Reconstructing AstroSat UV Deep Field Images

Abstract

The UltraViolet Imaging Telescope (UVIT) provides a significantly broader Field of View (FOV) than the Hubble Space Telescope (HST); however, this comes at the expense of re- duced spatial resolution. This limits the detection of finer structures in the observed images. Our goal is to enhance image resolution by reconstructing these finer details. To achieve this, we apply deconvolution techniques to UVIT images, aiming to recover features that are otherwise blurred by its broad Point Spread Function (PSF) and buried in Poissonian noise. Our approach primarily relies on Maximum Likelihood Estimation (MLE) methods, in- cluding the Richardson-Lucy (RL) Deconvolution Algorithm and the Alternating Direction Method of Multipliers (ADMM). First, we model the PSF of UVIT in the Near-Ultraviolet (NUV) and Far-Ultraviolet (FUV) channels and analyze its spatial variation across the field of view. Using synthetic galaxies and stars, we assess the performance of RLTV and ADMM in reconstructing Surface Brightness Profiles and recovery of clumps. We then apply these techniques to real UVIT observations in the GOODS-South field, comparing the results to high-resolution Hubble Space Telescope (HST) images. Our findings indicate that RLTV preserves photometric accuracy and produces smooth re- constructions, making it suitable for Surface Brightness Profile analysis. In contrast, ADMM better resolves clumpy structures with fewer artifacts but does not conserve flux, limiting its use in photometry. A hybrid approach, leveraging RLTV for photometric studies and ADMM for morphological analysis, may provide an optimal reconstruction strategy. This study demonstrates the effectiveness of hybrid deconvolution techniques in improving UVIT image quality and aiding future deep-field surveys.