Two Neural Methods for Astronomical Image Restoration

Resumo

Image analysis is a key issue in astronomy. One critical procedure is the photometric analysis of
astronomical images. Therefore, image restoration is a very important operation for the astronomy community,
being a permanent topic under investigation. In this paper, two different methodologies based on artificial neural
networks are explored. A convolutional neural network and the multiscale neural filter approaches are evaluated
on several astronomical images of different objects (stars, constellations, planets) and compared. The structural
similarity index measure is the image comparison metric used to determine the best restoration method.