A Modified Minimum Cost Flow Phase Unwrapping Method Based on Reliable Pixel Detection

Phase unwrapping has been one of the most focused and challenging topics in interferometric synthetic aperture radar (InSAR). It imposes a direct impact on the InSAR products, e.g., digital elevation model (DEM) and Earth surface deformation. In fact, except for a few deep-learning-based algorithms, the task of most single-baseline phase unwrapping algorithms is to appropriately locate discontinuities. However, for most of the current methods, misplacement of discontinuities on reliable pixels that do not contain discontinuities often happens, especially when dealing with interferograms that are distributed with dense noise and layover, which will lead to phase unwrapping errors in a large area. To mitigate this problem, a modified minimum cost flow (MCF) phase unwrapping method based on reliable pixel detection is presented. By introducing a phase quality index, reliable pixels are detected. Then, a modified MCF algorithm is applied to locate discontinuities outside of the detected reliable pixels. Finally, an integration guided by the quality index and the discontinuity location is employed to retrieve the unwrapped phase. Experiments conducted both on simulated data and realistic data have confirmed the effectiveness of the proposed method.