Abstract: Textural and structural characteristics are two of the principal features of human faces，both of which have advantages and disadvantages for face recognition. In this paper，we propose a Gaussian weights based new feature combined with texture feature and structure feature approach. In this method，we first use a standard fiducial point detector to locating five face key-points(e. g.，the mid-point of the eye)，and then a new feature will be generated dynamically by weighting the gray value of the pixel according to the distance between the pixel and its relative key-point. The new feature，originated from the texture feature，ties the geometric structural feature of the pixels and the key-point information and delivers better performance than the original texture feature for pose variations.Meanwhile，the five key-points are beneficial to part-based face recognition. Experimental results show that the new feature can improve recognition rates by about 5%than the original texture feature，and improved the part-based method by combining the new feature and the face blocking;achieving recogni-tion rates close to 100%for two different available databases.