Face contour construction with multiple information

In this paper, we present a new method to obtain the contours of the colored faces. First, we introduce a new nonlinear skin classifier to detect the human skin information in images. Then, a split machine (SM) places an adaptive grid over an image according to the low-level information detected in the representative regions. By filtering the redundant points extracted from the actively divided grid, we obtain a set of points that can coarsely construct the contour of the detailed skin region. Finally, we refine the contour by computing the local minimum energy of each point. Experimental results show that our proposed approach is capable of constructing the contours of the colored faces.