一种基于相机标定的图像配准方法

Method of Image Registration Based on Camera Calibration

图像配准是把同一场景的两幅或者多幅图像在空间上进行配准,它在图像分析领域应用广泛,如医学,遥感图像分析,图像融合,图像检索及目标识别等.匹配困难是图像配准的主要问题,造成匹配困难的原因之一是待匹配图像存在畸变.该问题是由于待配准图像相对于基准图像会产生大面积相同区域或产生平移、旋转、缩放等多种畸变.针对以上各种畸变,本文提出一种解决方案:利用相机标定信息求解畸变情况的平移矩阵、旋转矩阵和缩放比,通过反变换达到矫正图像畸变的效果.然而标定信息常会因为相机位置移动而发生改变.对于此问题,本文则是通过求解多角度下的云台转角来确定相机的标定信息.实验表明:本文提供的方案很好的解决了图像配准阶段的一些畸变问题,使图像配准有较高的精度.

Image registration is to register two or more images in the same scene in space. It＇s widely used in the field of image analysis such as medical science, remote sensing image analysis, image fusion, image retrieval and target identification. A main problem of image registration is difficulty in matching, which is caused by a lot of image distortion in original images. Compared to the reference image, the problem is caused by the floating image produces large areas of the same color or various distortions such as translation, rotation and zooming in the image. For all of these distortion problems, this paper proposes a solution, which uses the camera calibration information to get the translation matrix, the rotation matrix, and the scaling of the distortions, and corrects the image distortions by inverse transformation. The camera calibration information, however, often be changed by moving the camera position. To solve this problem, this paper is by getting the PTZ Comer in the multi-angle to determine the camera calibration information. Experiments show that this paper provides a good solution to solve the distortion problems in the stage of image registration and improve the accuracy in the stage of image registration.