FWHM of integration far higher than any individual sub

[Mark de Regt]

Hi,

I'm processing a set of data I acquired over the last several months.  Part of it is a group of 5-minute luminance subs, a group of 15-minute luminance subs, and a group of 30-minute luminance subs.

My intent was to make three masters, then do an HDR merge.

All went well (I thought).  Well, until I found out that the 30-minute integration has a FWHM of about 2.8 arcseconds, when the worst individual sub had a FWHM of 2.0 arcseconds.  I've re-aligned and re-combined several time, with no improvement.  All subs seem to have reasonably round stars, and the alignment, to the eye using Blink, is fine.

The 5-minute and 15-minute integrations are fine, with FWHM in the middle range of the individual subs.

What could be causing this, and can I do anything about it?

Thanks.

Mark

[ngc1535]

I think it would be amazing if you didn't generally have a slightly larger FWHM in the integrated result. The registration process necessarily needs to resample images and this process can smooth/blur/average values a tiny bit. PixInsight's default resample method (Lanczos) does a pretty good job. If you want to see how your results vary- try to combine with different resample methods. You will find bi-cubics tend to give you the smoothest/largest results. Surprisingly nearest neighbor, which has does not use a weighting scheme as all of the others, does surprisingly well if you have many frames to register (20-30 for example).

Of course, if you have a poor registration result (due to optical aberrations or something)- the random errors that combine will also "bloat" the final result.

Just for fun... try NN and see what happens!

-adam

[Mark de Regt]

Thanks, Adam.  I was afraid of that.  I've worked around it by doing an HDR of just the 5-minute and 15-minute images, which has a FWHM of 1.9, and using those data in the brighter areas, keeping the 30-minute data for the really faint stuff.