Abstract:Radio Frequency Interference (RFI) plays an important role on the quality of Ultra-Wide Band Synthetic Aperture Radar (UWB-SAR) imaging. Being modeled as sinusoidal signals, the RFI can be greatly suppressed by the method of iterative Maximum Likelihood (ML) estimation. However, the problems of local minimum and separation of close frequencies exist. In this paper, the difficulty is considerably reduced by the use of eigenstructure analysis to give a better initial frequency estimation. The algorithm is further simplified using a binary hypothesis test, on the number of sinusoidal signals in one spectral peak. The excellent performance of the algorithm is demonstrated by experiments on simulated and real RFI data.