基于平行直线组的可见光与红外图像配准算法

为实现可见光与红外图像配准,方便维修人员辨识电路板红外图像中的器件,提出一种新型的基于平行线组的可见光与红外图像的自动配准算法。首先,使用Canny算法提取图像边缘,并使用8邻域链码对其进行编码;然后,利用Freeman准则进行直线检测并提取平行线组;最后,利用图像中的平行线经仿射变换后仍为平行线的约束特性,将直线配准转换为点配准,利用鲁棒点匹配算法实现平行线间的配准。实验表明,当2幅配准图像中包含丰富的平行线时,算法能够充分利用图像中平行线在仿射变换时的平行约束性,实现可见光与红外图像的快速自动配准,且算法具有较好配准精度。

Fast aglorithm of visible and infrared image registration using parallel lines

To match the visible and infrared images and make it convenient for maintenance staffs to recognize the components in infrared image, an novel automatic registration algorithm between the visible and infrared images based on parallel lines was proposed. Firstly,Canny algorithm was used to extract the edge of the image and the edge encoded by eight fields chain code. Secondly, Freeman guidelines were used to detect the lines and the parallel lines extracted for registration. Finally, because the parallel lines remain parallel lines after the affine transformation,line registration was changed to point registration. The robust point matching (RPM) algorithm was used to match the parallel lines. The experimental results show that the proposed method can make full use of the parallel lines that remain parallel after affine transformation and realize automatic registration between the visible and infrared images quickly and efficiently.