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

作　　者：金智鹏[1] 郁梅[2] 

机构地区：[1]嘉兴职业技术学院信息技术分院,浙江嘉兴314036 [2]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《计算机应用与软件》2014年第4期223-226,共4页Computer Applications and Software

基　　金：浙江省自然科学基金项目(Y1101240);浙江省教育厅科研计划项目(Y201225496)

摘　　要：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.

关 键 词：移动终端 运动估计 提前终止 混合搜索 自适应 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]