Research On The Method Of Facial Pose Estimation Based On Chinese Feature Morphological Characteristic

Pose estimation of human faces plays an important role in computer vision and face recognition. It can be used in such as automatic control based on video, 3D face recognition and virtual movie-production etc. Because the direction of face and the position eyes watched can be detected through face posture, pose estimation is an important component of movement estimate and consciousness understand field too.This paper presents a fast approach to estimate the pose of human face from a monocular image, combining with the Chinese's appearance morphologic characteristic discovered by human appearance morphologic statistics, and cooperating with the theoretical foundation of projective geometry. It mainly includes:1) Definition of computer-assistant appearance morphologic measure model. Based on the knowledge in anthropology and morphologic, analyze the characteristics that pose estimation needed, ensure feature points involved, and form the calculate model. The implement of pose estimation is based on this model.2) Feature points detection. Research the method that detects feature points of face in the photo. Preprocess of 2D-Photo, including image boosting, image scaling.3) Statistic analysis and discovery of Chinese's appearance morphologic characteristic. Inputting a large number of facial frontal view photos, calculating out geometry distribute characteristic of peoples separately, and analyzing them, finally getting appearance morphology characteristic that support pose estimation method.4) Research of pose estimation method. The bases o.f method are the principle of monocular imaging, projective invariants and rules from morphologic.5) Besides theory research, we also built a pose estimation system, which named PoseEstimate, using Visual .Net and Direct3D on the PC platform. System use the method mention in paper, the result of the test has verified its feasibility.