提出了一种使用具有方向偏向性搜索模板和基于预测模式自适应选择运动估计方案的EPZS(enhanced predictive zonal search)改进算法。该算法将运动估计过程与H.264/AVC中多参考帧、多预测模式等编码工具紧密结合，充分利用视频中运动具有高度时空相关性的特点，综合考虑速度、图像质量及压缩效率等方面的性能。实验结果表明，该算法以极小的计算量得到了与全搜索方法接近的效果；与EPZS算法相比，运动估计的处理速度有了较大幅度的提高，同时保持了与之相当、有时甚至更优的率失真性能。此外，该算法具有很好的鲁棒性，并且能够产生平滑的运动矢量场。

An improved EPZS(enhanced predictive zonal search) algorithm with direction based search patterns and adaptive motion estimation scheme selection based on prediction block type is presented in this paper. Utilizing high special/temporal correlation of motions in video sequences and focusing on the interactions between motion estimation and new coding tools such as multiple references and prediction block types, the algorithm can obtain almost the same PSNR as full search at very low computation cost. Compared with EPZS, it can considerably improve the speed of motion estimation, and rate distortion performance. Experiments verify the superiority of the proposed algorithm. Furthermore, the proposed algorithm is robust and capable of generating a smooth motion vector field.