基于UMHexagonS的运动估计算法优化

An optimized motion estimation algorithm based on UMHexagonS

针对UMHexagonS算法冗余搜索的问题,使用大十字搜索判定结果,改进原有的运动估计算法。改进算法判断最优点可能分布区域,使用相应改进搜索模板搜索,降低搜索点个数,达到避免冗余搜索的目的,提高运动估计搜索效率。在多组视频序列中测试,图像质量相近情况下,改进算法比UMHexagonS算法有了显著的提高。特别是在剧烈运动的视频序列中,改进算法比UMHexagonS算法的运动估计时间减少了45.78%,编码耗时缩短了34.97%,比EPZS算法运动估计时间减少了35.25%,编码耗时缩短了25.45%。

Using the results of the Grand Cross search template, we propose an optimized motion es timation algorithm to avoid the unnecessary search of the UMHexagonS algorithm. In order to avoid redundancy search, we reduce search points and increase the efficiency of motion estimation, and a certain search template is used in the area divided by the Grand Cross search template. Experimental results on different video sequences show that the optimized algorithm can save encoding time effectively and maintain video quality simultaneously. Compared with the original UMHexagonS algorithm and the EPZS algorithm, the motion estimation time of our algorithm can be saved by 45.78% and 35.25%, and the coding time can be saved by 34.97% and 25.45% respectively in fast motion video sequences.