摘要
本文提出了一种运用图像的运动信息以及皮肤和嘴唇的彩色信息实现复杂背景下人脸自动检测的方法.该方法特别适用于图像中人脸区域和类肤色区域相连接的情况.实验结果表明,该方法具有实时性好,检测和定位精度高以及对光照等环境条件不敏感等优点,可以用于实时人脸跟踪以及实时人脸识别等应用领域.
In this paper, a method for automatic face detection in complex backgound is presented. This method can precisely detect and locate faces by combining motion and color information (including skin-color and lip-color). The difference between each pixel in the two successive images in calculated, and the moving parts arc obtained. Using the distribution of skin color in HIS color space, the candidate facial fields are obtained through skin-color filtering. By this method, fields in the background with skin-like color can be eliminated effectively. For the case of combination of skin fields and skin-like fields, the second color-filtering method is proposed. That is, a lip-color model is established to get the specific color information of the candidate facial fields, and the region is filtered again to eliminate the skin-like field and get the real facial field. Experiments show the proposed human face detection algorithm is effective, fast and rubust.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第2期205-210,共6页
Pattern Recognition and Artificial Intelligence
