Face Detection Color Image Of The Physical Model

With the development of network information technology, secure identity identification technology becomes key problem in many systems. Because of non-contact, less investment and easy to operate, face recognition has been become a hot research direction of the identity identification. And face detection as the primary and crucial link of face recognition has also got the high attention of people naturally.Automatic face detection is a relatively difficult task. The main difficulty is that the human face is a high degree non-rigid target. And face detection will be effected many factors as postures, facial expressions, and angles. In addition, the changes of illumination conditions are crucial factors of effecting face detection. Therefore, this is why there are so many methods of researching face detection in recent decades. These methods can be summarized into three categories:methods based on geometrical characteristics, methods based on statistical theory and methods based on skin color models. The general methods will be effected by large amount of calculating, the change of illumination conditions, expressions, poses, angles, etc. Human skin color is different from other objects of the nature, and the skin color of different ethnic people has clustering in a certain range, and the main affecting factors of skin color of the same race people are light intensity not but chrominance. And researchers found, the colors play very important role in face detection and recognition. Whether it is from the existing algorithm performance or the space for improvement in the future, face detection and recognition based on the color image are superior to grayscale or brightness image. Therefore, this paper embarks from the color image, face detection and recognition of the color image were studied, the main work and content are as follows:(l)First,we analyzed and summarized the development status and basic methods of face detection and recognition.(2)Secongd,we analyzed of several commonly used color space of face detection, taking into account skin color only related with the chrominace and not related with the luminance, we selected the brightness and chrominance detachable color space YCbCr. As the YCbCr color space, luminance is not entirely independent of the color information, this paper will convert YCbCr color space to YCb’Cr’color space by nonlinear transformation, and then use the elliptical model respectively to achieve face-color segmentation for the complex face images of multi-faces and simple face images of single face.(3)Based on the knowledges of medical physics, and combining with the reflection and absorption of human skin and hair about light of medical physics is unique. In this paper, we built the physical model and combined with the three primary colors of RGB color space, through reducing and eliminating the effect of illumination, the skin and hair from the background are segmentation.(4)In the stage of face detection, using the geometric features of facial organs, the human face was divided into three regions of the left eye, right eye and lip.Through locating the position of these three regional centers, the position of face was found. It reducd the detection time and improved efficiency.(5)Because the face detection methods based on geometric features of face organs are effected by many factors as facial expressions, gestures, ornaments and angles. According to the positional relationship of the face and hair, we defined a blur detection method of the full contained and half contained, which solved the problems. Through the experiments, good test results are achieved.