Face Boundary Extraction

摘要

In this paper, we propose a symmetry-based method for face boundary extraction from a binarized facial image, which contains a number of blobs after thresholding. The salient facial features such as eyes, mouth, and face borders are among the blobs. In our method, the symmetric axis transform is conducted on the exterior contours of the blobs in a binarized facial image by a Constrained Delaunay triangulation. A face shape is decomposed into a few components represented by the chain of three types of Delaunay triangles. The facial features are identified by the symmetry related distance analysis and then the face boundaries are traced with a group of connected Delaunay edges. With this method, face boundary with arbitrary shape can be traced precisely and small broken edges can be linked without obviously distortion.