The Research Of Face Detection In Complex Scenes Based On Colorful Image

With the development of digital technology and Internet, Recognition and detection of human faces has been widely used in such fields as personal identification recognition .human-computer interaction ,etc .At the same time , as the first step of human face recognition and other face applications , human face detection is very important. so human face detection has become a critical issue in the research fields of computer vision, pattern recognition.In this thesis , a novel methods of face detection in complex scenes based on colorful image is proposed. After analyzing common methods in human face detection based on complexion and template matching, this thesis brings forward a reasonable and effective methods on complexion pick-up . It proves that this methods reduces much time . In the end, the thesis puts forward a novel average face template based on edge detection which is used to match with the binary image reserving the details of human faces. It greatly reduces the influence of complex scenes. At the same time , for resolving the question of detecting obstructed and overlaped human faces, we bring forward a algorithmic mend. First of all, blocking the detected human faces. Secondly, matching the resulting binary image with the half template which has been put forward before. So we can detect the remaining obstructed and overlaped human faces. It is proved that the algorithmic mend is effective. The detect-rate of algorithm reaches 90.6% through detecting human faces on 30 test images. Except consuming more time, this novel methods is prior to the common methods based on complexion and template matching.In addition, the thesis probes into the sex detection, but the result is not satisfying. It is proved that we can't find suited threshold to distinguish man and woman through the methods of using female average eigenface.