A robust super-resolution imaging method based on echo denoising and L1 fidelity term regularization at low SNR

Abstract Radar forward-looking imaging has garnered significant attention in many fields. However, due to the ill-conditioned problem of the beampattern convolution matrix, the performance of forward-looking imaging deteriorates significantly. Especially in low signal-to-noise ratio (SNR) conditions, the robustness of forward-looking imaging is reduced by noise, and sidelobes appear in imaging results. A robust super-resolution imaging method based on echo denoising and L1 fidelity term regularization is proposed in this paper. First, the echo signal denoising process is formulated as a low-rank matrix approximation problem, aiming to reconstruct the corresponding noise-free echo matrix as the echo signal input. Subsequently, the total variation norm is introduced to preserve the target contours. Next, the L1 fidelity term and the denoised echo matrix are introduced to construct a robust optimization model, aiming to enhance imaging performance in noisy environments. In addition, the iterative reweighted least squares method and Newton’s iterative method are employed to simplify the solving process. Simulation experiments are designed to prove that the proposed super-resolution imaging method exhibits good performance under low SNR conditions.