Thank you, everyone, for the compliments.
I try to take a more minimalistic approach to my processing compared to when I first started using PixInsight. I discovered that I introduced many problems in my processing that I subsequently struggled to fix. And example of this would be stars with blown out cores, bloated stars, or misshapen stars from noise reduction techniques that were too aggressive.￼￼￼￼ I do all of my pre-processing manually in PixInsight. This is the only way I have gotten consistently clean results. Besides, I like to review my raw data frame by frame, anyway.￼￼￼￼
Once pre-processing is complete, I try to be as “gentle” with my data as possible. I do a lot of corrections in the linear state.
I usually start with a crop and and DBE on each narrowband master file.￼￼￼ Linear fit is next. Then I run two rounds of noise reduction on each mono image with TVGD￼ and MMT, according to the technique documented by Jon Rista.￼ I apply a very protective mask to the image when applying these processes because I want to preserve as much detail as possible.￼ I am willing to accept some residual noise to preserve detail.
I rarely do deconvolution anymore because I found it either creates too many artifacts or does not make enough of a difference to matter.￼ Next is RGBCombination, BackgroundNeutralization, the CorrectMagentaStars script, and a pretty aggressive SCNR to remove the inevitable green cast.￼￼ I may boost color saturation afterwards if SCNR results in washed out colors.￼
At this point, the image usually looks promising and is ready for a stretch. You specifically commented about my stars, and I believe the histogram stretching process is where stars can get damaged the most.￼￼￼ I start with a stretching technique known to preserve color saturation, such as masked or arcsinh stretch, and I only stretch to a background value of 0.01 or so. I always set the black clipping point to zero because clipping the blacks gives no leeway for smoothly altering the sky background and "shadows" later.
After the initial stretch, I use HistogramTransformation to stretch the image almost all the way to where I would like the final image to be. I do the HT stretch two or three times, setting the midtone value to 0.2 and the high dynamic range extension ("high range") to 1.1 each time. When I am done, the brightest stars have a core value of about 0.8 and the background sky value is around 0.15. Dynamic range extension is very important when trying to preserve star colors and minimize star size. Stars start to look big and blown out when the core gets oversized. By controlling the maximum brightness of the star during the stretch, star size is minimized and color maximized.
Depending on how bright the stars appear relative to the nebulosity, I may further dim the stars by making a star mask and applying a HistogramTransformation in reverse to tone down just the stars. The effect is subtle but makes a big difference in that it not only dims but shrinks the stars a bit, and does so without destroying the natural spectral contour of the star the way star minimization techniques often do.
A couple of rounds of CurvesTransformation is next, first to adjust the highlight/midtone/background balance and then to specifically highlight areas of interest. For example, if there is a luminous wall of gas and dust that doesn't quite "pop" out of the picture, I will make small adjustments to the portion of the curve that affects just that area until it looks the way I want. I may do this a few more times for different parts of the image.
Next, I make a luminance mask and boost the color saturation to the brighter areas. To minimize chromatic noise in the background, I will sometimes invert the mask and decrease color saturation in the shadows. If the noise is starting to get out of hand, I will do another pass of TGVD.
That's it for PI, really. The final step is importing the image into Lightroom and adjusting it until I am happy.
Edited by gundark, 22 January 2020 - 11:42 PM.