RPCA: a Resolved PSF reconstruction method for undersampled images

Abstract

The Point-Spread Function (PSF) is a critical component in various astrophysical research areas, including galaxy morphology, photometric redshift determination, and weak lensing studies. As the next generation of space telescopes prepares for operations, including the China Space Station Telescope (CSST), accurate PSF modeling has become increasingly important. However, accurately modeling the PSF remains challenging, particularly for poorly sampled images. To construct reliable PSF models based on star images, we aim to develop a novel and robust method, which accounts for the pixel integration effect in modeling for reconstructing high-resolution PSFs from undersampled or poorly resolved images. In this paper, we present Resolved Principal Components Analysis (RPCA), a novel approach that accounts for pixel integration effects to enable the direct reconstruction of resolved PSFs from undersampled images via an Expectation-Maximization PCA (EMPCA) technique. Our method has been tested on a simulated PSF dataset for the China Space Station Telescope (CSST), where both observational and detector-induced effects were considered during simulation. The results demonstrate that RPCA achieves superior precision in PSF shape parameters compared to existing methods such as Multi-CCD (MCCD), Resolved Components Analysis (RCA), and PSFEx. Notably, RPCA {could reduce the bias of $ΔR^2/R^2$ by approximately one order of magnitude with respect to MCCD, RCA and PSFEx. RPCA exhibits robust stability under realistic observational conditions, including varying signal-to-noise ratios (SNRs) and contamination from cosmic rays and bad pixels. This suggests that RPCA could be a promising candidate for future CSST PSF modeling efforts.