Abstract: Obtaining facial feature points from images are one of the fundamental research issues in computer vision and graphics. At present, most of face alignment methods obtain the position information of face feature points from a single static picture, i.e., a single still picture detection. Although these methods have obtained relatively accurate detection results, in the detection of feature points of continuous picture sequences, the feature point jitter may occur due to environmental changes such as illumination and background. Therefore, an embedded tracking system for a human face feature points is constructed using a tracking-based method for acquiring facial feature point positions from successive picture sequences. Simulation results show that this method is effective.

Key words: facial feature points detection, face tracking, embedded system, video processing