Research Image Mosaic Algorithm Based on Improved SIFT Feature Matching

Based on traditional SIFT feature point matching method and image pixel level fusion, an improved SIFT feature matching algorithm is proposed. Firstly the algorithm in the feature extraction process with Harris corner detector and picks out robust points; Followed use of 48 dimensional vector based on a circular window to dimension reduction and the feature points are refined by the geometrical consistency of the matching feature points, to improve the efficiency of the algorithm. Finally, the overlapped area linear transition fusion algorithm of image smoothing the transition to eliminate the splicing gap. The experiment verifies the robustness and rapidity of the proposed method.