Wide field astronomical image restoration with convolutional neural networks
Loading...
Date
2023
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Most image restoration methods in astronomy rely upon probabilistic tools that infer a best solution for a deconvolution problem. They achieve good performances when the Point Spread Function (PSF) is spatially invariable in the image plane. However, this later condition is not always satisfied for real optical systems. PSF angular variations cannot be evaluated directly from the observations, neither be corrected at a pixel resolution. The new method for the restoration of images affected by static and anisotropic aberrations is developed using Deep Neural Networks that can be directly applied to sky images. The network is trained using simulated sky images corresponding to the T80-S Telescope optical model, an 80 cm survey imager at Cerro Tololo (Chile), which are synthesized using a Zernike polynomial representation of the optical system and real observation conditions, like atmospheric turbulence and detector noises.
Once trained, the network is used directly on real images from the T80-S telescope, resulting in a prediction that is a corrected version of the image, characterized by a known and constant PSF across the field of view, zero noise and compensation for the effect of atmospheric turbulence. The method tested on real T80-S images was compared to the relative positions of objects in the GAIA survey catalog.
Description
Keywords
Astronomy, Image restoration, Convolutional neural networks, Artificial
intelligence, Spatially variable point spread function