基于改进SIFT的结构与纹理相似图像配准方法

 针对目前SIFT特征匹配算法在大面积结构和纹理相似的图像应用中存在较多的误匹配，导致图像拼接效果不理想的问题，提出了基于改进SIFT特征匹配的结构与纹理相似图像配准方法.通过结合色彩信息及空间信息进行匹配点对的筛选，进而得到更为精确的匹配点对，克服了传统SIFT算法在其寻找匹配点的过程中严重依赖灰度信息下的主方向，导致匹配的误差放大的缺点.实验结果表明，该方法提高了SIFT特征匹配的鲁棒性同时，进一步改善了拼接的效果.

Image registration for structure and texture similar images based on improved SIFT algorithm

This paper proposes a structure and texture similar image matching method based on improved SIFT algorithm in order to solve the problem that SIFT features matching algorithm exist many errors because it rely heavily on gray information,which would lead to non-ideal effect of stitching.We make use of color information and spatial information to filter the matching point-pairs,so as to achieves better accuracy and overcomes the traditional SIFT algorithm’s disadvantage.Experiments show that the proposed algorithm improved the robustness and the effect of stitching.