肤色建模和肤色分割的人脸定位研究

Localization of human face based on skin module and skin color segmentation

为了实现彩色图像中人脸的精确定位,提出了一种基于肤色模型、肤色分割处理的人脸定位算法。通过建立肤色模型计算得到图像的相似度分布图,经自适应阈值的二值化处理后,再进行肤色分割,将非人脸区域去除;最终利用眼睛与嘴巴构成三角形特征结合人脸椭圆模板匹配定位人脸。实验结果表明,该算法对于复杂背景下的彩色图像中的人脸正面定位和人脸转动一定角度后定位都有较好效果。

In order to localize the human face in color images accurately, a human face localization algorithm based on skin module and skin color segmentation was presented. Firstly, the similar degree distribution image was obtained by building the skin module. Then, the non-face region was removed in color image after binary image processing with adaptive threshold and the skin color segmentation. And finally the human face was localized by using the characteristic triangle composed of the eyes and mouth to match the elliptical module of human face. Experiments show that the algorithm is effective to localize the human front face and the face after turning an angle in color images under complex background.