用于移动终端的低复杂度快速运动估计算法

LOW-COMPLEXITY FAST MOTION ESTIMATION ALGORITHM FOR MOBILE TERMINALS

针对移动终端硬件计算能力不足的问题,提出一种高性能、低复杂度的运动估计算法。算法根据相邻块运动矢量的分布特性,判断运动趋势的可预测性,进而自动选择不同的混合搜索策略;再结合自适应的提前终止阈值,在保证搜索准确性的同时,大大减少了搜索点数。通过对各种类型的视频进行测试,结果表明,该算法对于不同运动类型、不同运动程度的视频序列都具有较强的自适应性,分别比UMHexagonS(Unsymmetrical-Cross Multi-Hexagon Search)和BBGDS算法节省55%和35.6%的运动估计时间,同时率失真性能保持不变。

Aiming at the shortage of computation capacity of mobile terminals hardware,we propose a motion estimation algorithm with high-performance and low-complexity. The algorithm judges the predictability of movement trend according to the distribution characteristics of the motion vector of adjacent pieces,and then automatically selects different hybrid search strategy; further combining the adaptive early termination threshold,it greatly reduces the number of search points while ensures the accuracy of search. By testing the videos in various types,the results show that the proposed algorithm has strong adaptabilities for the video sequences with different motion types and motion extents. It saves 55. 0% and 35. 6% of motion estimation time than UMHexagonS and BBGDS respectively,at the same time the ratedistortion performance keeps unchanged.