基于混合彩色空间的肤色区域分割算法

Skin Region Segmentation Algorithm Based on Mix-color Space

利用肤色在RGB、YCbCr和HSV空间中均有较好的聚类性质,以及模糊隶属度函数快速构造得到图象的肤色投影图;使用Active Contour的level-set算法对肤色投影图进行分割合并得到初步肤色侯选区域;再使用K均值方法对初步侯选区域进行修正,得到同质肤色区域。实验证明算法对简单,复杂背景下的单人脸和多人脸肤色区域分割均有较好的效果。

Using the clustering of skin in color space RGB, YCbCr and HSV, with fuzzy function, we construct skin probability map（ SPM）, and use level -set algorithm on SPM to unite and segment regions to get candidate skin regions. In order to improve the validity, k - means algorithm on color image will be used to adjust the candidate re- gions, and finally get the uniform regions . The algorithm proves to be effective on one face skin region segmentation or more than one face skin region segmentation under simple and complex background.