SAR Phase Unwrapping Using Path-Based Least-Squares Phase Estimation and Region-Growing with Polynomial-Based Phase Prediction

Differential SAR interferometry (DInSAR) has proven to be a processing approach that is well-suited to precisely identifying large-scale land deformation patterns. This is useful for many environmental monitoring applications, but the speckle noise and temporal decorrelation present in SAR images presents particular challenges in processing SAR images. This research focuses on the phase unwrapping problem, proposing two new approaches: Polynomial-Based Region-Growing Phase Unwrapping (PBRGPU), which expands upon the traditional region-growing approach to phase unwrapping; and Path-Based Least-Squares Phase Unwrapping (PBLSPU), which extends the least-squares phase unwrapping models in a path-based framework. Both algorithms were tested using simulated data and interferograms generated from RADARSAT-2 data. Both approaches significantly reduced the root mean square error compared to the algorithms they build from, and achieved a similar level of performance to the commonly-used SNAPHU algorithm without the need for masking low coherence areas.