To reduce the time complexity of the traditional SIFT algorithm related to the feature points detection and matching, an optimized image registration algorithm is proposed. Namely the feature points are extracted by Trajkovic algorithm, and the SIFT algorithm distribution descriptor method is adopted to allocate feature points descriptor parameters. Then sparse dimension reduction principle is used to reduce the feature points descriptor dimensions. Finally, the similarity measure algorithm based on bidirectional matching is used to match them. Six evaluation criteria including the feature points detection of image, matching pair, correct matching pair, matching accuracy, registration time and registration time drop rate are adopted in the simulation experiments. The simulation results show that the optimization algorithm in comparison with traditional SIFT algorithm is equivalent in the accuracy of feature point matching. However, it has obvious improvement in terms of feature points matching speed.