Study On Techniques Of Face Detection Based On Skin Model And Facial Feature Location

Until recently, many researchers concerned the research of face detection, that plays important role in applications such as face recognition, video surveillance, human computer interface and face image database management. But, up to now, due to the complexity of the purpose such as expression, illumination, age, pose, the quality of the images and glass, hairstyle, beard, mustaches may or may not be present and so on, many researchers can not resolve these problems completely even if they have studied it for long time. Learned from the domestic and international discourse and research papers concerning face detection and facial feature location in recent years, firstly we give a systematic summary of previous work. And we developed an automatic face detection system on the basis of the previous research on face detection.(1)The result of detection is easily influenced by illumination, so we present a new method of using "Gray World" to judge whether color warp exists in an image. If it is existent, we process light compensation. At one time, we improve on the method of "Reference White". Experimental results demonstrate successful face detection over different illumination and the missing rate(MR).(2) We establish a skin model based on the YC_bC_r color space. Using this model, the complexion is segmented and two-value processed. We can get backup face area using the filter based on mathematical morphology. And then, we utilize face geometrical characters and Euler number to roughly choose between these backup areas.(3) We find mouth candidates through the red feather of mouth in the YIQ color space and eye candidates through image complexity.(4) Besides these, we research face features algorithm, transform the relation of location to the relation of proportion and angle. Experiments prove that obviously improve the efficiency of face detection.(5) In the last part of the dissertation, we implement the algorithm using Visual C++ 6.0 and Matlab 7.0, and experiment in our face testing set. The results show that our method is robust, and strongly adaptive to different illumination, expression, variant pose, ages. In a word, hit rate is above 84 percent. The arithmetic can play in real-time detection because of its rate.