A new registration method for multi-spectral SAR images

The automatic registration of multi-spectral SAR images for the same scene is rather difficult due to their different appearance. In this paper, we introduce a new feature-based registration method for multi-spectral SAR images. It consists of two steps: in the coarse level registration, an edge feature consensus method is introduced to rectify the rotation and translation difference between the SAR images globally; and in the fine level of registration, the control point pairs are automatically from the globally rectified edge maps and the polynomial transform is adopted to refine the registration. Because the two steps are both applied on the edge maps, the method is very suitable for the registration of multi-spectral SAR images. It is also practical in that it does not need to extract the edge feathers with integrity. The root mean-square errors are below two pixels.