Recovering the wedge modes lost to 21-cm foregrounds

ABSTRACT One of the critical challenges facing imaging studies 21-cm signal at Epoch Reionization (EoR) is separation astrophysical foreground contamination. These foregrounds are known to lie in a wedge-shaped region (k?, k?) Fourier space. Removing these modes excises grave expense image fidelity, since cosmological information also removed by wedge filter. However, EoR non-Gaussian, meaning that lost correlated surviving some covariance matrix. We have developed machine learning-based method exploits this identify ionized regions within wedge-filtered image. Our reliably identifies largest and can reconstruct their shape, size, location an further demonstrate our remains viable when instrumental effects accounted for, using Hydrogen Array Square Kilometre as fiducial instruments. The ability recover spatial from images unlocks potential for current- next-generation instruments without relying on detailed models themselves.