复杂背景彩色图像中的人脸分割技术

Technology of Face Segmentation in Image with Complex Background

主要介绍了如何在含有复杂背景的彩色图像中进行人脸分割的技术。首先根据人的肤色在色度空间的聚类特性,寻找合适的色彩空间建立肤色模型。计算图像中所有像素的颜色与肤色的相似度值,并将其转换成0到255之间的灰度值,以得到图像的相似度灰度图,接着通过大津阈值分割进一步得到二值图,将人脸肤色区域与背景分离,从而定位了人脸区域。最后,利用人脸的结构特征,进一步缩小人脸区域范围,最终找到包含人脸主要特征部位(眼睛、鼻子、嘴巴)的人脸区域。

The technology of face segmentation in color image with complicated background is introduced in this paper. First,an appropriate color space must be found to establish the human skin color model based on the principle that human skin color clusters in color space.Calculate the similarity of every pixel＇s color and skin color,then convert it to grey level which is from 0 to 255.By this way the grey level image indicating the similarity is created.Second,to separate the face area from the image,maximum variance between two classes is provided to get the binary image. Finally ,using the property of configuration of the face ,we reduce the rang of the face area and get the face region.