Automatic Facial Feature Points Localization And Application Based On ASM

ABSTRACT:Facial feature points localization takes an important role in the face recognition, facial expression analysis, cartoon face synthesis, etc. It can provide the description for face shape and the information of the local features, such as eyebrows, eyes, nose, mouth and so on.By systematically analyzing and comparing of existing feature points localization methods, the main work of the paper includes:(1) First, the current common algorithms for facial feature points localization are summarized and compared. Then, we focus on the process of ASM searching and multi-resolution frames, then we analyze the problem and some improved algorithms based on the traditional Active Shape Models.(2) Design and implement a system of facial feature points localization and application based on ASM. Considering the accuracy and the computation workload, in this paper, we locate the facial feature points based on face detection and Active Shape Models. In order to improve the accuracy and efficiency of the points localization, we adopt a coarse-to-fine search strategy.(3) Through the data of facial feature points that we get from the algorithm, new points can be added to get a more accurate description by the relationship of the nearby points.(4) In the aspect of handling video with face in it, we not only use the mean shape that we get from training, we also make full use of the result that we get from the last frame, which contributes to higher efficiency.The results of our experiments demonstrate the feasibility of our methods. The use of Active Shape Models has been shown to be an efficient approach to image interpretation and pattern recognition. And the results of experiments can be used in facial sketch generation and facial animation. At last, we discuss the limits of the algorithm, and put forward suggestions for further studies.