一种适于实时应用的快速有效的自适应运动估计算法  

A Fast and Adaptive Motion Estimation Algorithmfor Real-time Application

作　　者：董海燕[1] 张其善[1] 

机构地区：[1]北京航空航天大学电子信息工程学院202教研室,北京100083

出　　处：《计算机工程与应用》2005年第7期1-2,197,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(编号:60372018)

摘　　要：运动估计是视频编码系统的关键部分,同时也是整个视频编码器中计算量最大的部分,运动估计性能的优劣直接影响到整个视频编码器的运行效率和整个视频序列的重构质量。该文提出了一种适用于实时应用的快速自适应运动估计算法,该算法首先利用运动矢量时空域的相关性预测初始运动矢量,然后对不同运动程度的图像块进行不同大小的菱形搜索,同时采用了高效的提前截止准则。实验结果表明,对于各种不同运动程度的视频序列,该文算法都可以获得和全搜索法基本相同的图像质量,同时可以节省超过99%的计算量,该文算法在搜索速度和搜索精度两方面比快速算法中性能突出的菱形搜索具有更大的优势,更适合实时应用。Motion estimation is an important part of video compression systems,since it could significantly affect the operation efficiency of whole video coding systems and the reconstructed quality of video sequence.This paper presents a fast and adaptive motion estimation algorithm for real-time application.The algorithm predicts motion vectors with the help of high correlation between adjacent blocks and uses diamond pattern with different size to blocks with different motion activity hereinto effective stopping criteria is adopted.The experimental results reveal that,to all kinds of video sequences,the proposed algorithm can obtain similar video quality to FS algorithm with reduced 99% computation cost.The proposed algorithm possesses more superiority than DS on speed-up and search accuracy,so it fits real-time applications.

关 键 词：视频压缩 运动估计 块匹配 菱形搜索 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]