Scattering-based Despeckling of Multi-frequency SAR Data

Speckle noise is a major factor impairing synthetic aperture radar (SAR) imagery interpretation and processing. In this paper, we propose a simple approach to deal with the despeckling of multi-frequency SAR data acquired over bare soil surfaces. The basic idea is to perform a proper normalization step in order to compensate the frequency dependence of the SAR measurements and make histograms of the multi-frequency images comparable. Once this pre-processing step is performed, conventional multi-temporal filtering can be applied. We tested this simple approach on both simulated and real-world multi-frequency datasets. Obtained results show that the proposed idea leads to increased speckle suppression capabilities with respect to both single-channel filtering and pure multi-temporal despeckling.