A Comparative Analysis of 2-Dimensional Model Fitting Algorithms for Astronomical Images

Abstract

Stars and galaxies are captured from light-years away by telescopes and other observational instruments, which produce a pixel grid containing values indicating the light emitted by the object. These pixels have no meaning on their own; important information on the observed object is obtained by fitting models. AstroPhot is a Python-based photometry tool built for fitting such models to astronomical images and uses chi-squared 2D forward modeling to analyze the information they contain. AstroPhot performs sub-pixel integration to ensure high accuracy. In common galaxy models such as Sérsic, brightness within pixels may vary rapidly, so sub-pixel integration is essential to precisely portray these models. Efficiently and accurately performing this type of integration is challenging and many techniques exist to solve this problem. We seek to determine the optimal algorithms/parameters to ensure speed and reliability. To probe this question, AstroPhot galaxy models were compared to those generated by GALFIT, an established photometry solver that also uses chi-squared minimization for fitting. Ideally AstroPhot and GALFIT represent the same model, but due to their differing sub-pixel integration methods, there will be subtle variations in their values. Comparing models from the two algorithms required careful unit conversions due to their different surface brightness parameters. Using a high-resolution image generated with GALFIT as a reference, we sought to determine which pixels in AstroPhot models need more integration, as well as how to get the most accurate pixel value most efficiently.