Facial Components Segmentation for Extracting Facial Feature

초록

Most face recognition systems require the segmentation method of facial components such as mouth, eyes, and nose, etc., to extract facial feature. In this paper, we present a method of finding eyes from the extracted facial components. The proposed method consistes of preprocessing, binarization, segmentation, detection of the candidates of eye regions. The proposed method locates a pair of eye regions using heuristic rules devised from facial geometric characteristics. We develop the gradient magnitude and the intensity image based methods for the binarization step. Face model used in this paper consists of two eyes, two eyebrows and mouth. However, each binarization method cannot be a solution in general environment. To overcome the problem described above, we proposed method has been tested with various face images taken from diverse conditions and archieved very encouraging results.