中文摘要：

基于图切分理论,提出了一种自动分割人脸的方法。对于一幅人脸图像,首先通过人脸检测算法,确定人脸图像的大致前景和背景区域;然后利用高斯混合模型建立其前景和背景区域的颜色模型;最后迭代地运用图切分算法准确地分割出人脸目标。该方法创新之处在于实现了人脸图像前景目标的全自动分割。实验结果说明了该方法的有效性。

英文摘要：

This paper presented a method of extracting faces automatically based on graph cut. For a face image, firstly, identified the approximate face region by the algorithm of the face detection. Then, built the color model for background and foreground utilizing the GMM （ Gaussian mixture model）. Lastly, extracted successfully the accurate face region using the iterative graph cut. The innovation of the method lies in that can extract the special foreground object full-automatically from a face image. Experimental results indicate that the method has application values.