人脸视频图像编码是近年来图像编码领域里的一个研究热点问题,它在通信、互联网等方面有着广泛的应用前景。人脸图像编码的研究包含很多子问题,主要的三个方面为:精确的人脸目标定位算法,实时的人脸目标跟踪算法和高效的人脸图像编码方法。本文的主要研究工作在于: 1) 提出帧间差分和背景差分相结合的人脸目标定位算法,在人脸特征选择上,主要使用形状特征,通过椭圆拟合来得到人脸的位置。 2) 采用基于Kalman滤波的运动预测方法对人脸的区域变化进行预测,以此缩小目标可能存在的区域(ROI).-Facial video image coding image compression in recent years a hot research topic in the field of the problem, it is in the communication, the Internet have a wide range of applications. Coding of face images contain many sub-problems, the main three are: accurate target location algorithm for face, real-time face tracking algorithm and efficient facial image coding method. This major research work are: 1) of the frame difference and background difference goal of combining face localization algorithm, feature selection in the human face, the main use of shape features, by ellipse fitting to get the location of the face. 2) Kalman filtering based motion prediction method changes the face of the region predicted to narrow the possible target area (ROI).