Research Of Lip Contour Extraction In Face Recognition

The lip contour extraction technology in face recognition is a hot research topic in computer vision. This technology is widely used in facial expression classification, lip recognition, fatigue driving test and other fields. Its research and development not only has great theoretical significance, but also has wide application prospect and practical significance.This paper studies the lip contour extraction problem in face recognition. Acoording the practical difficulties in traditional lips contour extraction methods, we proposed two improved level set methods based on Fisher transformation and semi-supervised learning algorithm for lip contour extraction. Firstly, illumination compensation is performed based on the adaptive Gamma correction. Then, we constructed a skin-color model for skin detection in CbCr color space so that the approximate lip region can be obtained based on the geometry features of human face. Next, we proposed two improved level set methods to extract the lip contour. Finally, we realize the complete lip contour extraction system by MATLAB, the results show the algorithm is feasible to extract the lip contour. Research of the dissertation is concentrated on the following aspects:(1) An improved level set approach based on Fisher transformation is proposed to strengthen the gradient information of the lip contour and improve the accuracy of contour extraction.(2) Fisher transformation requires a lot of artificial tag data, to solve this problem, we apply the semi-supervised learning combined with level set method to extract the lip contour. The method comprehensive utilized the marked and unmarked image pixel information, it achieved good contour extraction results in the cases with fewer tag information.(3) We realize the complete lip contour extraction system by MATLAB, the results show the algorithms are feasible to extract the lip contour.