Phase Unwrapping via Deep Learning Based Region Segmentation

Phase unwrapping (PU) transforms wrapped phase into the real one, so that the products of interferometric synthetic aperture radar (InSAR) can provide meaningful information, such as surface height and deformation. By analyzing the characteristics of phase integer modulus, we propose a new phase unwrapping algorithm which combines deep learning and region segmentation with ambiguity number. Because this algorithm calculates the ambiguity number directly, it can improve the phase preserving property of the unwrapped phase and greatly reduce the computation complexity. The simulation and real data experimental results demonstrate the effectiveness of our methods.