A Statistical-cost Approach to Unwrapping the Phase of Insar Time Series

Fully 3-D phase-unwrapping algorithms are commonly based on the central assumption that the phase difference between neighbouring sample points in any dimension is generally less than half a phase cycle (the Nyquist criteria). In the case of InSAR time series, however, signals are correlated spatially but uncorrelated over the repeatpass time, due chiefly to changes in atmospheric delay. Here I present an alternative 3-D phase-unwrapping algorithm that treats the problem as a series of maximum a posteriori probability (MAP) estimation problems. This is achieved by generating probability density functions for the unwrapped phase difference between neighbouring points through analysis in time, and then searching for the solutions in space that maximise the total joint probability.