An Effective Image Reconstruction Enhancement Method With Convolutional Reweighting for Near-Field SAR

Compressive sensing (CS) technology has been widely applied in synthetic aperture radar (SAR) imaging. However, the typical sparse reconstruction model,
$\bm {L}_{{{\mathbf{0}}}}$
regularization, can overlook sparse features in target cluster structures, leading to reconstructed SAR images with isolated scatterers and making it impossible to discern the underlying target structure. Furthermore, the iterative hard thresholding method shows great potential in sparse imaging. However, it requires transforming the echo data and imaging grid matrixes into vectors and constructing a large observation matrix. This leads to a significant computational overhead in large-scale scenes. In this letter, we propose a convolutional reweighting
$\bm {L}_{{{\mathbf{0}}}}$
norm regularization method, rooted in structural sparsity within CS theory. This method aims to enhance the sparse features of cluster structures in radar images. To expedite computation, we introduce the Armijo gradient descent line-search step size criterion to adapt the step size in the iteration process and the imaging support set. Ultimately, two sets of experimental data demonstrate the algorithm's accuracy and evaluate its viability.