search

astrostat / pylira

A Python package for Bayesian low-counts image reconstruction and analysis
9 stars 7 forks source link

Add disk source test dataset #23

Closed adonath closed 2 years ago

adonath commented 2 years ago

This PR adds a disk_source_gauss_psf test dataset. The disk has an extension comparable to the size of PSF and in addition I introduced a gradient to the exposure. This is what the data look like:

image

The idea was to impose a little challenge to the algorithm, by (a) having a structure that has an extension comparable to the PSF and (b) a symmetry that is not adapted to the multiscale reduction scheme.

adonath commented 2 years ago

@infinitron Test wise I ran the algorithm with high statistics on the data and I got a rather poor result:

image

I assume the result can be much improved by adapting the multiscale prior more towards extended emission. Would you have any recommendation for this? Getting a reasonable result here is secondary, but it doesn't hurt either. I think step by step we should add 2-3 other (possibly) challenging test cases and then we have a reasonable coverage of the current state of the algorithm to develop it further.

infinitron commented 2 years ago

@adonath Interesting! Can you try switching the prior to this: kappa1=1, kappa2=5, kappa3=1? I tried this prior on a different source and it appeared to capture some diffuse emission.

adonath commented 2 years ago

Thanks @infinitron! This didn't really help:

image

For this PR it's probably fine, but we should keep in mind to document a bit how to get reasonable results for this test case. I'll merge this PR now..