计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (3): 185-189.

• 信号处理 • 上一篇    下一篇

UMHS运动估计算法的优化研究

梁  静,张  凡,唐作其,陆安江,张正平   

  1. 贵州大学 计算机科学与信息学院,贵阳 550025
  • 出版日期:2014-02-01 发布日期:2014-01-26

Research on advanced UMHS motion estimation algorithm

LIANG Jing, ZHANG Fan, TANG Zuoqi, LU Anjiang, ZHANG Zhengping   

  1. School of Computer Science & Information, Guizhou University, Guiyang 550025, China
  • Online:2014-02-01 Published:2014-01-26

摘要: 运动估计技术是H.264的关键技术,其消耗整个编码时间80%左右。提出了一种更加合理的模板策略,该策略可以更快地接近最优运动矢量;引入了运动类型判断机制以避免不必要的搜索。在UMHS算法上应用了这两种策略,保持了原有图像质量的同时有效地节省了运动估计时间。实验结果显示,优化后的算法与原UMHS算法和EPZS算法相比,在重建图像质量和码率接近的情况下,运动估计时间分别节省了约30%和15%,特别是当视频内容比较剧烈时分别节省了约40%和25%。

关键词: 运动估计, H.264, 非对称十字型多层次六边形格点搜索(UMHS), 增强型预测区域搜索(EPZS)

Abstract: Motion estimation takes almost 80% of the overall encoding time in H.264. This paper proposes a more reasonable pattern strategy that can access the best MV faster and introduces the tactics that can avoid unnecessary search; then realizes the proposed methods on UMHS. The novel algorithm saves encoding time effectively, which maintains the video quality simultaneously. Experimental results show that the advanced one can save about 30% and 15% ME time compared to UMHS algorithm and EPZS algorithm but has no change in the reconstructed video quality and bit-rate, which even reaches about 40% and 25% ME time-saving under fast motion video.

Key words: motion estimation, H.264, Unsymmetrical-cross Multi-Hexagon-grid Search(UMHS), Enhanced Predictive Zonal Search(EPZS)