在极低码率视频编码运动估值过程中，针对视频对象的边缘形状特征，将宏块按不同形状的模板自适应地进行分割，同时引入近期、远期2个参考帧进行块匹配。为提高搜索速度，首先判断出不需编码的宏块，并引用了一种快速搜索算法。实验结果表明，与H.263先进预测模式相比，采用所建议的运动补偿方法，图象的信嗓比得到明显的提高。

In the process of motion estimation for very low bit rate video coding, a macroblock is partitioned adaptively with a pattern which matches the shape of edges of the objects in the macroblock, and two time differential reference frames are used. To speed up the process of motion estimation, the macroblocks which need not to be coded are distinguished before the process, and a fast search algorithm is used during the process. Simulation results demonstrate that, compared with the advanced prediction mode in H.263, the proposed method for motion compensation makes higher signal noise ratio (SNR) for the reconstructed image.