为了提高图像配准精度，本文提出了基于改进的 SIFT(scale invariant features transform)和 RANSAC(Random Sample Concensus)的图像配准算法。该方法首先采用 SIFT算子进行特征点提取和特征描述；其次采用改进的RANSAC阈值参数设置算法去除误匹配点；最后对提取的特征点进行分布均匀化，以进一步提高估计精度。基于实际拍摄图像序列的实验证明了本文所提算法的有效性和鲁棒性。“,”To increase the accuracy of the automated image-pair registration, a new registration method based on SIFT and RANSAC is proposed in this paper. Firstly, key points and features from image pairs are extracted based on SIFT. Secondly, a new version of threshold setting method based on RANSAC is proposed to remove mismatched points. Finally, a new feature points’ distribution uniformalization method is used to refine the estimation. Experimental results on captured image pairs show that the proposed algorithm is effective and robust.