Statistical and morphological component separation of foregrounds in convolved HI skymaps

Download PDF
Open Webpage

Yanis Guimard, Ming Jiang, Zhenghao Zhu, Huanyuan Shan, Jalal Fadili, Jean-Luc Starck

Published: 2022, Last Modified: 06 Nov 2025SiPS 2022EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: The Epoch of Reionization (EoR) signal is hidden in the overwhelming foreground contamination, which is a challenge in the EoR research. Its detection is often probed by the 21 cm signal originated from neutral hydrogen. Therefore, we address a foreground removal problem to recover the 21 cm signal. We evaluate different blind source separation methods applied on the bright foregrounds in HI skymaps that we have simulated, in order to retrieve the 21 cm signal. We describe our simulation process, review how different frameworks can solve the same problem, then compare the statistical approach FastICA to the morphological and sparsity-based GMCA on mixings of the true skymaps, assuming perfect deconvolution has been performed. We then attempt to deconvolve instrumental effects and compare those methods to another morphological approach, DecGMCA, which performs both deconvolution and component separation jointly.