Compressive sensing based multi-frequency synthesis

摘要

Most modern radio telescope arrays observe with wideband receivers to optimise signal-to-noise. However for such wideband visibility data, the changing shape of the source with frequency may limit the performance of existing deconvolution methods. In such a case, it is necessary to estimate explicitly the change in brightness with frequency. This is called multifrequency synthesis (MFS). The current MFS methods either work only for the linear spectral model or take a long time to converge. We propose a new method, MFS-CS, to solve the MFS problem based on the theory of compressive sensing (CS). Experimental results show that it provides superior reconstructions compared to a normal deconvolution method (MSCLEAN) and an MFS-based extension (MFS-MSCLEAN). The main advantages of our method are improved efficiency, compatibility to any spectral model and simplicity implementation. MFS-CS is a potential candidate solution for the next generation telescope.