Font Size: a A A

Research On Face Tracking Approach Based On Skin-color And Template Matching

Posted on:2008-05-02Degree:MasterType:Thesis
Country:ChinaCandidate:X L LiuFull Text:PDF
GTID:2178360215978973Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Nowadays,the human face researching received more and more attention for the increasing demand of commercial and security application, and the development in many fields such as the Human-Computer Interface(HCI) application and the biological authentication and the object detection technology. Besides technology of human face recognition, detection and segmentation developing, the human face tracking technology which is applied in processing the video images becomes more and more popular,and becomes the main content in human face researching. While video media develops, the human face tracking technology is revealing the broader prospects comparing with other face processing technologies. The normal method of face tracking is often to detect the human faces in one or some frames first, as the initial face positions, and then find the faces in other frames using the moving object tracking technologies. The tracking result can be the foundation of the following segmentation and the human authentication.This article summarized the basic concept and the principle of human face detection and tracking technology first, and then analyzed and compared the existing theories of face detection and tracking technologies.According to the database, we proposed a highly effective human face tracking algorithm which is based on the YES color space and the inter-frame differential theories. In this algorithm, we process the first several frames of the video, map the RGB color space to the YES color space, and locate the skin color regions while synthesize the result of moving detecting process according to the difference between frames. Next use multi-templates to search the human faces in the moving regions. Then, we detect the positions of human faces in the following 200 frames,as the content of the frames are consecutive. Based on the proposed algorithm, we carry on the simulation experiment on the computer focused on the threshold determination of the YES color space, the removal of non-face regions by the templates matching and the shifting between frames.The experiment result indicates that, this human face detection algorithm based on the YES space may separate the face and the non-face regions fast and effectively. We can expect the certainly brighter, highly efficiency, broader future of the face tracking technology.
Keywords/Search Tags:Face Tracking, YES Color Space, Inter-Frame Differential, Template Matching
PDF Full Text Request
Related items