Deep learning-based galaxy image deconvolution

With the onset of large-scale astronomical surveys capturing millions images, there is an increasing need to develop fast and accurate deconvolution algorithms that generalize well different images. A powerful accessible method would allow for reconstruction a cleaner estimation sky. The deconvolved images be helpful perform photometric measurements help make progress in fields galaxy formation evolution. We propose new based on Learnlet transform. Eventually, we investigate compare performance Unet architectures image astrophysical domain by following two-step approach: Tikhonov with closed-form solution, followed post-processing neural network. To generate our training dataset, extract HST cutouts from CANDELS survey F606W filter (V-band) corrupt these simulate their blurred-noisy versions. Our numerical results simulations show detailed comparison between considered methods noise levels.