ABSTRACT:As an important visual pattern, human face reflects the visual information which plays an important role in human communication and interaction. Human face detection is a key technology in face information processing and its application field is far beyond the face recognition system. Face detection has been increasingly a major concern by the researchers because of its huge value in video supervision, access control, video conference and human-computer interaction, etc. As a key part in face recognition, face detection research also shows the academic value of itself, at the same time, variety of face detection algorithm comes to appearance as the research methods and tools are coming to maturity increasingly.The article studies the face detection method based on skin-color and AdaBoost algorithm, which includes the following aspects:First, studying the face detection algorithm based on skin-color. By comparing the common color space and color models, we choose the YCbCr color space with better clustering performance and Gaussian model which is easy to carry out to describe the facial skin. We have achieved good color segmentation result by researching image processing techniques to satisfy the need of color segmentation, image binary based on the threshold segmentation, mathematical morphology processing method and further removing non-facial area method using of color size and aspect ratio.Second, studying the face detection algorithm based on AdaBoost. We discuss the 3 parts which are made up of Haar-like features, integral image and cascade classifier based on AdaBoost algorithm. Aiming at the weak ability problem on single Haar-like feature classification, we preliminary study the combination Haar-like features. For the large number of Haar-like features and long-term training, we use the optimization method of the characteristic value of the quantitative to shorten the training time.Finally, realizing the face detection method based on skin color segmentation and AdaBoost algorithm. Input the image as the cascade after the color segmentation to increase the detection speed and detection rate.The article summarizes the paper at last and looks into the further work. |