Ground-based galaxy images sharpened to Hubble quality by new algorithm

Studying the galaxies that host active galactic nuclei is hard because the brilliant central point source drowns out the fainter surrounding structure. This algorithm splits each image into a smooth extended component (the galaxy) and a sparse point-source component (the AGN), then applies a novel 'balance constraint' on their pixel-wise product to prevent over- or under-subtraction. Applied to Hyper Suprime-Cam images of AGNs at redshifts 0–1, the recovered host-galaxy morphologies match Hubble Space Telescope quality. The method is designed to scale to wide-field surveys like Rubin, Euclid, and Roman, enabling large statistical studies of how galaxy structure relates to nuclear activity.