A Refined Maximum-Likelihood Inspired Tomographic SAR Imaging in High Noise Level

We consider using the maximum-likelihood (ML)-inspired methods, such as ML-inspired adaptive robust iterative approach (MARIA), for tomographic synthetic aperture radar (TomoSAR) imaging in scenarios with high noise level. Traditional MARIA method faces two key challenges in this case: first, the positive feedback mechanism likely leads to error propagation in high noise level; second, the assumption of exactly known noise power is usually invalid, both of which result in significant performance degradation. Therefore, we propose a refined ML-inspired TomoSAR imaging method suitable for high noise level, improving the signal and noise estimation process of the typical MARIA. First, we derive a new iterative expression based on ML criterion without the positive feedback mechanism for signal estimation, which suppresses the error propagation in high noise level. Second, we regard noise power as a variable to be estimated and introduce an iterative method based on ML criterion for estimation. The simulation and real unmanned aerial vehicle (UAV) experiment results both verify the effectiveness of the proposed method. Finally, we theoretically give a convergence guarantee for the proposed method.