Proposes an image registration algorithm of FAST corner detection with the description of the characteristics of SIFT. First applies FAST to extract image feature points, then uses the ring structure operator for SIFT description of the feature points, finally conducts coarse matching of the points by means of K-D algorithm, and pre screening through disparity gradient and purification with RANSAC algorithm achieves feature points matching. The experimental results show that compared with SIFT algorithm, the improved SIFT algorithm reduces the number of false matches and improves the accuracy and stability of matching.