Since the matching result of scale invariant feature transform(SIFT) algorithm has many false matching points, a image matching algorithm based on SIFT descriptors with double scales and limited searching region(DSLSR SIFT) is proposed. The method uses double scales descriptors to calculate the preliminary matching points set, then local limited searching region condition is added in the preliminary matching points set to remove the points which are far away from the limited region condition and obtains refined matching result. Finally, the random sample consensus (RANSAC) algorithm is used to evaluate the matching results. Experiment results show that the correct rate of this algorithm is better than SIFT algorithm on average of about 17%, which significantly improves the matching accuracy.
