Back to forums

Comparison of processes to separate stars in PI

Processing Deep Sky NGC7000 NGC Objects
16 replies•717 views
Which process / program do you use to separate stars and nebulae?
Single choice poll 67 votes
0% (0 votes)
45% (30 votes)
36% (24 votes)
3% (2 votes)
16% (11 votes)
You must be logged in to vote in this poll.
Daniel Arenas avatar
Daniel Arenas avatar
Yesterday I was processing my NGC7000 data (58 lights with 300 seconds of time exposure each, total 4.8 h) and in the non-linear process with PixInsight I wanted to compare different processes to separate the nebula and the stars. I usually did it with Starnet v1 but I decided that it would be interesting compare the results. So here you are some screenshots.

That was the starting image:



With Starnet v1:



With Starnet v2:



With the trial version (30 days) of StarXterminator:



It seems that, in this case and maybe not in others, the Starnet v2 process (free license) is doing a better job.

Which one do you use?

Kind regards and clear skies!
Helpful Concise Engaging
Sean van Drogen avatar
Sean van Drogen avatar
As your test sort of shows already its situational.
Tend to use both Starnet 2 and Starxterminator and see which one gives me the better result. As of yet I have not found the pattern of when Starnet 2 gives the better result vs Starxterminator and vice versa. Have not kept track but feels 50/50 at this point. Usually the brighter stars has more artifacts in Starxterminator but Starnet 2 sometimes leaves some stars in.

CS,
Sean
Daniel Arenas avatar
Daniel Arenas avatar
Sean van Drogen:
As your test sort of shows already its situational.
Tend to use both Starnet 2 and Starxterminator and see which one gives me the better result. As of yet I have not found the pattern of when Starnet 2 gives the better result vs Starxterminator and vice versa. Have not kept track but feels 50/50 at this point. Usually the brighter stars has more artifacts in Starxterminator but Starnet 2 sometimes leaves some stars in.


Yes, I think you are right.

The best, would be to have both StarNert2 and StarXterminator and try one and another in the image and use the best result. Nowadays, I have the 30-day-trial version of StarXterminator. Maybe in the future I'll pay for the license.
Stjepan Prugovečki avatar
Stjepan Prugovečki avatar
I have both. It really depends about situation which one gives better results. On my PC StarXterminator is somewhat faster, so I always use it first.
Manuel Peitsch avatar
Manuel Peitsch avatar
I agree with the situational use and relative benefits. I had however far better success with StarXterminator on linear images than with SN2.
CS
Manuel
Stuart Taylor avatar
Stuart Taylor avatar
Sean van Drogen:
As your test sort of shows already its situational.
Tend to use both Starnet 2 and Starxterminator and see which one gives me the better result. As of yet I have not found the pattern of when Starnet 2 gives the better result vs Starxterminator and vice versa. Have not kept track but feels 50/50 at this point. Usually the brighter stars has more artifacts in Starxterminator but Starnet 2 sometimes leaves some stars in.

CS,
Sean

I agree. I did my own comparison btw starnet2 and STX and decided on balance that the latter was better. But if I am oversampled and have flabby stars it does tend to produce a mottling in the starless image.

I actually think starnet2 is probably pretty much as good overall
Bob Lockwood avatar
Bob Lockwood avatar
As most everyone has said, I also use both Starnet-V2 and StarXterminator. Both seem to be inconstant and show different results on the same image. So I do both, and use the one that looks best.
Stuart Taylor avatar
Stuart Taylor avatar
this discussion prompted me to do a direct comparison on my latest image. I must admit the free option is looking clearly better in this particular case
Tommy Wlasichuk avatar
Tommy Wlasichuk avatar
Usually StarXterminator does a better job for me, but I often run both and pick the better of the two. Sometimes I’ll actually use both and combine them in gimp with a mask
Alicia Rossiter avatar
Alicia Rossiter avatar
I am very happy with StarXterminator and I plan to pay for the license as soon as the trial period ends. It was super easy to download and install in both PI and Photoshop. I did try several times to install Starnet V2 and it was a real nightmare, and as of today it does not show up in the new modules to add to PI  (yes, I did read all the blogs on how to make it work with a MacBook M1… ). The PI team does not support issues like this. 
Oh well!, enough with venting about Starnet!. 

Clear skies!
Alicia
Alex Woronow avatar
Alex Woronow avatar
Alicia…
While I had no trouble installing Starnet v2 (windows 10), I thought that it did not show up…but it did. StarNet v1 was under Object Recognition, but StarNet v2 is under "etc" in the PI menu. Maybe you missed it there? Anyway, good luck!

Alex W
Helpful
Die Launische Diva avatar
Die Launische Diva avatar
I haven't tried StarXterminator but given several comparisons between StarXterminator and Starnet v2, I can't justify paying just for a single plugin an amount of money which is a significant fraction of PixInsight's price.

I have great respect for Nikita, the developer of Starnet. Not for the price, but because he shared with the community the internals and theory behind Starnet. Other products just vaguely refer to AI and deep learning, which without any further details, are just catchphrases to my ears.

With the exponential advancements in machine learning, after a while, saying that a piece of image processing software uses deep learning without providing further details, will be as funny as if nowadays an image processing application advertises that it uses the fast Fourier transform.
Well Written Insightful Engaging
Dan Vranic avatar
Dan Vranic avatar
For me, StarX left too many blurred spots.
Sean van Drogen avatar
Sean van Drogen avatar
Just ran a comparison on my latest
Left: Starnet2 right: Starxterminator

Both have issues with Sadr, but clearly in this case Starnet2 gives the cleaner result
Stuart Taylor avatar
Stuart Taylor avatar
I sent my NGC281 comparison (posted above) to Russell (creator of Star XTerminator). He replied to say he was close to releasing an improved version and asked me to send him some images to try it out on. I'll report back.
Well Written Engaging Supportive
Eddie Bagwell avatar
Eddie Bagwell avatar
I was using the SXT free trial when SNv2 just came out and ran comparisons between the two and found that SNv2 gave better results in my images.
Danny Lee avatar
Danny Lee avatar
I use both. 

On my own data I find StarNet2 leaves me with a cleaner result when stars are extracted from a pre-stretched image.

Although StarNet2 doesn't work well for me on linear images, I have better linear results from StarXterminator and the resulting star mask looks better. At least for me.