Need help with my color data - purple star halos on RGB

[roofkid]

Hello everyone,

first of all I want to thank the Pixinsight team for their amazing product I have been using it for multiple years and I'm very happy with it. Keep up the good work!

Since I am out of ideas on a current problem I want to ask for help here. I have a problem with purple star halos on RGB data. An example of my recent data of M63:


I have tried multiple workflows but have yet to find a solution or the root cause of this problem. The colors just look "off" to me. The problem gets compacted even more when l MaskedStretch the color data. I am not sure if it's my processing, the changing light pollution due to switching to LED, my scope or my capturing technique (gain / length)

What I usually do:
StarAlignment -> DynamicCrop -> AutomaticBackgroundExtraction -> ColorCombination -> PhotometricColorcalibration (with BackgroundNeutralization) -> linked STF

I have also tried LinearFitting to G and to L with similar results. I have also tried BackgroundExtraction on a combined RGB.

The individual exposures are 300s each at gain 0 with ASI1600 and Esprit 100 with motorized focusser. You have my permission to play with the original data here: https://drive.google.com/open?id=1WPQxtW_2dwRzJsV64lhQOKzfiURCZ2Mf
If you post an edited version of the data anywhere but this thread, please list me as the author of the data (-> https://www.astrobin.com/users/roofkid/)

Thank you for looking and trying to help me,
Sven

[RickS]

Use SubframeSelector to check the FWHM of the individual colour channels.  A likely cause is that the stars are bigger in red and blue than green (or become so when stretched.)

[roofkid]

Hi RickS,

thank you for answering. Indeed the FWHM are different and I also think this is a probably cause.

For Moffat 4, -1 sensitivity:
R -> 2.3 px
G -> 1.8 px
B -> 2.3 px

I used Convolution to have G match the other colors but the issue persists.
Can you think of another strategy to handle this?

[RickS]

Hi Sven,

You could try stretching the RGB image, then extract the colour channels and then use star masks and MorphologicalTransformation to shrink the red and blue stars a little.  Recombine with ChannelCombination.

Another alternative might be to use deconvolution on the linear red and blue masters, perhaps through a star mask.

A third option would be a magenta mask created with ColorMask, clean up with CloneStamp if needed, apply & tweak with CurvesTransformation (bump up green, reduce saturation, etc.)

Cheers,
Rick.