一种新的快速块运动估计的分级筛选法  

A New Fast Multi-Level Elimination Algorithm for Block Motion Estimation

作　　者：陈韩锋[1] 戚飞虎[1] 

机构地区：[1]上海交通大学计算机科学与工程系,上海200030

出　　处：《计算机学报》2004年第6期828-832,共5页Chinese Journal of Computers

基　　金：国家自然科学基金 ( 60 0 72 0 2 9)资助

摘　　要：提出了一种新的分级筛选法用于快速的块匹配运动估计 .该方法中最佳匹配块通过若干个筛选过程得到 ,每个筛选过程用不同大小级别的子块作为基元进行匹配 .在初始的级别中用很少的运算代价通过简单的特征匹配先淘汰一部分候选匹配块 ;然后在上一级剩余的候选块中 ,逐级用更加细致的特征继续筛选 ;直至找到最佳匹配块 .对不同运动复杂性的视频的实验表明 ,在搜索范围为 (- 7,7)× (- 7,7)的情况下 ,分级筛选法不仅在速度上比一些常见的快速运动估计方法要快 ,而且估计精度与全搜索法非常相近 .A multi-level elimination algorithm (MLE) for fast motion estimation is proposed. The optimal matching block of current block is detected in the next frame by several elimination levels. In different levels, sub-blocks of various sizes are adopted as basic elements in block-matching. In the first level, some candidate blocks are eliminated by simple characters based block-matching, needing only very few computation costs. Then in further levels, more detailed characters are used for block-matching to eliminate candidates until the most matched block is selected from all the candidates in the last level. Experiments on videos of various motion show that the proposed algorithm is faster than usual fast motion estimation algorithms and can get very close performance to the full-search algorithm when the search area is set as (-7,7)×(-7,7).

关 键 词：分级筛选法 块匹配运动估计 特征匹配 视频编码 视频分析 

分 类 号：TN919.81[电子电信—通信与信息系统]