Abstract: In order to improve the image registration speed and the matching stability，a sub-pixel image registration algorithm based on comparative distance is presented.And also，theoretical deduction and the result of experiments are given.And it has been proved that the algorithm has sub-pixel accuracy，and the precision is between 0.01 and 0.1.The results of matching experiment for 12 couples images have indicated the matching accuracy of presented algorithm has achieved 100%，and the time consumed is less than the RANSAC algorithm which is in common used in present.

摘要： 为了提高全景图全自动拼接中图像配准的速度和匹配稳健性，提出了一种基于相对距离法去除外点的亚像素级图像配准算法，并给出了分析和实验结果，亚像素级像素定位误差在0.01~0.1之内。在对12组图像匹配的实验结果表明，该匹配算法的匹配正确率达到100%，且匹配的时间小于目前通用的RANSAC匹配算法。