An advanced image registration method based on weighted nuclear norm minimization and SIFT

Synthetic aperture radar (SAR) image registration is significant for mapping, measurement, navigation, and so on. However, it is problematic for the SAR image alignment because the multiplicative speckle noise in the image and the coherent imagery mechanism of the SAR. Therefore, an advanced feature-based image registration method for the SAR image is proposed in this paper. Firstly, the speckle noise is filtered in natural SAR image based on the weighted nuclear norm minimization, which makes the amount of the false feature point reduce. Secondly, with the defined gradient for the SAR image, the improved SIFT method is employed to extract the feature point and generate the descriptor. The experimental results show that, compared to other methods, the proposed method improves both the accuracy of alignment and utilization of feature point significantly.