Human Facial Features Detection and Tracking in Images and Video

摘要

Face detection is the first important step for facial analysis algorithms whose goal is to determine whether faces or not in image or video, if present return the location and the bounding box of each face in the image. In this paper, we propose an efficient and fast multi-view face detection method in face images and videos. We first propose to adopt a color filtering based efficient region scanning method to detect face rapidly; it skips over regions that do not contain candidate faces. For facial feature detection from the detected face region, Haar-Like feature utilized along with AdaBoost as a learning algorithm that constructs a strong classifier composed of weak classifiers, the incorporating facial parts improve the detection accuracy. We adopt affine motion model estimation, which can accommodate the shape change, as a tracking method and by using the coordinates of both eyes and a mouth with the origin of a nose the head pose is estimated. Experimental results show the computational cost of our approach is low and it gives a better detection performance in factors that affect the appearance of faces such as variations in illumination, poses and facial expressions, occlusion, make-up, beard, mustache, glasses, etc. In addition to these factors, the face images are large or small and the details of face parts are clearly visible or not.