Tidy up 2D smoothing

[MatthewWhiting]

The 2D smoothing, while working, it still a bit awkward. The kernel needs normalising so that the fluxes of things don't increase (even though we make measurements in the original image array, having the noise quoted in such large units is awkward).

[MatthewWhiting]

[1187] implements the correct normalisation of the Gaussian, dividing the kernel by 2*pi*sigmaX*sigmaY (and finalised in [1194]). This makes the flux in the smoothed image go down (in the same way as for the spectral smoothing - see paper), and it seems to be doing the right thing now.

I also had to rotate the PA of the Gaussian so that it now works as the proper astronomical position angle (measured east of north).

Appears to be working fine now. Closing this ticket.

[MatthewWhiting]

Reopening this, as there are a couple of things that need addressing.

The S/Nmax value for the detection is now inconsistent - it is based on the peak, measured from the original array, divided by the noise, measured in the smoothed array. To be correct, the peak used must be taken from the smoothed array as well, as that is where the detection is done.

Also, the flux errors use the noise from the smoothed array - for these it doesn't really make sense to calculate these errors, since we would need the noise in the original array (as that is where we are measuring the flux). So this may be a case where we simply don't report the errors. This should be the case for spectral smoothing as well.

[MatthewWhiting]

This was done in [1198], essentially as described in the text above. I had to fix up some of the verification results, as the S/Nmax value changed, but that is fine.

Closing again.