人脸是一个非刚体 ,其运动分割一直是一个难点 .在首先建立色度和象素点运动矢量幅值一致性的区域分割的基础上 ,通过运用考虑到模型失败的运动模型图像分割技术 ,从而得到含人脸的前景分割 .新的算法可以准确地将包含人脸的前景与背景分开 .实验结果显示了算法的有效性 . Human face is a non-rigid body, and its motion segmentation has been a difficult point.On the basis of first establishing the region segmentation of the consistency of chrominance and pixel motion vectors, by using the motion model image segmentation technology which takes into account the failure of the model, So the foreground segmentation with face is obtained.The new algorithm can separate the foreground from the background that contains the face accurately.Experimental results show the effectiveness of the algorithm.