基于头肩图像中脸部轮廓的运动估值

对于可视电话中的头肩图像，在提取出脸部轮廓后，在脸部区通过搜索出全局的二维运动参数来确定各像块运动矢量的初始搜索点。在此基础上搜索精确的运动矢量时，搜索范围很小，搜索结果更符合真实运动，保证了运动矢量场的一致性。因此，既加快了运动估值的速度，又节省了编码运动矢量所用的比特数，使图像的质量有所提高。

Motion Estimation Based on the Contour of Face in Head-and-Shoulder Image

In this paper,a new method of motion estimation for head and shoulder image in videotelephone is proposed After extracting the contour of the face,a parameter of 2 D global motion is calculated to get the initial searching point of motion vector for all blocks in the face The motion estimation can be accomplished within a small range around the initial searching point for each motion vector The accuracy of motion estimation and smoothness of motion vector fields are improved The experiment results show that the proposed method spends less time for the motion estimation and provides higher image quality