Line-Wise and Pixel-Wise Approaches for RFI Detection in SAR data: a Comparison

In this paper we present a comparison between two recently proposed approaches for the detection of Radio Frequency Interference (RFI) corrupting Synthetic Aperture Radar (SAR) data. Specifically, we focus on a technique based on binary hypothesis tests [1], [2], which is able to provide one detection decision for each line of the radar samples acquired at a fixed azimuth position, and an approach involving the minimization of the Bayesian Information Criterion (BIC), which provides detection decisions on a pixel-by-pixel basis [3]. The main differences between these line-wise and pixel-wise methods are highlighted from both theoretical and experimental perspectives. In particular, the reconstruction results of simulated data corrupted with RFI, achievable through the considered detection approaches, are compared in terms of coherence and Normalized Mean Square Error (NMSE).