关键词: 图像配准, 尺度不变特征变换（SIFT）, 自适应阈值, 随机抽样一致（RANSAC）, 分层区域

Abstract: Aiming at the problem that SIFT（Scale Invariant Feature Transform） algorithm extracts feature points that are impure and susceptible to noise, it is first proposed to preprocess the image before SIFT algorithm extracts feature points to eliminate some external interference. Secondly, for the 128-dimensional high-dimensional descriptor in SIFT algorithm, the matching speed is reduced. A feature descriptor based on the hierarchical region is proposed to reduce the descriptor sub-dimension and shorten the running time of the algorithm. Finally, aiming at the problem that the fixed threshold does not have wide applicability in the SIFT algorithm matching process, an adaptive threshold method is proposed to solve the problem that the fixed threshold does not match the actual image and improve the matching accuracy. The experimental results show that the improved algorithm can improve the matching accuracy and efficiency, enhance the robustness and reliability of the algorithm and has wide applicability.

Key words: image registration, Scale Invariant Feature Transform（SIFT）, adaptive threshold, RANSAC, hierarchical region