Phase unwrapping for large SAR interferograms: Statistical segmentation and generalized network models

Two-dimensional (2-D) phase unwrapping is a key step in the analysis of interferometric synthetic aperture radar (InSAR) data. While challenging even in the best of circumstances, this problem poses unique difficulties when the dimensions of the interferometric input data exceed the limits of one's computational capabilities. In order to deal with such cases, we propose a technique for applying the statistical-cost, network-flow phase-unwrapping algorithm (SNAPHU) of Chen and Zebker to large datasets. Specifically, we introduce a methodology whereby a large interferogram is partitioned into a set of several smaller tiles that are unwrapped individually and then divided further into independent, irregularly shaped reliable regions. These regions are subsequently assembled into a full unwrapped solution, with the phase offsets between regions computed in a secondary optimization problem whose objective is to maximize the a posteriori probability of the final solution. As this secondary problem assumes the same statistical models as employed in the initial tile-unwrapping stage, the technique results in a solution that approximates the solution that would have been obtained had the full-size interferogram been unwrapped as a single piece. The secondary problem is framed in terms of network-flow ideas, allowing the use of an existing nonlinear solver. Applying the algorithm to a large topographic interferogram acquired over central Alaska, we find that the technique is less prone to unwrapping artifacts than more simple tiling approaches.