摘要： 针对全局运动估计过程常伴有局部运动的问题，提出一种基于LMedS&LS的全局运动估计算法。通过获取相邻2帧图像上对应特征块的坐标，将该算法应用到仿射变换模型中，估计出全局运动参量，并根据运动参量采用双线性内插法完成图像配准。实验证明，该算法相对于LMedS算法在精确度和实效性2个方面都有改进。

关键词: 全局运动估计, 特征块预检验, 改进型全搜索匹配算法, LMedS&LS算法, 图像配准

Abstract: For global motion estimation process is often accompanied by the influence of local campaign, this paper puts forward a global motion estimation algorithm based on LMedS&LS. It gets the corresponding characteristics of the coordinates of the piece from two adjacent frames, uses it in affine transformation model, estimates the global motion parameters, and completes image registration through the bilinear interpolation according to the motion parameters. Experiments show that compared with LMedS algorithm, it is improved in both the accuracy and effectiveness.

Key words: global motion estimation, feature block pre-test, improved full search matching algorithm, LMedS&LS algorithm, image registration

中图分类号: 

• TP391.41