An InSAR image registration strategy applied to complex distortion

Image registration is a key step in signal processing of interferometric SAR. Traditional real correlation method is a common and robust approach for InSAR images registration. When the distortion between the two SAR images is large and complex, the registration accuracy of this traditional method is limited. To enhance the registration accuracy, this paper proposes an improved registration strategy with three improvements comparing with traditional algorithm. Firstly, adjacent sample windows and overlapping search windows are employed in the window partitioning. Secondly, a two-step singular points removal and modification method is proposed for the low-coherence regions. Finally, cubic spline interpolation is used to precisely explain the complex image distortion. Airborne dual-antenna data are processed to validate the effectiveness of the proposed registration strategy.