Superresolution Reconstruction of Severely Undersampled Point-spread Functions Using Point-source Stacking and Deconvolution

Point-spread function (PSF) estimation in spatially undersampled images is challenging because large pixels average fine-scale spatial information. This problematic when fine-resolution details are necessary, as optimal photometry where knowledge of the illumination pattern beyond native resolution image may be required. Here, we introduce a method PSF reconstruction point sources artificially sampled an and combined together via stacking to return finely estimate PSF. then deconvolved from pixel-gridding superresolution kernel that can used for optimally weighted photometry. We benchmark against < 1% photometric error requirement upcoming SPHEREx mission assess performance concrete example. find standard methods like Richardson--Lucy deconvolution not sufficient achieve this stringent requirement. investigate more advanced with significant heritage analysis called iterative back-projection (IBP) demonstrate it using idealized Gaussian cases simulated images. In testing on real recorded by LORRI instrument New Horizons, able identify systematic pointing drift. Our IBP-derived kernels allow accuracy significantly better than individual exposures. broadly applicable variety problems combines computationally simple techniques way robust complicating factors such severe undersampling, complex PSFs, noise, crowded fields, or limited source numbers.