基于SIFT的图像配准方法

针对大尺度图像配准和不同传感器图像配准问题，介绍了一种基于SIFT的图像配准方法。首先提取图像中适应尺度变化的不变特征点，在提取过程中加入多尺度Harris检测算子，提高了匹配点对的重复率，通过聚类和归一化互信息准则对候选匹配点对的角度、尺度和位置特征进行迭代筛选，删除错误的匹配点对，最后得到正确的匹配点对，对图像进行配准。实验结果表明：该方法能处理相似变换的图像配准。

Image registration approach based on SIFT

To resolve the large scale and multisensor image registration, an improved method based on scale invariant features transform(SIFT) is proposed. First, the scale invariant features of images are extracted, and a multi-scale Harris corner detection operator is added in the process, which increases the repeatability of matching point pairs. Then, after deleting the false matching points by clustering and normalized mutual information(NMI) for the rotation angle, scale and position of the candidate matching point pairs, the correct matching points are found. Finally, through the resolution equations formed by correct matching points, the image registration can be finished. Experimental results show that the method can deal with similarity transform in image registration.