The Research About ASM Facial Feature's Locating Method Under The Front Small Angle Deflection

Facial feature point locating in the Face Recognition, three-dimensional virtual Face Synthesis, Facial Expression Analysis and synthesis, and has a very important role, is the current computer vision, computer graphics, research in the field of pattern recognition, and other hot spots. At present, a number of algorithms and models proposed, but due to the complexity of the face and facial image imaging environmental impacts, still do not have a general model or algorithm is able to resolve the problem on.In this paper, the current common facial feature point locating algorithm and the model are summarized, and their predecessors on the basis of an initiative based on the shape model (ASM) enhancement. Because traditional methods ASM difficult point in each feature are the result of more accurate, and the traditional method of matching ASM by the initial results of the impact of very large. ASM against the traditional method of defects and shortcomings, especially for light and robust enough attitude change, in order to make the ASM more robust posture changes to achieve positive deflection of the small-angle locating facial features, the paper raised a number of improved methods : (1) a selective and sensitive marker of light points (eyebrows, eyes, nose) using 2-D profile. (2) the use of stacked two ASM series model search, the initial improvements installed improper search failure. (3) a separate extension of the eyes of the marker, so that the characteristics of the eye locating more robust. (4) when used in the search Face Detection Operators Face automatic locating and location of the eyes, and people around the face when small-angle deflection characteristics of a more robust position. Experimental results show that the improvement in the accuracy of the ASM methodology and robustness are greatly enhanced.