Active Shape Model is a useful method for the facial feature point location and the face recognition. Several improvements to the traditional Active Shape Model are proposed in this paper, including: (i) key points based initial shape estimation method; (ii) two-dimension local texture model of the normal and tangent direction for each feature point; (iii) decreasing the sampling pixels on each side of the feature point in the multi-resolution framework approach. Experimental results show that the improved Active Shape Model performs more accurate and efficient than the conventional one.
展开▼
