革新知能統合研究センター

Speaker: Gerald Baier (German Aerospace Center)

Title: Noise Removal for Synthetic Aperture Radar Interferometry

Abstract: Synthetic aperture radar (SAR) is a remote sensing technology for earth observation. Today several SAR satellites continuously monitor the earth. This talk starts with an introduction into SAR and its applications, before diving into the field of SAR interferometry, which can measure deformations and the height of the imaged terrain. For the latter, German Aerospace Center (DLR) has currently two satellites in orbit, which generated a globally available digital elevation model. As SAR acquisitions are inherently corrupted by noise, a denoising filter is a mandatory, intermediate processing step. The remainder of this talk focuses on a SAR interferometry filter following the nonlocal filtering paradigm, which was originally proposed for removing Gaussian noise in optical images. Several techniques make the proposed filter applicable to a wide range of terrain, leading to improvements to the accuracy of the derived digital elevation model. Due to the vast computation cost of nonlocal filters their efficient implementation on contemporary hardware, either CPUs or GPUs, is briefly touched upon.