A high-accuracy interferometric SAR images registration strategy combined with cross correlation and spectral diversity

For the image registration of interferometric SAR, there is a contradiction between the registration accuracy and the robustness based on the traditional cross correlation and spectral diversity methods. To eliminate the contradiction, this paper proposes an image registration strategy combined with the respective advantages of above two methods. Its improvements are summarized as follows. Firstly, it divides the image registration process into two main steps. Step one is oriented to the image registration in the range direction based on cross correlation to enhance the robustness. Step two is oriented to the image registration in the azimuth direction based on spectral diversity to ensure the registration accuracy. Secondly, a method for detecting and correcting the singular values in the range pixels offsets is proposed in step one. It can reduce the impact of low-correlation regions. Finally, a hybrid filtering method is introduced for the azimuth pixels offsets in step two to reduce the impact of phase noise. Airborne dual-antenna data are employed to validate the effectiveness of the proposed strategy.