An enhanced Active Shape Model for facial features extraction
Chengzhi Sun, Mei Hua Xie · 2008
Active Shape Model (ASM) has been widely recognized as one of the best methods for image features extraction. In this paper, we propose an Enhanced ASM (EASM)algorithm for face image features extraction The ASM search algorithm only use the local texture constraint around each landmark points. The EASM combines both local texture constraint and global texture constraint in EASM search. In the EASM algorithm, each landmark is firstly matched by its local constraint in its current neighborhood, and we adjust the shape model parameters to get the current shape model. Then, the novel image global texture is reconstructed according to the current shape model. After that, we evaluate the fitting degree between the novel image global texture and the reconstructed global texture for the novel image, and update the shape parameters according the fitting degree. Experiments show that our proposed EASM algorithm can extract the facial features more accurately than tradition ASM.