Abstract: There are some corners in motion regions, resulting in low executive efficiency and precision, when using corners to compute the motion estimation. For solving this problem, motion estimation algorithm based on non-motion region partitions is proposed in this paper. The reference frame image is divided into non-motion regions and motion regions by grid method. Corners of non-motion regions are selected in the reference frame image, and matched with corners of the current frame image. After removing the mismatch of corners by using distance invariant criterion, motion vector is estimated by residual corners. Experimental results show that the algorithm can improve the precision and executive efficiency.

Key words: corner, region partition, motion estimation, grid method

摘要： 针对利用角点进行运动估计时，存在一定数量的运动区域角点，导致估计速度慢、精度低的问题，提出了一种基于非运动区域划分的运动估计算法。算法采用栅格法划分参考帧图像上的非运动区域与运动区域；对参考帧图像上提取的角点进行筛选，利用非运动区域的角点与当前帧图像上角点进行匹配；根据距离不变准则去除误匹配后，估算运动矢量。实验结果表明，该算法有效地提高了运行效率和估计精度。

关键词: 角点, 区域划分, 运动估计, 栅格法