Real Time Simple Face-Tracking Algorithm based on Minimum Facial Features

Abstract

This paper presents an algorithm to detect the human face from a video file using the knowledge-based approach and track the motion of the individual by detecting face in each frame. The algorithm detect human face by the geometric correlations between location of face and hairs in each frame of a video file. Range of skin color are used to figure out possible face regions so as to initially localize the face, furthermore, the probable hair blocks in an image frame are determined by means of hair color spectrums. Combined skin and hair blocks decide candidate face areas in light of the geometric relation. The phase correlation motion estimation algorithm is used to analyzing successive frames in a video sequence to identify faces that are in motion and track the human faces from the video file.. The accuracy of single-face tracking is higher than 90% with the frame rate of 10fps. Several algorithms have already been proposed and developed for various applications and employed successfully. But, those algorithms are quite complicated and hard to meet the realtime requirements of specific frame-rate. Therefore, the proposed is able to be expectedly transplanted to an embedded system, like the developing pet robot so as to perform dynamic face detection and tracking. The algorithm can be used for surveillance. The algorithm can be used for developing secure PC camera and web camera. The algorithm is being used for providing laptop security. General Terms Image Processing, algorithms.