Proceedings Article | 21 October 2014
KEYWORDS: Interferometry, Synthetic aperture radar, Speckle, Stanford Linear Collider, Backscatter, Phase interferometry, Buildings, Seaborgium, Image processing, Radar
In this paper, the authors have investigated whether the noise whitening procedure developed as a preprocessing step before despeckling of detected images may be useful also in contexts where phase information is exploited. In a preliminary test set, an interferometric pair of COSMO-SkyMed StripMap images, featuring industrial buildings and vegetated areas, has been: 1) focused without Hamming window (aimed at improving the focusing of targets at the cost of introducing a spatial correlation of background noise), starting from raw data. 2) focused with Hamming window, starting from raw data; 3) preprocessed for complex noise whitening, starting from data at point 2). From the complex interferograms, coherence and interferometric phase maps have been calculated for the three cases by means of boxcar filtering. In case 1) coherence is low on vegetation and also suffers from spreading of areas characterized by strong backscattering because of the presence of high sidelobes. In case 2) points targets and buildings in general are much more defined, thanks to the sidelobe suppression achieved by Hamming filtering, but the background coherence is abnormally increased, due to the introduction of a spatial correlation. Case 3) is the most favorable because whitening operation carries out low coherence on vegetation and high coherence on buildings, where the effects of Hamming filtering are retained. An analysis of the phase field reveals that case 3) should be expedited also in terms of phase unwrapping. Thus, the whitening procedure, devised as a blind preprocessing patch of SLC data, with the goal of a better despeckling, is useful also for SAR interferometry, in which the tradeoff, dictated by the coefficient of the Hamming window, between the ideal situations of focused targets and uncorrelated speckle may be relaxed.
