StarXTerminator - New Russell Croman PS Action for Star Removal

[Xiga]

Hi guys

Just came across this tonight. It's similar to Starnet++, in that it uses AI to remove stars, but it's all done within PS and apparently it leaves far fewer artefacts behind. 

https://www.rc-astro.com/resources/StarXTerminator/

Personally, i find $60 to be a bit steep for what basically amounts to just a single action (especially when you can get a whole suite of actions for a fraction of that price) but it certainly seems to be very effective. It's also very early days in the development, and Russell promises that it will only get better as he continually improves it.

Here is a video of someone doing some early tests of it. 

 

[mackiedlm]

I've done some trials with this, head to head with starnet++ in PI and I have found that it is much much better. No halo artefacts where bright stars are removed - the main reason I gave up on starnet  was simply not there.

He has already updated the AI to address a couple of small issues.

Personally I think its a vast improvement.

But I agree 100% - it's a lot of cash for a one trick pony.

[tooth_dr]

Did a few tests myself. Seems better than Starnet but also a lot handier to use.  £45 seems reasonable enough if I’m going to use it on every image/part of image processing routine. 

[tooth_dr]

I've tried it on a few images now and I'm getting a much coarser grain with Star XTerminator compared to Starnett++.

Images below, any input welcomed.  I was using a 2600MC with an NBZ filter.

 

[Xiga]

38 minutes ago, tooth_dr said:

I've tried it on a few images now and I'm getting a much coarser grain with Star XTerminator compared to Starnett++.

Images below, any input welcomed.  I was using a 2600MC with an NBZ filter.

 

I haven't used it myself Adam, so i can't really offer much in the way of help, other than to say you should post your issue and images over in the CN thread (linked below). Russ is very active on it, and asks for images that don't work the best so he can improve the AI. 

https://www.cloudynights.com/topic/789257-new-ps-plug-in-for-removing-stars-starxterminator/page-7

[tooth_dr]

[tooth_dr]

12 minutes ago, Xiga said:

I haven't used it myself Adam, so i can't really offer much in the way of help, other than to say you should post your issue and images over in the CN thread (linked below). Russ is very active on it, and asks for images that don't work the best so he can improve the AI. 

https://www.cloudynights.com/topic/789257-new-ps-plug-in-for-removing-stars-starxterminator/page-7

Cheers Ciaran, posted on the forum.

[ollypenrice]

Thanks for the heads up. I get very mixed results from Starnet and also find that either it or my PC cannot handle the images from the ASI2600 I'm now processing. These are around 140meg. Does anyone know what kind of processing demands StarXterminator makes? I would spend £45 on it in a heartbeat if I were sure it would run.

I know it's quite a price but de-starring is so powerful when it works. I don't want to publish starless images, I want to de-star and re-star with smaller stars, which is dead easy in Ps. Put the starless image on the bottom, paste the linear original on the top, set the blend mode to lighten, and stretch the linear image till the stars appear at the level you like. This will improve star colour intensity by stretching them less, preserve their relative sizes as captured and let you keep them small. And it saves you the hassle of masking and star-reducing as well.

Olly

[emyliano2000]

I tried it too and I do like it more than starnet but compared to starnet it seems to leave the area behind the bright stars very, very soft and blurred. 

Emil

[Laurin Dave]

1 hour ago, ollypenrice said:

Thanks for the heads up. I get very mixed results from Starnet and also find that either it or my PC cannot handle the images from the ASI2600 I'm now processing. These are around 140meg. Does anyone know what kind of processing demands StarXterminator makes? I would spend £45 on it in a heartbeat if I were sure it would run.

I know it's quite a price but de-starring is so powerful when it works. I don't want to publish starless images, I want to de-star and re-star with smaller stars, which is dead easy in Ps. Put the starless image on the bottom, paste the linear original on the top, set the blend mode to lighten, and stretch the linear image till the stars appear at the level you like. This will improve star colour intensity by stretching them less, preserve their relative sizes as captured and let you keep them small. And it saves you the hassle of masking and star-reducing as well.

Olly

Hi Olly

I've tried it in Photoshop and it works on images from my ASI2600mc on my i7 machine, faster than Starnett in Pi...  as others have said  it does leaves very smooth patches where the star has been removed..   Were you aware that you can get a months free trial? 

Dave

[ollypenrice]

47 minutes ago, Laurin Dave said:

Hi Olly

I've tried it in Photoshop and it works on images from my ASI2600mc on my i7 machine, faster than Starnett in Pi...  as others have said  it does leaves very smooth patches where the star has been removed..   Were you aware that you can get a months free trial? 

Dave

I am now! Thanks Dave, that clinches it...

Olly

[gorann]

I downloaded it and tried it on the HSO image I just made from the present IKO data release (Cave Nebula). I had to use it on my new Mac where I have the latest OS (11.6) and PS-version. Would not work on my old Mac (OS10.10) with PS CS5 (which is still my favourite). The star removal appears to be virtually without any annoying artefacts (except the smooth area around the biggest stars but that must be unevitable).

Cheers, Göran

Here is the original

Here is the Star Xterminator version

Here it is with stars blended in (layer and curve in PS) at a much reduced level

Edited October 6, 2021 by gorann

[Simon Pepper]

I saw a tutorial on this and as people have mentioned it also does not have the halos. Question when you add the stars back do the halos come back? I ask because I actually use  starnett to clone stamp out this horrible halos. 
 

can it be used as a halo xterminator as well?

[gorann]

7 hours ago, Simon Pepper said:

I saw a tutorial on this and as people have mentioned it also does not have the halos. Question when you add the stars back do the halos come back? I ask because I actually use  starnett to clone stamp out this horrible halos. 
 

can it be used as a halo xterminator as well?

The halos of bigger stars remain, at least if you use it on a stretched image. So far I have used it late in the processing and brought back stars from a stretched but darkened (with a curve) image. I am still experimenting but it is clearly a very powerful tool.

[david_taurus83]

On 06/10/2021 at 08:47, ollypenrice said:

Thanks for the heads up. I get very mixed results from Starnet and also find that either it or my PC cannot handle the images from the ASI2600 I'm now processing. These are around 140meg. Does anyone know what kind of processing demands StarXterminator makes? I would spend £45 on it in a heartbeat if I were sure it would run.

I know it's quite a price but de-starring is so powerful when it works. I don't want to publish starless images, I want to de-star and re-star with smaller stars, which is dead easy in Ps. Put the starless image on the bottom, paste the linear original on the top, set the blend mode to lighten, and stretch the linear image till the stars appear at the level you like. This will improve star colour intensity by stretching them less, preserve their relative sizes as captured and let you keep them small. And it saves you the hassle of masking and star-reducing as well.

Olly

Hi Olly. Many thanks for this tip! For the last couple of years I've always struggled with adding stars back in. Starnet sometimes makes a mess of the stars and creates a hatch pattern on the bright ones. I used to use Pixelmath to simply add them back in and was never happy with the result so I've largely avoided using Starnet to process. I've just tried your method and it's so easy to control them! Thanks again! 🙂

[michael.h.f.wilkinson]

Very interesting. I have students helping develop a new tool (non-AI but using morphological connected filters) to do something very similar.  Would be interesting to see the differences side by side (speed wise, there is no contest, ours beats Starnet++ by a mile)

[gorann]

10 hours ago, michael.h.f.wilkinson said:

Very interesting. I have students helping develop a new tool (non-AI but using morphological connected filters) to do something very similar.  Would be interesting to see the differences side by side (speed wise, there is no contest, ours beats Starnet++ by a mile)

On my three years old  Mac powerbook it takes about a minute for Star Xterminator to do its job on a 150 Mb ASI2600MC image. So very fast.

[michael.h.f.wilkinson]

11 minutes ago, gorann said:

On my three years old  Mac powerbook it takes about a minute for Star Xterminator to do its job on a 150 Mb ASI2600MC image. So very fast.

Haven't compared to Star Xterminator, only to Starnet++ (which I find dreadfully slow). Ours should be able to go up to 30 Mpixel per second on a quad-core machine

[gorann]

1 hour ago, michael.h.f.wilkinson said:

Haven't compared to Star Xterminator, only to Starnet++ (which I find dreadfully slow). Ours should be able to go up to 30 Mpixel per second on a quad-core machine

Yes, Starnet++ is much much slower. I usually start that process before I have to do something else, like taking the dog for a walk. I have not clocked Star Xterminator but I can do it when I will use it later today.

[Xiga]

How do people find StarXterminator's handling of Newtonian diffraction spikes? 

[gorann]

20 hours ago, michael.h.f.wilkinson said:

Very interesting. I have students helping develop a new tool (non-AI but using morphological connected filters) to do something very similar.  Would be interesting to see the differences side by side (speed wise, there is no contest, ours beats Starnet++ by a mile)

I just clocked it and Star XTerminator takes 45 s for a 150Mb image on my Mac Powerbook. Not even time enought to go to the loo😉

[ONIKKINEN]

2 hours ago, Xiga said:

How do people find StarXterminator's handling of Newtonian diffraction spikes? 

Only just recently started experimenting with starnet and StarXterminator but i would slightly favour the latter with my tries so far.

 

Bright stars still have spikes, especially if the star is close to any nebulosity, but it looks like there are fewer remaining artifacts compared to starnet++.

[tooth_dr]

It's been while since I've used it, and I thin StarXTerminator has probably has some improvements over the past couple of months.  Has anyone else noticed any differences?

72 minutes of Ha on the Cave Nebula:

[AKB]

32 minutes ago, tooth_dr said:

Has anyone else noticed any differences?

Just halfway through an evaluation, so I only know what it’s like at the moment.  Certainly fewer artefacts than Starnet++, but perhaps a slightly softer result?

Is your example using linear or stretched data?

Tony

[tooth_dr]

6 minutes ago, AKB said:

Just halfway through an evaluation, so I only know what it’s like at the moment.  Certainly fewer artefacts than Starnet++, but perhaps a slightly softer result?

Is your example using linear or stretched data?

Tony

Not sure, but agree def less artifacts.  My example was done on a stretched TIFF out of APP