• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

基于UMHexagonS的运动估计算法优化

彭磊,彭勇,吴锡生,徐赟,胡旻涛   

  1. (江南大学物联网工程学院,江苏 无锡 214122)
  • 收稿日期:2015-06-08 修回日期:2015-10-20 出版日期:2016-10-25 发布日期:2016-10-25
  • 基金资助:

    江苏省产学研联合创新资金(BY2013015-35)

An optimized motion estimation algorithm
based on UMHexagonS

PENG Lei,PENG Yong,WU Xi-sheng,XU Yun,HU Min-tao
  

  1. (School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)
  • Received:2015-06-08 Revised:2015-10-20 Online:2016-10-25 Published:2016-10-25

摘要:

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%。

关键词: UMHexagonS, EPZS, 视频压缩, 运动估计

Abstract:

Using the results of the Grand Cross search template, we propose an optimized motion estimation 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.

Key words: UMHexagonS, EPZS, video compression, motion estimation