Optimal weighting and stacking

Angus Lau

Hi Adam,

Thank you
very much for the excellent recent series on weighting methods in PixInsight.
To me, I think the most robust method currently is the PSF Scale SNR and NSG. They
will maximise the SNR of the stacked image, which is the main purpose of
stacking images. While the PSFSW may have good intentions, it is mixing up
frame selection in an arbitrary way which the user has little control.

Ideally, we
want a system which will give us optimal SNR for the faint areas while making best
use of those frames with good resolution (tight stars). I think this may be achievable
to a certain extent. For example, I have been thinking about making two stacks.
One uses NSG for optimal SNR. The other only stacking those frames with good
resolution. Then combine the two stacks together to give us the best of both
worlds. For instance, using something like a luminance mask. Because resolution
is more important for brighter features while SNR is more critical for fainter
features. So they are not contradictive to each other.

I guess a
new stacking engine may be possible to achieve the above. While stacking the brighter
areas, only use frames with good resolution (or give them higher weightings).
While stacking the fainter areas, use weights similar to NSG.

On a
related note. My understanding of current noise evaluation is a global measure
of the image. In order to maximise SNR of the stacked image for faint features,
I believe noise should be measured for the background only or for the fainter
areas of the image. Because this is the noise we care most about.

Any
comments or corrections to my thoughts? Thanks!

Angus Lau