I would like to present a better than ever re-process of the Hubble palette rendered Rosette nebula above. But I'm not. It sucked. That's my old attempt which is better looking. But I'm quite sure my new process was technically better than my previous attempts.
Old process involved stacking the RGB data from each narrowband filtered, then selecting the best colour channel from each, extracting that, recombine and finish off.
New process was to reduce the redundant data up front by 2x2 binning on the raw output, thus the dominant colour channel(s) will dominate. With only one channel per band to work with, the software has less opportunity to get confused. I even managed to recombine them in PixInsight but the DSLR noise still ruled the day.
I did notice PixInsight behaving a little oddly for reasons I'm not sure about yet. Bright stars seemed to acquire a dark halo around them, which I suspect is some side effect of the stacking process but I'm not sure how yet. I am getting better at DBE parameter setting now at least. Had enough practice... also I think the 2x2 binning is not helping. Even if the unbinned version previously contained little extra detail, I think the effective extra sampling (like planetary drizzle) helped with the overall process. So I might try again with that.
What am I getting into? This isn't tweak a few sliders in a raw converter...