Preventing and masking out unreliable results for critical quantitative applications of phase unwrapping

Phase unwrapping is the reconstruction of a phase function on a grid given its cyclical values (i.e., the values module 2/spl pi/). Phase unwrapping is usually considered successful when errors are concentrated in small regions, and do not propagate throughout the data set. However, for certain applications, even errors on isolated pixels can be critical. For example, when building a digital terrain model (DEM) from synthetic aperture radar (SAR) interferometry, a wrong phase ambiguity can cause an height error of a hundred meters, which is clearly undesirable even for a few pixels. The authors present two approaches for preventing and masking out possible phase unwrapping errors, associated with the branch cut and the network flow based phase unwrapping methods, respectively. The methods are statistically validated on SAR interferometric data simulating the Shuttle Radar Topography Mission (SRTM) data. The tests performed show the validity of the proposed methods, and the importance of excluding possible unreliable phase unwrapping results.