Minimum mean square error space-varying filtering of interferometric SAR data

This paper addresses the problem of filtering Interferometric Synthetic Aperture Radar signals both in presence of non-planar topography and different Doppler centroids, to mitigate geometrical decorrelation effects. The problem is space-variant; we assume knowledge about the scene topography and derive an optimal, Minimum Mean Square Error, filtering procedure. The algorithm is flexible and, beside the standard stripmap-stripmap interferometry, it may be applied to interferometric SAR data acquired in any operative mode: for instance in scan-scan, scan-strip and scan-spot interferometry. The scene topography contribution may be either derived from an external rough Digital Elevation Model or directly estimated from the SAR data. Experimental results carried out on real data confirm the validity of the theory and show that this filtering procedure allows us to obtain a reduction of the interferometric noise content. Its gain is particularly marked in the cases of steep topography, where application of the standard common band filters could deteriorate the signal quality, or for large Doppler Centroid shifts.