Improving phase unwrapping techniques by the use of local frequency estimates

In multi-pass space-borne SAR interferometry, the two acquisitions often present low correlation levels and very noisy phase measurements which are incompatible with automatic phase unwrapping. Instead of dealing with many residues due to erroneous wrapped phase diierences, we propose to use the local frequency as measured by a spectral analysis algorithm presented in a previous paper 1]. For this purpose we present two conventional unwrapping algorithms, one local, the other global, that we revisit to beneet from the robust local frequency estimates. For a local approach based on path following techniques, we use the frequency estimates in a slope compensated lter which extend the complex averaging up to a suucient number of look to eliminate residues due to the noise. Then we connect residues due to non-interferometric features along mask components resulting from the detection of lay-overs and uncorrelated areas. For a global approach such as weighted least squares methods, we demonstrate that the use of noisy discrete phase gradient leads to a biased solution. To avoid this drawback, we propose to use the local frequency estimate and associated measure of conndence as phase gradient and weight. Results are presented on both topographic and diierential interferograms obtained from ERS-1 European radar satellite over various landscapes and over the displacement eld of the Landers 1992 earthquake.