Presentation #108.14 in the session “Missions and Instruments (Poster)”.
Astronomical images in the high and very high energy range typically show low statistics and are distorted by non-uniform exposure, instrumental and astrophysical background and limited angular resolution of the telescope. Fully exploiting the scientific potential of the data requires careful statistical treatment using Poissonian statistics and additional effort to reduce the observational imprint on the data. The most challenging part is the correction for the point spread function of the telescope, also called “deconvolution”. Deconvolution is an ill-posed problem, however by taking into account prior assumptions it is possible to solve. One existing Bayesian algorithm for this is the LIRA algorithm, initially written in the late 2000s as a package for the R language.
Pylira is a new Python package for Low Counts Image Reconstruction and Analysis. It features a Python API to the original C based LIRA algorithm and extends the functionality with high level convenience methods for the visualization, convergence diagnosis, serialization and combination of LIRA deconvolution results. The package also includes online documentation and tutorials. We will present the package with its main features and show simulated as well as real analysis examples based on Chandra and Fermi-LAT data. We will outline future ideas to extend the algorithm for taking into account data from different instruments, including systematic errors and more flexible prior assumptions. The work is supported by NASA APRA grant 80NSSC21K0285.