JM模型是JVT（jointvideoteam）发布的H．264标准测试模型，对算法学习和研究有着重要的意义。根据JM测试模型的参数设定，其中的运动估计算法有3种可选模式。本文结合JMIO．2的源代码对UMHexagons算法进行了分析，并对该算法进行改进，能够在保证视频序列各分量信噪比的情况下缩短运动估计的耗时。本文利用UMHexagons算法的准确预测以及运动估计代价的相关性来设置阈值达到提前结束搜索的目的。在JMI0．2的测试模型上进行了算法验证。实验结果表明，利用块与块之间运动估计代价的相关性，在保证编码性能的同时，可以减少运动估计所需时间的10％以上。

JM is the standard testmodel published by joint video team(JVT), which has great significance in algorithm learning and research. In JM, there are three optionalmodes by setting parameters differently. Based on the study of the principle of the UMHexagonS algorithm, an improved method is proposed to reduce the time ofmotion estimation with minimal loss in bitrate and reconstructed quality compared with those of the originalUMHexagonS algorithm. Correlation between adjacent blocks is used to end the search. In JM10.2, results show that the improved UMHexagonS algorithm can reduce the time of motion estimation with the original performance of the original UMHexagonS algorithm.