基于局部结构特征的红外与可见光图像匹配

摘要

Due to prominent distributional variations of grayscale between multi-source images, traditional image matching methods rooted in grayscale correlations exhibits limited effectiveness on infrared and visible images matching problems. To avoid undependable grayscale correlations, a local structure-based infrared visible image matching method is presented. Harris corner and its neighbor contour segments oriental angles are combined to build Local Oriented Harris Feature(LOHF), affine estimation is derived based on LOHF, and feature matching is realized by point set matching hypothesis-verification combined with RANSAC robust estimation. Experimental results show the capability of proposed method when other methods are failure on infrared visible image matching task.%使用灰度特征对红外与可见光图像进行匹配时,其效果受异源图像间灰度分布差异的限制.为此,提出一种基于局部结构特征的匹配方法.提取Harris角点及其关联的局部轮廓曲线切向角度,使用角点与角度信息构造局部有向Harris特征点,推导基于该特征点的仿射估计方程,结合特征点集匹配假设检验与RANSAC鲁棒估计进行匹配.实验结果表明,在其他方法失配的情况下,该方法可有效实现红外与可见光图像的特征点匹配.