From DESI to Euclid: A Generative Bridge to Unbiased Galaxy Structures

Abstract

Ground-based seeing imprints a size-dependent bias on galaxy structural parameters, yet the space-based imaging needed to remove it currently covers only a small fraction of the sky. We close this gap with a generative model that translates DESI imaging of Bright Galaxy Survey (BGS) targets into Euclid VIS images. A Fourier-domain analysis confirms that it recovers structure down to 0.37'' (from the 1.41'' DESI r-band baseline), an approximately 3.8-fold improvement in resolution. Although it stops short of the 0.16'' Euclid VIS resolution, this recovery already de-biases the structural parameters relative to the DESI r-band structure measurements: the Petrosian radius bias falls to +0.075'' (from -0.870''), independent of galaxy size; the Sérsic-radius bias drops to -0.018'' (from -0.322''); and the Sérsic-index bias to +0.093 (from +0.262). We release these translations over the Euclid DR1 footprint as the Euclid-resolution BGS (E-BGS), which can be blindly validated once DR1 is public.