Anders Hjelset
Active member
Does anybody know how to deal with red halo rings on high-luminance stars, predominated created on stars with strong Ha..? See a typical example from M101 Pinwheel. Thanks a lot 
How to remove red halo rings around stars
- Thread starter Anders Hjelset
- Start date
fredvanner
Well-known member
This is the unavoidable result of a non-colour-preserving non-linear stretch. Here are two synthetic gaussian stars of exaggerated uniform colour, a red star with R:G:B ratios 5:1:1 and a blue star with ratios 1:1:5, against a low neutral background, scaled 4:1:
If I apply a standard STF, I get:
The bright white core is simply the result of the non-linear stretch "forcing" the brighter core pixels closer and closer to the maximum value, and so closer and closer to pure white. So it is the core that is artificially white, not the border that is artificially coloured.
If I apply a standard STF, I get:
The bright white core is simply the result of the non-linear stretch "forcing" the brighter core pixels closer and closer to the maximum value, and so closer and closer to pure white. So it is the core that is artificially white, not the border that is artificially coloured.
Last edited:
fredvanner
Well-known member
For comparison, this is the same image with a maximum (stretch factor 1000) colour-preserving ArcsinhStretch:
So the answer is: experiment with alternative, colour-preserving non-linear stretch techniques.
So the answer is: experiment with alternative, colour-preserving non-linear stretch techniques.
Anders Hjelset
Active member
Thanks, but shouldn't there then be rings of kind of colors..? It's only the stars with high Ha emmissions that show up like this.
It could however very well be a focusing issue, since the focuser used Green when AutoFocusing.. Maybe Red/Ha was a bit too far off?
Anders Hjelset
Active member
Thanks, this is indeed what I also thought would be the reason. But why Red only? There are no stars with rings if not just Red ones. Could it be that these stars happen to have a very high Ha emmission that make them bloat far sooner when stretched..? I mean, Red do catch Ha ..even if I've heard that our eyes have limited receptors..
How would you have tried to at least surpressed the clear rings, if possible? The stars should however still appear with a red tone. Any idea would be much appreciated
fredvanner
Well-known member
This effect is most noticable for dimmer redder stars. Hot "blue" stars are usually very unsaturated blue-white; cooler red stars are increasingly more saturated red.
Could you upload an unstretched image for testing.
This is not true. I can see a less saturated "ring" around the other bright stars in your screenshot - but since they are hotter and whiter anyway it is less prominent.
Could you upload an unstretched image for testing.
Anders Hjelset
Active member
Sorry, I didn't see your proposal of trying ArcsinhStretch above when asking for a solution proposal
Many manu THANKS..!!
Anders Hjelset
Active member
Yes indeed, I see that also now. Thought they were too faint to bother about. But in perpective of your clear explanation I will certainly give color-preserving non-linear stretch techniques a try..!!
fredvanner
Well-known member
BTW, stars with stronger signal in the H-a filter are not strongly emitting H-a - they are just cooler. What you are seeing in the H-a image is almost entirely the H-a part of the continuum spectrum.
fredvanner
Well-known member
A brief warning: colour-preserving stretchs will inevitably result in a dimmer image. It is dimmer because there is less highly unsaturated near-white content.This is mostly a matter of changing your expectations - the images full of bright white stars are just a convention; the slightly dimmer images with correctly coloured stars are a better convention if you want to avoid artefacts like the red rings.
cloudbait
Well-known member
There's another answer... keep the stretch that results in the center of stars moving towards unsaturated color and the halo showing color. Which I think is the more aesthetically pleasing result and more accurately represents the telescopic view, as well.
Anders Hjelset
Active member
Thanks Chris ..but I should still use ArcSinh to stretch, or what..?
I figured that running Factor=10 repeatedly seems to be a "smooth way" of doing this. And of course watch a zoomed-in star carefully to know when to stop and start slowing down the factor..
cloudbait
Well-known member
I rarely use ArcSinh stretch. I'm generally happy with either GHS or HT.
Anders Hjelset
Active member
Same here Chris, at least until now. However, I find GHS a bit overengineered (overparameterized
). The combination of HistTrans and CurvTrans offers everything for my stretching needs. Except for the function 'Linear' in GHS where BP could be lowered "below zero" to restore a decent lighter/hazier background. If the background by some reason should have been made too black earlier.. which indeed happens (uhuu).
fredvanner
Well-known member
GHS has an Arcsinh mode. If you use this in "Lightness" colour mode you should avoid coloured rings.
Anders Hjelset
Active member
Ahaa, I never noticed that, Fred.
Well, I didn't actually bother to dig into "for me appearently overcomplicated things" like ArcSinh until you hinted me about its features and capabilities.. Many thanks again though..!!
Anders Hjelset
Active member
In Lightness mode the image just turns all white. In RGB and Color mode the rings are less visible, perhaps a bit more visible in RGB mode. Or did I misunderstand something here, Fred?
fredvanner
Well-known member
When I tried it out I got reasonable results with "Lightness" mode and a stretch factor of ~4. I found RGB mode tended towards the "coloured ring" effect with higher stretches. "Colour" mode seemed to oversaturate the colour. I'm testing on my synthetic images (since I know what the real data is there); if you post a test image (raw integration without any other post-processing), I could experiment with it. I guess trial and error is the order of the day for this problem.
Anders Hjelset
Active member
I think there is a dilemma since SPCC should be done in linear which causes ArcSinh stretch to alter the SPCC color balance in non-linear. And the necessary "dimming" is also somewhat tricky to compensate for. I think I have to put this issue on-hold ..until I really need to dive deeper into it..
But, many thanks anyway for your most engaged assistance..!!Best // Anders