在研究红外图像成像原理与特征、传统红外图像配准方法的基础上, 提出了一种改进的联合点特征与灰度特征的红外图像配准算法。首先采用经典的 Harris角点检测算法提取一次角点, 在一次角点的基础上对其进行下降排序, 对排序结果进行不同份数的等分并提取每一部分的中间值, 最后根据图像配准需求有效利用中值选择不同的特征点; 该算法充分利用环形区域的旋转不变性和特征点区域灰度差异性很小的原则进行特征点匹配。实验结果表明该算法能够提取出更加精确的匹配点, 能够有效的完成红外图像配准。

On the basis of research of principle and characteristics of infrared image and the traditional infrared image registration method, this paper presents an improved infrared image registration algorithm based on point feature and gray feature. The algorithm uses the classical Harris corner detection algorithm to extract the first corner, then down ranks the first corner, divides the result into different parts and extracts the median of each part. Finally, it effectively uses median to select different feature points according to the image registration needs. The algorithm matches the feature points by making full use of therotation invariance of annular region and feature point region of which gray difference is very small. The experimental results show that the algorithm can extract more accurate matching point, can effectively complete the infrared image registration.