Research On Facial Feature Point Based On ASM

Facial feature point location is one of the fundamental and crucial problems in the field of facial recognition, computer vision and graphics. It aims to locate the facial feature points and the shape information of eyes, mouth and so on based on the facial detect. The state-of-the-art AFR system can perform identification successfully under well-controlled environment. However, evaluation results and practical experience have shown that AFR technologies are currently far from mature. A great number of challenges are to be solved before one can implement a robust practical AFR application, especially the accurate facial feature location problem, which is the prerequisite for sequent feature extraction and classification. In this thesis, facial feature point location using Active Shape Models is studied after a recent overview of AFR research and development. Study facial feature point location problem and provide a thorough survey of the algorithms, and then focus on facial feature point location using Active Shape Model.Point Distribution Model is described, and three aspects of ASM are discussed: aligning the training set, modeling shape variation and choice of number of modes. To improve active shape model (ASM) accuracy in facial feature point location in facial images, an improved ASM method based traditional ASM algorithm is proposed. The irises are located and utilized to initialize the average shape model based on the result of facial detection and three relative parameters are calculated to serve for the initial of average shape model; Facial similar configuration, feature subspaces of global face shape model and salient feature local shape models are all employed to constrain the movement of feature points; Log-Gabor coefficients are used to describe the local texture distribution and built texture model for each feature point to increase the robustness to illumination change and other noise. What's more, gray information andLog-Gabor coefficients are connected to describe the character of each feature point and then to build texture model for each feature point; Edge constrain strategy is used to reduce the search region of each feature point. Experimental results show that our improved ASM algorithms perform significant better than traditional ASM and also have more robust on illumination change and noise. Reason for the fail results are investigated, like changes of illumination, pose, expression and so on.