I’ve used StarX on a drizzled narrowband image, and it is taking out some of the nebulosity. I’ve had this happen before, but it hasn’t been anything significant. In this instance, I’m going to be adding in separate RGB stars, not the original NB stars. I’d like to find a way of easily adding the removed nebulosity back into the starless image, whilst keeping it in the linear state. Is that possible?
For reference, this is the result after running StarX (zoomed in on the offending area):
Oh hey, I recognize that object ;)
I recently found that making a HRR/OGG images help to keep starx from removing filaments like these, as well as small blobs that look like stars but aren’t.
Basically, you make an NB/BB/BB image, background neutralize it, apply Color Calibration such that the average star is white. You should be left with an image where all emission signal is red and continuum signal is grayish. Now you can extract stars and, hopefully, get a cleaner result.
I believe this works because you’re essentially giving starx a cleaner star field reference.
Gordon Pegue · Aug 21, 2025 at 09:14 PM
What is this an image of?
The Soap Bubble, next to the Crescent nebula.
lucasjacobson · Aug 21, 2025 at 09:34 PM
Have you tried running StarX again but just on the star image? It might take it out, and you could add the result back to your original starless image.
This worked quite well. Not 100%, but certainly workable for my needs. What’s the best way of adding it back in to the starless image?
andrea tasselli · Aug 21, 2025 at 09:50 PM
Don't use SXT. Try Starnet++ V2.
The thought crossed my mind, I’ll give it a go!
Mikolaj Wadowski · Aug 21, 2025 at 10:04 PM
Oh hey, I recognize that object ;)
I recently found that making a HRR/OGG images help to keep starx from removing filaments like these, as well as small blobs that look like stars but aren’t.
Basically, you make an NB/BB/BB image, background neutralize it, apply Color Calibration such that the average star is white. You should be left with an image where all emission signal is red and continuum signal is grayish. Now you can extract stars and, hopefully, get a cleaner result.
I believe this works because you’re essentially giving starx a cleaner star field reference.
Very interesting! How on earth did you come up with that idea?! BTW, I’ve just seen your version of this target and it’s an absolute beaut, congratulations!
Jamie Macdougall · Aug 21, 2025, 11:53 PM
Mikolaj Wadowski · Aug 21, 2025 at 10:04 PM
Oh hey, I recognize that object ;)
I recently found that making a HRR/OGG images help to keep starx from removing filaments like these, as well as small blobs that look like stars but aren’t.
Basically, you make an NB/BB/BB image, background neutralize it, apply Color Calibration such that the average star is white. You should be left with an image where all emission signal is red and continuum signal is grayish. Now you can extract stars and, hopefully, get a cleaner result.
I believe this works because you’re essentially giving starx a cleaner star field reference.
Very interesting! How on earth did you come up with that idea?! BTW, I’ve just seen your version of this target and it’s an absolute beaut, congratulations!
I think it was 2am and I was just experimenting while trying to get Blurx and Starx to stop frying the corner stars in my image haha. I forgot to update filter offsets after some changes in the optical train which resulted in less-than-ideal star performance. And yeah, this technique also seems to improve Blurx’s performance on narrowband stars with heavy aberrations.
Thanks for the kind words! I’m sure your image will turn our fantastic too, you seem to have a lot of signal based on the star-only image you posted.
Jamie Macdougall:
I’ve used StarX on a drizzled narrowband image, and it is taking out some of the nebulosity. I’ve had this happen before, but it hasn’t been anything significant. In this instance, I’m going to be adding in separate RGB stars, not the original NB stars. I’d like to find a way of easily adding the removed nebulosity back into the starless image, whilst keeping it in the linear state. Is that possible?
For reference, this is the result after running StarX (zoomed in on the offending area):
John Stone · Aug 22, 2025 at 01:34 AM
then use PixelMath and subtract the stars from the original image to remove them again but leave the nebulosity alone.
That sounds like a good easy method! What’s the pixel math expression you’d use?
Jamie Macdougall:John Stone · Aug 22, 2025 at 01:34 AM
then use PixelMath and subtract the stars from the original image to remove them again but leave the nebulosity alone.
That sounds like a good easy method! What’s the pixel math expression you’d use?
Are you running SXT on a linear or non linear images, I have had problems with SXT when running it on non linear and now only run it on linear images before applying NXT.
Aaron H. · Aug 22, 2025 at 07:57 AM
Jamie Macdougall:
John Stone · Aug 22, 2025 at 01:34 AM
then use PixelMath and subtract the stars from the original image to remove them again but leave the nebulosity alone.
That sounds like a good easy method! What’s the pixel math expression you’d use?
By default, unless the "unscreen" button is selected, it's a simple subtraction.
Run StarXTerminator, then undo on the main image, stamp out any problematic nebulosity on the extracted stars, then run "myimage - myimage_stars". You'll now have a new matching pair with your tweaks applied.
Thanks!
Tom Engwall · Aug 22, 2025 at 12:41 PM
Are you running SXT on a linear or non linear images, I have had problems with SXT when running it on non linear and now only run it on linear images before applying NXT.
Linear
I found success with both Starnet and the clone stamp method. However, both methods have downsides/minor problems so I’m going to try a combination of them both.
Thank you everyone for your input!
TL;DR: Managed to create a perfectly starless image using multiple techniques.
If anyone is interested, and for prosperity, I thought I’d give a quick run down of my findings below.
StarX worked great on the Ha and Sii, but would remove nebulosity from the Oiii. Starnet2 worked well on the Oiii but not perfect. So I used StarX on Ha and Sii, and tried multiple methods to get the best result on my Oiii:
The clonestamping of the nebulosity in the stars-only image (from StarX) and then performing a manual removal of this image from the original (nebula and stars) image using pixelmath did work quite well and was quite fun, but inevitably I couldn’t get it perfect and ended up cooking the edges of some of the stars close to or inside the nebulosity.
Performing 2x upsample with Starnet2 worked, but would unfortunately leave residue around the brighter stars in the starless image. It also took forever and crashed Pix a couple of times - when reloading I found Starnet2 disappeared from the processes list but remained in the repositories. Long story short, I had to delete the “updates” file for my repository and then perform a “check for updates” in the repositories to reinstall all of them fresh. There may be a better way to fix this but it wasn’t too difficult and in any case that’s a topic of discussion for another time!
Reading the tips on Starnet2 website, I tried creating a HOO image and reducing the STF stretch slightly before running star removal. This was tricky because I didn’t know exactly how much to reduce the STF by and therefore couldn’t get a good result.
In the end, I used Starnet2 on the Oiii image. The result left one or two larger stars in the nebula only image, so I then ran StarX on that starless image which removed the larger stars perfectly, but unfortunately also removed some of the nebulosity. Because this new stars-only image contained only the brighter stars and the minor nebulosity, I was easily able to clone stamp out the nebulosity from this new stars-only image then use pixelmath to manually remove it from the starless image (having first hit undo on the StarX run). This left me with a perfect starless Oiii image.
Reading back, that seems excessively complicated. I may not have done a great job of explaining, but I can assure you it was far easier and far more accurate than any of the other methods I tried! Thank you for everyone who chimed in, if it wasn’t for your suggestions I would be looking at sub-par starless data.
The lesson learnt: no two images are the same, different methods will always be required!
Had the same thing happen but in my case it identified a small planetary in a wide field image as a star - but only in the O3 where the signal was strong. It was fine in the Ha, S2, RGB, and Luminance. Attached is a snip from the O3. The indicated nebula was totally gone from the standard SXT result.
📷 PN_O3.png
I used the technique I normally do, which is the subtraction method described above to restore it, which worked fine.
Great to see you have found a solution. All variations on a theme I guess, but I have used successfully a method originally described by Adam Block. It basically uses the clonstamp tool to create a mask of the left-over nebulosity and then run SXT with that mask applied. It allows for very precise control and almost no visible remains. I’ve described it in more detail in my processing of the Crescent Nebula.
Tom Engwall · Aug 22, 2025, 12:41 PM
Are you running SXT on a linear or non linear images, I have had problems with SXT when running it on non linear and now only run it on linear images before applying NXT.
It depends on what it is making it’s mistakes on. Sometimes it works better on linear, sometimes on non-linear.
Jamie Macdougall · Aug 22, 2025, 12:43 PM
Aaron H. · Aug 22, 2025 at 07:57 AM
Jamie Macdougall:
John Stone · Aug 22, 2025 at 01:34 AM
then use PixelMath and subtract the stars from the original image to remove them again but leave the nebulosity alone.
That sounds like a good easy method! What’s the pixel math expression you’d use?
By default, unless the "unscreen" button is selected, it's a simple subtraction.
Run StarXTerminator, then undo on the main image, stamp out any problematic nebulosity on the extracted stars, then run "myimage - myimage_stars". You'll now have a new matching pair with your tweaks applied.Thanks!
Tom Engwall · Aug 22, 2025 at 12:41 PM
Are you running SXT on a linear or non linear images, I have had problems with SXT when running it on non linear and now only run it on linear images before applying NXT.
Linear
I have only used “unscreened” stars with this “cloning out the problematic nebulosity” technique and had good results with the following pixel math:
Expression: ~((~original image) / (~stars with nebulosity removed))
The only issue may be that with bright saturated stars, the “repaired” starless image may have very small slightly darkened spots at the core of where a bright star was removed. (this is due to the expression’s division of a small number by a similarly small number). This is easily cloned out on the repaired starless image if it causes any issues. I mostly used this method to recover small faint fuzzy galaxies that starX slopped out of the ‘starless’ and into the ‘stars’.
To satisfy my own curiosity, I will compare this above method with the “direct subtraction” method (not selecting the unscreen button).
Jamie Macdougall · Aug 21, 2025, 09:02 PM
I’ve used StarX on a drizzled narrowband image, and it is taking out some of the nebulosity. I’ve had this happen before, but it hasn’t been anything significant. In this instance, I’m going to be adding in separate RGB stars, not the original NB stars. I’d like to find a way of easily adding the removed nebulosity back into the starless image, whilst keeping it in the linear state. Is that possible?
For reference, this is the result after running StarX (zoomed in on the offending area):
I´ve noticed stuff like this.
Ive started very gently stretching the RGB with only stars in mind.
Then ill run a very light Noise X then Star X. Its seems to pick up less background like that. Then screen the stars back in, sometimes large overlay works better than unchecked and screen stars selected
If you are combining RGB stars with NB the blur your RGB stars and layers your NB over.. or at least thats what i do I might be mad ….
I sometimes have this problem when I run SXT on nonlinear images but it seems to always work on linear images. I usually apply SXT on linear images then apply NXT on the linear starless image.