I agree that ‘Do It Once And Do It Right’ should be the approach to work with StarTools and this is what I prefer also. I don’t have too much experience using NBAccent, but when sometimes the ‘wrong way’ is necessary and yields more pleasing results – well, then that’s fine for me, too.
I also fully agree to your remarks about using NBAccent and about using NB data as luminance. So I gave this approach a try and here is what came out:
What do you think? The faint parts suffer from weak broadband data, exactly as you described. But detail is quite a bit stronger I would say. So choosing between the “wrong way” NBAccent approach and this “NB for luminance and broadband for chrominance” approach this is a trade-off between colour saturation and detail representation. And in the end a question of personal preference (as this is often the case).
(As for processing: Super Structure was skipped, this prevents the background from being mottled. And as a last step I dialed in some skyglow.)
I also reread Carles’ explanations how to layer an NBAccent version for colour, and an NB version for luminance. Interesting, but I have to admit, I’m too lazy to go this way. And my skills working with the Layer module are very limited
Best regards, Dietmar.
As you said, detail is very good. And color is fine across all the brightness range I think. IMHO it's the best version posted in this thread. It would be my 'final' version.