Radio-interferometric Image Reconstruction with Group Sparsity Residual Regularization

While compressive sensing (CS) has been widely utilized in this field, tradi tional CS approaches adopt fixed sparse bases for the entire astronomical image, resulting in inadequate sparsity and compromised reconstruction quality. Astronomical images possess prominent nonlocal self-similarity, as extended radio structures, galactic shells, symmetric jets, and background emis sions generate numerous similar patches throughout the field of view.

To improve the reconstruction performance, this paper introduces an image reconstruction method named Group sparsity Residual regularization for Radio-interferometric Imaging (GRRI). The GRRI approach takes the groups of nonlocal image patches with similar structures as the basic unit for sparse image representation. Addi tionally, a low-complexity adaptive dictionary learning strategy based on principle component analysis is used for each group.