Realization On Color Face Detection Based On Skin Colour Segment And Statistical Methods

With the social and technological development, pattern recognition and computer vision in the current society have increasingly widespread, especially after the 9.11 incident in the United States, the use of personal identification not only in banking and finance, e-commerce, companies attendance and family security ,The field, more elevated to a national security point of view. At present, the main characteristics applied to the identification of human have a fingerprint, iris, face, palm prints, and so on. Face as a biological characteristic and because of its intuitive, simple, informative and the advantages of convenience, more and more researchers aroused the concern about the face detection. At present, the applications based on facial characteristics include automatic face recognition, video surveillance, Facial Expression analysis, video conferencing and human-machine interactive systems. in all the applications, the first problem we must solve is to face detection , It will directly impact on the performance of the whole system.In past years, because of face detection in the real application has widespread, it has attracted more and more attention to scientific research workers, has developed into an independent research direction. As people give the more in-depth studies on the face detection, a variety of practical face detection algorithm have proposed.In this paper, we focused on the merit about existing algorithm, and take into account the fact of the popularity of color images in real-life applications, to achieve a face detection technology by using the combination of skin color segmentation and statistical classification. Firstly getting all possible candidate face regions by using color segmentation from the input image, then using the classification based on statistics for all candidates regions to confirm, thus achieving the face detection. Specific content include:1,give the introduction about skin color segmentation technology in detail, and do a detailed comparison about the performance using different skin color model under different skin color space, by the analysis of advantages and disadvantages about the color the Gaussian skin color model and the simple skin color model, imposed a improved algorithm by using a combination the two above model, which takes into account the efficiency and accuracy.2,implement the two current popular face detection methods - support vector machines (SVM) and AdaBoost algorithm. And by combination with skin color segment to achieve color face detection. For SVM, give the performance comparisons under different features (pixel features and pca characteristics).3,impose a new face detection method based on newest statistical classification-minmax probability machine and minimal error minmax probability machine , by the combination with the skin color segment to achieve color face detection. Similarly in SVM give the performance comparisons under different features.