RFI Removal From SAR Imagery via Sparse Parametric Estimation of LFM Interferences

One of the challenges in spaceborne synthetic aperture radar (SAR) is modeling and mitigating radio frequency interference (RFI) artifacts in SAR imagery. Linear frequency-modulated (LFM) signals have been commonly used for characterizing the radar interferences in SAR. In this letter, we propose a new signal model that approximates RFI as a mixture of multiple LFM components in the focused SAR image domain. The azimuth and range frequency modulation (FM) rates for each LFM component are estimated effectively using a sparse parametric representation of LFM interferences with a discretized LFM dictionary. This approach is then tested within the recently developed RFI suppression framework using a 2-D SPECtral ANalysis (2-D SPECAN) algorithm through LFM focusing and notch filtering in the spectral domain. Experimental studies on Sentinel-1 single-look complex (SLC) images demonstrate that the proposed LFM model and sparse parametric estimation scheme outperforms the existing RFI removal methods.