassed on the four-color theorem this model can use only two level set functions to represent any partition in the piecewise smooth case. Then the Hausdorff distance is used as a similarity between a general elliptical face model and the segmented closed curves. Based on the predefined threshold, the final face contours are achieved in this validation step. Experiments are performed on a large test set and show the efficiency of the method.