中国科学院云南天文台机构知识库

在太阳观测研究中,高分辨图像与全日面像的配准是一项非常有意义的工作,但由于他们之间存在着旋转、缩放和平移,因此很难高精度地进行匹配。提出一种结合局部统计信息和控制点匹配的图像配准方法,核心思想为将视场等间隔划分为大量重叠的局部区域,通过相关匹配在全日面像上寻找对应的局部区域,然后计算每一对局部区域间的亚像元偏移,根据偏移量确定每一对特征点的坐标位置,据此作为点匹配中的特征控制点;最后根据控制点建立仿射变换的转换方程,采用最小二乘求解整个视场的转换参数,根据解出的参数重新对图像进行迭代,得到收敛后的结果并进行配准。通过对高分辨观测图像和全日面SDO/HMI连续谱图像进行配准,拟合结果的偏差在0.25”以内。 

In the research of the Solar observation, the registration between high-resolution observation images and full-disk Solar images is a greatly meaningful work. However, it is difficult to register with high accuracy due to the rotation, scaling and translation between them. This paper presents an image registration method combining local statistical information and control point matching: At first, between the high-resolution observation image and the full-disk Solar image is preliminary estimated about the orientation, scale, and positional parameters. Then the images are pretreated based on the estimated parameters. And then the pretreatment of the field of view is divided into a large number of overlapping local regions equally. Each corresponding local area is determined by the Correlation matching on the full-disk after preprocessing. Next, the sub-pixel offset between each pair of local regions is then measured, and the coordinate positions of the feature point are determined according to the sub-pixel offset between each pair of local regions, which is used as the feature control point in the point matching. Finally, the conversion equations of the affine transformation are established according to the control point, then use the least squares to solve the entire field of view of the transformation parameters. The image is  0.25 arc-seconds.