Natural looking stars

[Vallantho]

I’m using an L-Enhance filter and after processing in Siri my stars are looking a bit on the fake side  they’re just white dots.  No colour, no falloff  I’m wondering what is the best way to get a more natural look.

Im thinking the best thing to do is shoot a bunch of subs without the filter and stack them  Then when I use starnet to recombine with the stars from the broadband image instead of the ones from using the filter, if that makes sense?

What is a problem I think is that I have to remove the camera to remove the filter  I don’t have a filter draw at the moment. So realigning the  camera is gonna be a bit hit or miss  

what other ways are there I could preserve some star colour?

Maybe do a photometric colour calibration and layer the images up in affinity and do it that way?

 

 

[ONIKKINEN]

Would you be willing to share the raw stacked file, or just a single subexposure? It would be easier to give advice with the raw data at hand.

But anyway, how are you stretching the data here? You shouldn't drop the white point below the point of star saturation, and it looks like that may have happened considering that every star here is just white and colourless. Its also easy to destroy star cores if you use one of the more advanced hyperbolic stretch tools and place some of the sliders somewhere other than the best spot, personally i dont like using them too much. If you ran deconvolution you may have also destroyed the stars if it was ran too aggressively. The image looks fairly noisy so i would recommend not running that in this case.

[Vallantho]

I've uploaded a sub and the raw stacked file for you.

In Siril i did a background extraction, photometric colour calibration, green noise removal and then used starnet. I used a generalised hyperbolic stretch on both the starless and on the stars.  What form of stretching would you recommend for the stars?

I didn't do any deconvolution.

result_10320s.fit Cygnus_Wall_2023-06-13_Light_nofilter_050.fits

[ONIKKINEN]

1 hour ago, Vallantho said:

I've uploaded a sub and the raw stacked file for you.

In Siril i did a background extraction, photometric colour calibration, green noise removal and then used starnet. I used a generalised hyperbolic stretch on both the starless and on the stars.  What form of stretching would you recommend for the stars?

I didn't do any deconvolution.

result_10320s.fit 227.03 MB · 0 downloads Cygnus_Wall_2023-06-13_Light_nofilter_050.fits 37.84 MB · 0 downloads

I ran a very quick process (in Siril) to the stack:

Looks like star colours are alive and well, so at least its not a capture issue where stars were saturated to begin with. So its all fixable it seems. Stars are bi-colour because of the filter passing only the blue-green O3 and the very deep red Ha. Personally for this reason i would work on the stars-only layer in Photoshop and manually fiddle with the levels until the starcolours are the most natural to my eyes, but this one isn't too far off either so not strictly necessary.

What i did: Bin x3 (average), Slight crop, background extraction (crude, could have done a better job), Photometric colour calibration followed by a manual white reference on that bright-ish white looking star in the top left corner just because i thought it might take the edge out of the very red stars. Stretched first with an Asinh stretch at 300 power and then a simple histogram transformation.

Where you probably took a wrong turn would be the hyperbolic stretches, especially if you ran it on the stars-only image. I recommend first processing the image to the point where stars look good to your eyes and only then separating them with starnet. Then keep working on the starless layer until you like what you see and re-introduce the stars. This way the star layer needs minimal work because they were good before you separated them. I usually do all this in Photoshop because its much easier to see what is going on with the layer system. But typically the stars only layer sees only minimal work with maybe some saturation or levels adjustments, but definitely no hyperbolic stretching at that point.

Edited June 21, 2023 by ONIKKINEN

[Vallantho]

Hey that’s great. Definitely better. Thanks for the8 help. I’ll go through it and try things like you say and see how I get on 👍👍