ADAPTIVE DECONVOLUTION OF ASTRONOMICAL RADIO IMAGES

The novel image-forming methods used in the Molonglo Observatory Synth esis Telescope (MOST) are not amenable to a direct application of the self-calibration techniques that have been so effective in improving t he performance of many other aperture synthesis radio telescopes. Spec ifically, self-calibration of antenna-based errors and the use of phas e closure conditions is not possible because vital antenna-based infor mation is lost irretrievably as the MOST forms interferometer beams in real time. However, it is shown in this paper that suitable models of errors introduced by the atmosphere and telescope can be adaptively c onstructed from the properties of synthesised images. These models can then be used iteratively with standard deconvolution methods to impro ve significantly the fidelity of MOST images. The method would have ge neral application to tomographic imaging systems in which non-ideal be haviour could be represented by a small number of unknown parameters.