Low-rank EM-based imaging for large-scale switched interferometric arrays

Interferences and computational cost pose significant challenges in large-scale interferometric sensing, impacting the accuracy and numerical efficiency of imaging algorithms. In this paper, we introduce an imaging algorithm using antenna switching based on expectation-maximization (EM) to address both challenges. By leveraging the low-rank noise model, our approach effectively captures interferences in interferometric data. Additionally, the proposed switching strategy between different sub-arrays reduces significantly the computational complexity during image restoration. Through extensive experiments on simulated datasets, we demonstrate the superiority of the lowrank noise model over the Gaussian noise model in the presence of interferences. Furthermore, we show that the proposed switching approach yields similar imaging performance with fewer antennas compared to the full array configuration, thereby reducing computational complexity, while outperforming non-switching configurations with the same number of antennas.