Abstract: A lot of colors similar as skin color exist in nature，they bring many difficulties to detect human face color.This paper presents a skin color detecting method that it combines circular clustering distribution of Cg-Cb color space with a parallelogram clustering distribution of Cg-Cr color space.The experimental results show that under complicated background,this method is simple and the velocity is fast，it also can wipe off most of the non-skin color and reserve the skin color perfectly.

摘要： 自然界存在很多和肤色相似的颜色，给人脸检测带来了很多的困难。提出了结合Cg-Cb颜色空间的圆形肤色聚类分布和Cg-Cr颜色空间的平行四边形肤色聚类分布的肤色检测法。实验结果发现在人脸处于复杂背景的条件下，该方法简单易行，检测速度快，能去除大量的非肤色，并且极好地保留了肤色。