I am currently migrating everything I have developed over the past few months from my private repo to this public repo. But I am still thinking of how the code organisation will be. The idea is that this package contains multiple utilities (for radio astronomy, optical, etc), combining different kinds of analysis. The code aims is to be as general as possible and easy to use.

Module Sersic Fitting/Image decomposition of morphen

It would be critical to stablish on-the-fly (during the minimisation) which model components are 'compact' and which are 'extended'. Hence, we can get a better report of statistics to highlight compact and diffuse structures.

To do:
Compare minimisation runtime using morphen, IMFIT and GALFIT.

There is an unknown issue where jax.scipy.signal.fftconvolve and scipy.signal.fftconvolve yields different results with LMFIT.

• Investigate possible errors
• a simple convolution test with both cases result in the same output, but not when running an optimisation

The following experiments must be conducted:

• Compare results when the background is fixed to be the residual map from deconvolution and when is set to be a multiplicative scalar of it (e.g. an additional parameter to me minimised)
• Compare results with a simple flat-sky model
• Compare results when a bkg estimator is used as a model of the background

Need to improve source detection, in order to work well for a wide range of images with different resolution, SNR, etc.

Source extraction must be improved. Tricky to find 'default' detection parameters since the noise structure and SNR changes significantly across images with distinct angular resolutions (e.g. eMERLIN -> JVLA).

Check the interferometric decomposition implementation.

• CASA task imsmooth is providing different results than scipy.signal.fftconvolve, setting proper normalisation parameters
• Check where CASA task imsmooth should or should not conserve flux density.
• Not adding a background estimate for complicated structures during minimisation is leaving significant negative residuals on VLA maps.
• Investigate scaling factors between different PSF beams.

An issue needs to be fixed for the following selfcal run:

• Only steps 'p0' and 'ap1' are executed
• If step 'p0' uses combine='spw' and step 'ap1' uses the bright template, the corresponding spwmap for ap1 will be wrong, as it will not use the spwmap from step p0.
  This is a typical run for e-MERLIN observations. Sources may be bright, but we should avoid performing multiple steps of selfcal, specially if the data has been flagged out significantly during phase referencing calibration.

When using usemask='auto-multithresh' and outlierfile, CASA gives and error.

• investigate this further to see the real origin of the issue and if it can be fixed.
• this is critical if one would like to run automatic self-calibration.