Abstract

Poster - Splinter General   (Kuppelsaal / virtual plenum)

20-FOLD RESOLUTION IMPROVEMENT IN RADIOASTRONOMICAL SINGLE-DISH IMAGE SCANNING

Johannes Ebersberger
Sternwarte Nürnberg, Astronomische Gesellschaft der Metropolregion Nürnberg, Fachgruppe Radioastronomie

Scanning image resolution using small radio telescopes suffers very strongly from the diffraction limit. Since interferometers are so efficient in terms of resolution, no efforts were made to develop solutions for single dishes. An iterative multi-frame superresolution technique using an imaging model combined with Bayesian optimization and stochastic regularization turned out to be most advantageous. Apriori information about the imaging process as well as an adaptive procedure regarding the pixel intensity statistics allows to reconstruct the fine-structure content in the subpixel domain to an unexpected high extent. Generally, Bayesian optimization is very sucessful in artificial intelligence applications, multi-frame superresolution algorithms are already standard in some mobile phones. Images from an 18-hours-scan of the northern part of the milky way with a 2.6-meter single dish equipped with 7 feeds for 21cm will be shown. Combined with the above mentioned algorithm, an image resolution improvement down to 0.25° is achieved, instead of the standard resolution comparable to the beam size (5.5°). A comparison with images from the EBHIS HI survey will be shown. A 40sec-YouTube-Video shows the development of the final high resolution image during processing (dtCi7cm1E5Y).