一种精确而快速的块匹配算法  

Accurate and Fast Block Matching Algorithm

作　　者：杨志云[1] 郝红卫[1] 陈骥 

机构地区：[1]北京科技大学信息工程学院,北京100083 [2]微开半导体研发(上海)有限公司C-Model Group,上海200233

出　　处：《计算机工程》2008年第2期223-225,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60475003,60675006);北京科技大学科学研究基金资助项目

摘　　要：为提高基于块的运动预测的搜索精度,该文基于块匹配准则,提出最小绝对差分误差和(SADD)算法。SADD以SAD为基础,利用块匹配中相邻像素点误差之间相关性,将块内各像素误差按一定顺序进行差分,消除了相邻点误差之间的冗余。该算法更精确地表达了块与快之间的相似程度。实验结果表明,该算法可有效降低码率,提高PSNR值,而且其鲁棒性好,可适用于各种序列。In order to improve the search accuracy performance of motion estimation based on block, this paper researches the block matching algorithm, brings forward a new matching algorithm named the minimum Sum Of Absolute Different Differences（SADD）. Based on SAD, the algorithm SADD takes advantage of the relativity in the adjacent differences inside one block, adopts the difference operation in a certain order to remove the redundancy, which enables the algorithm SADD to describe the similarity between two blocks much more accurately. Simulation result shows that this algorithm can reduce the bitrate effectively and improve the PSNR performance well. And it can be adapted to kinds of video sequences.

关 键 词：运动补偿 运动估计 块匹配 绝对差分误差和 视频压缩 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]