运用运动矢量的中心偏向性改进三步搜索算法

Improving Three Step Search Algorithm According to Central Preference of Motion Vectors

在视频图像压缩算法中,运动估计的目的在于在参考帧中寻找最相似的匹配块,使匹配后的残差最小。由于目前的运动估计算法仅适用于平移运动,小残差的运动匹配往往发生在搜索窗口的中心位置附近,快速的三步搜索算法没有考虑到这些情况,文章就此对其改进,提出了改进的三步搜索算法,实验结果证明了改进的合理性和有效性。

The purpose of motion estimation in video encoding is to search the best fitness of the block with little residue,As the ME only performs well with the displacement parallel to the image plane,which usually happens proximately at the points near the center of search window,the nature of this central bias property should fully be considered in ME process.The paper proposes an improved algorithm of the Three Step Search based on the above opinion. The experiment results have proved its efficiency.