Radioastronomical Least Squares image reconstruction with iteration regularized Krylov subspaces and beamforming-based prior conditioning

Abstract

A fast iterative method based on projection onto Krylov subspaces has been proposed for Radio Astronomical (RA) image reconstruction from telescope array measurements. The image formation problem is formulated as a linear least squares (LS) estimation problem by discretizing the Field of View (FoV) of the telescope array into a number of pixels. The ill-posed imaging problem is regularized by the Krylov iterations and the system matrix is prior conditioned by the weights attained from the matched filter beamformed data. The performance of the proposed method is shown based on simulated data from a single station of the the Low Frequency Array Radio Telescope (LOFAR) antenna configuration on a test radio astronomical image. It has been shown that the prior conditioning of the system matrix results in a more accurate image estimate by reducing the artifacts introduced in the empty parts of the image. Furthermore, it was shown that Krylov-based methods fit very well in the context of large scale RA image reconstruction due to their speed and computational benefits.