Abstract: Motion estimation is the key to remove the temporal redundancy in video compression, the existing algorithms are mostly SAD matching algorithm based on full search strategy. The compression performance of these algorithms are very good, but the computation is complexity and it has a poor real-time. This paper proposes a new fast motion estimation algorithm. It splites a block into multiple sub-blocks, and calculates every sub-block gray value and the gray value of the sum of squares of every sub-blocks, then as a whole parameter in combination with the three proposed matching criterion, calculates the optimal frame motion estimation between the current frame and candidate. Experimental results show that using this algorithm, the motion estimation significantly reduces the complexity and compression performance is very close to full search strategy based on SAD algorithm.

Key words: motion estimation, video compression, matching criterion, block partition

摘要： 运动估计是剔除视频压缩中的时间冗余的关键，现有算法大都是基于全搜索策略的SAD匹配算法，这些算法虽然压缩性能很好，但计算复杂，实时性差。提出一种快速运动估计新算法，将块分割成多个子块，计算每个子块的灰度值之和与灰度值的平方和，将其整体作为一个参数再结合提出的三个匹配准则，求出当前帧和候选帧之间的最优运动估计。通过实验表明，采用该算法后计算的复杂度明显减小，实时性得到较大提高，其压缩性能却非常接近基于全搜索策略的SAD算法。

关键词: 运动估计, 视频压缩, 匹配准则, 块分割