A Method for Phase Unwrapping in the Presence of Dense Fringe Patterns

The phase unwrapping problem is always solved in the hypothesis that the variation rate of the phase function to be reconstructed is not too high, i.e. the phase fringes are not too dense, for significant large areas. There exist cases where this assumption is not true and the classical phase unwrapping approaches fail. For example, Synthetic Aperture Radar (SAR) differential interferometric measurements of fast glacier motion can be characterized by very dense and convoluted phase fringes (Moiré and bull's-eye patterns). In order to circumvent this difficulty, we explore the possibility of relaxing the usual hypothesis about the variation rate of the unwrapped phase. Instead, we assume that the curvature of the phase function is sufficiently small. This assumption can remain valid in a larger number of cases. The proposed phase unwrapping approach is tested on data sets coming from ERS SAR differential interferometric measurements of glacier motion in Greenland. Comparison with the results obtained by using the classical approach shows that significant improvements are achieved in the areas characterized by very dense fringes, while consistent results are obtained in the other cases.