WR-134

[Rodd]

I have been trying to bring out the OIII shell for a long time. I finally succeeded to reveal a complete circle. How?  Two words that are new for me….starless processing.  It still needs allot of work  but the next hurdle might take another year to work out.  Bicolor images often give me trouble.

TOA-130 and asi 1600. 133 300 sec Ha and 114 300 sec OIII

[teoria_del_big_bang]

Thats a great image.
Yes removing the stars and processing the nebula and then adding the stars back in works a treat on many images.
It is something I have just started using as well and if it works then for me I think it can create some wonderful images that allow you to bring out all the detail in the nebula leaving the nebula as the main focal point instead of overblown stars attrating your attention.
For me it seems to work far better than using masks.
Still a bit of a learning curve for me too and part of the issue can be getting a starless image I am happy with and not one with unwanted artefacts. I also like the idea of being able to add RGB stars to NB data.

Steve

[Budgie1]

4 hours ago, Rodd said:

I have been trying to bring out the OIII shell for a long time. I finally succeeded to reveal a complete circle. How?  Two words that are new for me….starless processing.  

No stopping you now, Rodd! 

If you're feeling like trying another trick, what about Generalised Hyperbolic Stretch? 

1 hour ago, teoria_del_big_bang said:

For me it seems to work far better than using masks.
Still a bit of a learning curve for me too and part of the issue can be getting a starless image I am happy with and not one with unwanted artefacts.

I still use masks or colour masks on the starless background, especially for the main target, as it allows more control. 

Using both StarXT & StarNet2, I've found that if one tool is leaving artefacts or removing more than you want it too, then try the other. One always seem to give a better removal than the other on different images.  

[Rodd]

3 hours ago, teoria_del_big_bang said:

Thats a great image.
Yes removing the stars and processing the nebula and then adding the stars back in works a treat on many images.
It is something I have just started using as well and if it works then for me I think it can create some wonderful images that allow you to bring out all the detail in the nebula leaving the nebula as the main focal point instead of overblown stars attrating your attention.
For me it seems to work far better than using masks.
Still a bit of a learning curve for me too and part of the issue can be getting a starless image I am happy with and not one with unwanted artefacts. I also like the idea of being able to add RGB stars to NB data.

Steve

Thanks Steve. You cover a lot of ground in your post.  So much to learn

[teoria_del_big_bang]

1 hour ago, Budgie1 said:

Using both StarXT & StarNet2, I've found that if one tool is leaving artefacts or removing more than you want it too, then try the other. One always seem to give a better removal than the other on different images.  

Yes I also tend to use both and choose the best, often they are pretty indistinguishable but often one does a better job than the other.

1 hour ago, Budgie1 said:

I still use masks or colour masks on the starless background, especially for the main target, as it allows more control

I have not had much experience as yet in this bit yes I still do use masks just trying to eliminate having to use the star masks when stretching to stop them getting blown and overtaking the image.

Steve

[Budgie1]

48 minutes ago, teoria_del_big_bang said:

I have not had much experience as yet in this bit yes I still do use masks just trying to eliminate having to use the star masks when stretching to stop them getting blown and overtaking the image.

The GHS script I linked for Rodd above is good at controlling the stars during stretching, but watch the videos for the basic settings etc, as it much more involved than a normal Histogram Transformation stretch. 

Something else, which I haven't tried yet but could prove useful, is to do the star extraction when the image is still linear, then stretch the stars & background separately. Just a theory that came to mind while at work but I'll give it a try tonight and see how it turns out.  

[Rodd]

2 hours ago, Budgie1 said:

If you're feeling like trying another trick, what about Generalised Hyperbolic Stretch? 

I have been meaning to learn this--but I can't find it in Pixinsight.

[Rodd]

15 minutes ago, Budgie1 said:

Something else, which I haven't tried yet but could prove useful, is to do the star extraction when the image is still linear, then stretch the stars & background separately. Just a theory that came to mind while at work but I'll give it a try tonight and see how it turns out.  

I know starnet does not work on linear images--I am not familiaer with Staenet2 or starxterminator.  There should be a way to do it though.  Perhaps a pixel math formula that uses an in-function screen stretch and then removes the stars.  I like this Idea. 

[teoria_del_big_bang]

10 minutes ago, Budgie1 said:

Something else, which I haven't tried yet but could prove useful, is to do the star extraction when the image is still linear, then stretch the stars & background separately. Just a theory that came to mind while at work but I'll give it a try tonight and see how it turns out.  

That's pretty what I have been doing, getting a starless image and a star image, whilst linear. I tried doing it to the SII, Ha and OII separate then combining the 3 starless images and also on the already combined, but still linear, colour image. I tend to think doing it in the monos stage was better (obviously if not OSC) but not sure that was true I need to do some more processing.
But I think in future if at all possible as I take mostly NB images I will always try to take some RGB data for the stars. It shouldn't take long to get good enough data for the stars as they can be pretty quick exposures and can wait until a near moonless night (although I see very few clear moonless nights last year or two 😞 ) .

I think you have to be a tad careful though if any brightest stars in the fov not to end up with halos around them if glow from the brighter stars is stretched far when doing the nebula processing and masks are probably needed for the brighter stars anyway or needs some way of removing the glow on the nebula image, or a halo removing process after the recombination which is not always adequate.

As I say still learning so maybe all this is 2nd nature to some 🙂 

Steve

[Rodd]

7 minutes ago, teoria_del_big_bang said:

hat's pretty what I have been doing, getting a starless image and a star image, whilst linear.

Now for the million dollar question....."How?"

[The Lazy Astronomer]

13 minutes ago, Rodd said:

I have been meaning to learn this--but I can't find it in Pixinsight.

GHS appears under utilites>script once installed. Link to install is here: https://ghsastro.co.uk/links-2/

Also a few video tutorials on how to use it - I'd recommend the recent one from the astro imaging channel as it's essentially a walk through from both of the script creators.

[The Lazy Astronomer]

4 minutes ago, Rodd said:

Now for the million dollar question....."How?"

Starnet v2 does linear de-starring

[Rodd]

6 minutes ago, The Lazy Astronomer said:

GHS appears under utilites>script once installed. Link to install is here: https://ghsastro.co.uk/links-2/

Also a few video tutorials on how to use it - I'd recommend the recent one from the astro imaging channel as it's essentially a walk through from both of the script creators.

I have trouble getting scripts in PI--they are never in the repository and adding them doesnt seem to do anything.  

[vlaiv]

4 minutes ago, The Lazy Astronomer said:

Starnet v2 does linear de-starring

But still uses only 16bit data format?

[Rodd]

6 minutes ago, vlaiv said:

But still uses only 16bit data format?

Do you mean for heh stars of for both the starless and starsonly images?  It would be unfortunate to turn a 32 point floating image into a 16 bit image

Edited June 17, 2022 by Rodd

[vlaiv]

Just now, Rodd said:

Do you mean for heh stars of for both the starless and starsonly images?  It would be unfortunate to turn a 32 point floating image into a 16 bit image

It looks like it is necessary. I tried running Starnet v2 on 32bit image and it complained and refused to work.

[teoria_del_big_bang]

18 minutes ago, Rodd said:

Now for the million dollar question....."How?"

As @The Lazy Astronomer says Starnet2 does this and also StarXTerminator.

What turning it into a 16 bit image does I am not sure, @vlaiv can you enlighten us ? 

Steve

[Rodd]

19 minutes ago, vlaiv said:

It looks like it is necessary. I tried running Starnet v2 on 32bit image and it complained and refused to work.

How significant do you see this?  Also--I run Strnet V1 on 32 floating images

Edited June 17, 2022 by Rodd

[Budgie1]

33 minutes ago, Rodd said:

I know starnet does not work on linear images--I am not familiaer with Staenet2 or starxterminator.  

Both of these allow you to work with a linear image, I can't see any reference to 32 or 16 bit requirements for StarXTerminator though.

StarXTerminator has a paid-for and trial version which can be obtained from here: https://www.rc-astro.com/resources/StarXTerminator/

For StarNet2, there's download links & an installation tutorial for PixInsight here: https://www.galactic-hunter.com/post/starnet2 

[vlaiv]

1 minute ago, teoria_del_big_bang said:

As @The Lazy Astronomer says Starnet2 does this and also StarXTerminator.

What turning it into a 16 bit image does I am not sure, @vlaiv can you enlighten us ? 

Steve

I'm not a fan of using 16bit format in any stage of image processing (apart from acquisition).

This is more pronounced these days with progress in CMOS arena. CMOS sensors have lower read noise and thus drive ever shorter exposures.

It is not uncommon for people to use 10-20 times shorter exposures with CMOS then it was common with CCD cameras. This means that signal levels are much lower - for same brightness of target, x20 shorter exposure will result in x20 lower signal value.

Take for example instance where you have faint signal in 10 minute exposure - say 250e. Same signal in 30s exposures will be x20 times lower or 12.5e on average. Problem with 16bit image format is that you can't record 12.5e - you can either record 12e or 13e and you introduced 0.5e just by virtue of using 16bit format.

32bit float point does not have this limitation. It can record numbers with much greater precision - and any introduced error is minimal (easily swamped by noise).

Short exposures "cram" all the signal into lower part of 0-16bit range, and although signal is fine grained after stacking - it will get rounded up and unnecessary errors will be introduced. This leads to "posterization" of faint parts of image on short exposures.

One way how to avoid this happening with StarNet V2 and still use linear data would be to linearly stretch your data - setting white point to just above brightest target parts. This will clip the stars, but we don't really care as stars will be removed. We just need starless version (we can later subtract starless version from full version to get unclipped stars with some math).

 

 

[vlaiv]

2 minutes ago, Budgie1 said:

Both of these allow you to work with a linear image, I can't see any reference to 32 or 16 bit requirements for StarXTerminator though.

Don't know about StarXTerminator - but this happens to me:

(this is with version x2 of starnet - not sure why it still says StarNet++ for GUI).

Readme also emphasizes this:
                           
 

                            StarNet++ v2.0.2
                                Finally!
                            
-> As before: Only TIF files with 16bit per channel are supported. <-

              -> Yes, AVX is still needed. Sorry! <-
   -> If the previous version did not work - this won't either. <-
   
          -> TO BE USED FOR ASTRO IMAGE PROCESSING ONLY <-

                            -> Have fun! <-

 

[vlaiv]

7 minutes ago, Rodd said:

How significant do you see this?  Also--I run Strnet V1 on 32 floating images

If you run PI script - then it is very likely that it converts to 16 bit for StarNet and then converts back to 32bit - in the process rounding happens so error is introduced regardless of data being 32bit in the end.

[Rodd]

9 minutes ago, vlaiv said:

t is not uncommon for people to use 10-20 times shorter exposures with CMOS then it was common with CCD cameras. This means that signal levels are much lower - for same brightness of target, x20 shorter exposure will result in x20 lower signal value.

Isn't total exposure time the important thing.  If you have a stack nade from 20 10 minute exposures (taken with CCD) and I have a stack of 3.3 hours made from 200 60 sec exposures using CMOS, with a lower read noise andmuch higher efficiency, wont teh stacks be comporable?

[Rodd]

9 minutes ago, vlaiv said:

If you run PI script - then it is very likely that it converts to 16 bit for StarNet and then converts back to 32bit

This should be verified.  I never process in TIFF format, so Starnet works with xsif.  

[Rodd]

19 minutes ago, Budgie1 said:

Both of these allow you to work with a linear image, I can't see any reference to 32 or 16 bit requirements for StarXTerminator though.

The strnet tool in my PI does not work with linear images--the results you get are way, way horrible.