Skin color detecting based on clustering in YCgCb color space under complicated Background

As everyone knows nature exists a lot of colors similar as skin color, such as sand, sand beach, soil, stone, cement road, wood, bole, filemot leaves, and so on. The above mentioned things are similar to skin color and exist largely in our images. So they bring many difficulties to detect human face color. And the problem hasn't been solved completely. In this paper we present a circular fitting of distribution of Cg-Cb color space. And we combine a red Gaussian model in Cg-Cb color space aim to make the method of circular model more ideal. The experimental results show that our method can wipe off most of the non-skin color and reserve the skin color perfectly under complicated background than many former methods.