Improve star selection algorithm
Summary
Currently the stars, which we attempt to reconstruct, are selected in a very simple way: we take all stars which have magnitude in V-band less than predefined value (default:7) and are situated within two degrees around the pointing direction.
This approach works well in sparsely populated parts of the sky, but we would have issues when facing a more densely-populated sky region which may contain double stars or clusters of stars or another bright extended object.
What is the expected correct behavior?
One option (simple) is to filter out double stars and other alike objects from the selection list and mask them in a subsequent analysis. The affected pixels should not enter in the calculation of the NSB baseline and should not be considered as reference stars for the fit.
Second option would be to consider such object, but with prior knowledge about their origin to provide a suitable treatment. This may become quite complex especially if we have a variable object (either in brightness or in geometry) with a variation timescale comparable to the observation time.