Image stitching has become an important part of image processing and an increasingly popular research direction. Image registration is the core of image stitching. In order to overcome the performance imbalance of traditional algorithms, an improved image registration algorithm is proposed in this paper. Firstly, the algorithm detects the feature points based on Hessian matrix, and then describes the feature points by rBRIEF method. Finally, in the matching stage, the nearest neighbor ratio and next nearest neighbor algorithm are used for a rough matching, and the RANSAC algorithm is used to confirm the rough matching point pairs to eliminate false matching. The test results show that the improved surf image registration algorithm has certain efficiency and robustness.