Blue Halo's(?) Dots left after StarxTerminator Pixinsight - Forums

Topic starter

Feb 21, 2023

Dear forum,

I'm experiencing some trouble processing a picture. After I use the Starxterminator I'm left with an almost starless picture, and a Star Image. But when I look at my Starless picture I see blue dots almost everywhere where the stars were, and even green ones. When I zoom in to me it looks like small halo's from where the stars used to be.

Her are a few examples from the picture im processing. Is there anyone who could lead me the good way to get rid of these blue spots, or to tell me where I maybe went wrong in the first place?

Much appreciated!

Melvin

jewzaam

Feb 21, 2023

These look like halos I get from my l-extreme filter, SXT doesn't remove those either. I don't think this would change the result, but try extracting each channel then run SXT on each individually. The way I deal with the l-extreme halos is to repair them in Affinity Photo using infill painting. Also you could try stretching less on the stars and see if the halo is more controlled.

Helpful Concise Supportive

andrea tasselli

Frequent contributor

Feb 21, 2023

If they exist only one channel then paint them over with the nearby background and you'll be done with them once overlaid with them stars back in.

Bob Lockwood

Feb 21, 2023

One thing to keep in mind is SXT removes stars, not halos. As far as STX is concerned, the halos are nebulas.
Find out why you are getting halos, could be the filters.

Minh Lết

Feb 21, 2023

Use a clone stamp and stamp them out or use spot remover in photoshop

AstroNOM

Topic starter

Feb 21, 2023

Thanks so far for the replies, I will get to work with all your input! If any better I will post new pictures.

More input always welcome

Willem Jan Drijfhout

Feb 21, 2023

Have you tried BlurXTerminator first with an aggressive setting setting to reduce halos, and then do SXT?

Tim McCollum

Feb 21, 2023

Not an ideal solution, but in PI open the color saturation process, hover over the blue while watching the hue line to see where the exact color is, then place a point there. Next place 4 more points, 2 either side of the center point. Adjust the spacing from the center point as a bandwidth setting, open real time preview and drag the center point down to kill the color. You could also make a blue color mask and then using curves kill the blue as well, hech maybe also with the blue mask do SCNR to it.
These are all just work arounds to save the data you have vs figuring out the exact cause of the halos.
Tim

Helpful

AstroNOM

Topic starter

Feb 21, 2023

Willem Jan Drijfhout:
Have you tried BlurXTerminator first with an aggressive setting setting to reduce halos, and then do SXT?

No not yet, but I will definitely try that! Im now trying to perform SXT for all filters seperately before combining the filters into a coloured image. Might help too.

AstroNOM

Topic starter

Feb 21, 2023

Tim McCollum:
Not an ideal solution, but in PI open the color saturation process, hover over the blue while watching the hue line to see where the exact color is, then place a point there. Next place 4 more points, 2 either side of the center point. Adjust the spacing from the center point as a bandwidth setting, open real time preview and drag the center point down to kill the color. You could also make a blue color mask and then using curves kill the blue as well, hech maybe also with the blue mask do SCNR to it.
These are all just work arounds to save the data you have vs figuring out the exact cause of the halos.
Tim

Thanks Tim for your effort. I will have to dive into this one with a bit more time, but will also definitely try your way to see how it goes!

Well Written Respectful

Alan Brunelle

Feb 21, 2023

StarXterminator does not always work to perfection. I get those as well sometimes and sometimes not.

Recommendations to infill or clonestamp the halos out generally work well. But beware, not always. The halos, and sometimes partial stars left in the "de-starred" image are part of the information within the complete image. You can prove that by recombining the star image to untouched, de-starred image and confirm an image faithful to the original. The implication (and reality) is that in filling and clonestamping to remove the halos and partial stars in the starless image will remove the information that would have completed the image after the stars are returned. This does not cause big issues in weak halos left in the starless image, and indeed can be a help to removing halos. But some spikes and partial stars removed can skew the color and the shape of stars after the recombination. So beware the fixes proposed might required additional work after recombination to fix the issues created in that process.

In my

The Coma Cluster - Added Data - Pushing the Limits with BlurX-, StarX-, and NoiseX-Terminators image. StarXterminator left a good number of stars in the starless image. Some of the biggest and brightest stars. Very perplexing. Yet removed many of the very small background galaxies, even the small smudge galaxies that are clearly not round. So more learning for the AI library needs to be done? Well, maybe, or maybe not. The new library works very much better than earlier versions, even on this image. It just may be that StarXTerminator cannot be expected to do everything. But if improvements can be had, the Coma Cluster might be a place to teach the AI.

Helpful Insightful Respectful

AstroNOM

Topic starter

Feb 21, 2023

Alan Brunelle:
StarXterminator does not always work to perfection. I get those as well sometimes and sometimes not.

Recommendations to infill or clonestamp the halos out generally work well. But beware, not always. The halos, and sometimes partial stars left in the "de-starred" image are part of the information within the complete image. You can prove that by recombining the star image to untouched, de-starred image and confirm an image faithful to the original. The implication (and reality) is that in filling and clonestamping to remove the halos and partial stars in the starless image will remove the information that would have completed the image after the stars are returned. This does not cause big issues in weak halos left in the starless image, and indeed can be a help to removing halos. But some spikes and partial stars removed can skew the color and the shape of stars after the recombination. So beware the fixes proposed might required additional work after recombination to fix the issues created in that process.

In my

The Coma Cluster - Added Data - Pushing the Limits with BlurX-, StarX-, and NoiseX-Terminators image. StarXterminator left a good number of stars in the starless image. Some of the biggest and brightest stars. Very perplexing. Yet removed many of the very small background galaxies, even the small smudge galaxies that are clearly not round. So more learning for the AI library needs to be done? Well, maybe, or maybe not. The new library works very much better than earlier versions, even on this image. It just may be that StarXTerminator cannot be expected to do everything. But if improvements can be had, the Coma Cluster might be a place to teach the AI.

Thanks for your input Alan. I saw your picture and I respect the work that you put into it. Amazing photo! I have to say that for me there is a lot of learning in PixInsight and i still have to develop a solid workflow considering bias frames, pre processing and pos processing. Eventually ill get to the answer of thes problems, the more I learn the better. Thats why, all the input is welcomed and much appreciated!

Respectful Supportive