Assessment of GPU-Based Enhanced Wiener Filter on Very High Resolution Images

Despeckling of Synthetic Aperture Radar (SAR) images is still an open issue in the remote sensing community. Many filtering approaches have been adopted by the scientific community in order to reduce speckle noise effects including, such as spatial-based and non-local techniques. Mainly all existing approaches present some limitation in terms of accuracy or computational time or edge preservation. Recently a frequency-based solution presented as an improvement of the classical Wiener filter was proposed in order to limit the previous mentioned critical issues. The algorithm was already tested on both simulated and Sentinel-1 data. In this paper, we assess and validate the effectiveness of the proposed solution exploiting different very high-resolution (VHR) data such as, COSMO-SkyMed and RADARSAT SAR images. The results are interesting proving that the developed filter can effectively deal with the case of different VHR SAR images and can provide accurate results within a very limited time.