基于SIFT的红外与可见光图像配准方法

摘要

An improved method of IR and Visible Images Registration based on SIFT is proposed. Firstly; the feature points are extracted by SIFT and an constraint strategy based on SUSAN operator is added in the process ; then the feature descriptors are created. Secondly; the nearest-neighbor method based on k-d tree is used to find the corresponding matching points and the RANSAC algorithm is used to acquire transformation matrix. Finally; carrying out transformation and bilinear interpolation in the image prepared for registration to realize image registration. Experimental results show that the method is stable; reliable and efficient.%介绍了一种基于SIFT的红外与可见光图像配准方法.首先用SIFT算法提取特征点并构造特征描述子,在特征点提取过程中采取了基于SUSAN算子的约束策略,其次用基于k-d树的最近邻方法对特征点进行匹配,用RANSAC算法求解变换矩阵,最后对待配准图像做相应参数的坐标变换及双线性插值,从而实现图像配准.实验表明,该方法具有稳定、可靠、快速等特点.