Pixinsight Losing detail in the Cygnus loop that I can see on the starless image when recombining

[SiD the Turtle]

Hi all,

Having trouble processing the Cygnus loop, I feel like I'm losing data whenever I stretch, or try to reduce the stars. Would appreciate advise, all thumbnails are STF-auto-stretched for visibility:

Here's my raw image, after running WBPP. Pretty decent, I have some backgrounds to deal with, and way too many stars.

After dynamic background extraction:

Now I've used StarXterminator:

Then HDRMT and NoiseXterminator. This is the starless image (perhaps HDRMT is a little agressive):

This, IMO is pretty special, I see plenty of detail here I'd like to keep. So I use PixelMath [~((~starless)*(~stars)) to recombine:

Now the detail is lost in the noise of the stars.

Doing the best I can with a stretch, then using 'Star Reduction' from Bill Blanshan and Mike Cranfield, which has worked well for me before, I can get something like this:

I feel like I'm missing a lot of the detail that was in the starless images. Any ideas? More stretching? A different star reduction tool?

 

Edited July 29 by SiD the Turtle

[Vroobel]

From my experience, I can highly recommend watching carefully PixInsight's tutorials about their 'GradientCorrection' (even a few times...) and leaving other methods of gradient removal. Proper values/sliders settings work wonders. 

 

Edited July 29 by Vroobel

[SiD the Turtle]

Thanks. I took another stab in a different direction:

• Was a bit less lazy in the picking of dynamic background extraction points.
• Didn't use HDRMT.
• Used a mask and goosed the saturation of the nebulosity separate to the stars.
• Used a good-old fashioned histogram transform to trip out the stars.

The output I'm happy enough with for now:

 

[TiffsAndAstro]

45 minutes ago, SiD the Turtle said:

Thanks. I took another stab in a different direction:

• Was a bit less lazy in the picking of dynamic background extraction points.
• Didn't use HDRMT.
• Used a mask and goosed the saturation of the nebulosity separate to the stars.
• Used a good-old fashioned histogram transform to trip out the stars.

The output I'm happy enough with for now:

 

It's a very very nice image

[Fegato]

When do you take the stars out?  

I tend to stretch the whole image lightly, to the point where the stars are about where I think I want them.  I then remove the stars and work on the subject, stretching further, noise reduction etc. Then when I'm happy with that, I'll add the stars back. You can also brighten or reduce intensity of the stars only image, as well as increase colour saturation, before you add back. If you follow this sort of approach, there is no need to do any star reduction on the combined image.

BTW - what was your intention in using HDRMT?  I only use this in extreme cases of bright areas completely masking details, and will often use masks if I do, to protect other areas.  I think it's added a lot of artifacts, blotching etc. to your image, and also makes it look less "natural" (whatever that means!). As a result, I prefer your third image above to the fourth one...

 

[Clarkey]

I normally use starX, but in images such as this I find it destroys a lot of detail. With careful use of curves you can stretch the darker areas of the image without overdoing the star stretch. Sometimes it give a better result.

If you do have to remove the stars, try using the image blend script to recombine, rather than the simple pixelmath statement. There are a number of options that allow the stars to be returned in a less intrusive way. You can also alter the intensity of the blended layers.

[Whistlin Bob]

15 hours ago, Fegato said:

I tend to stretch the whole image lightly

+1, this seems to work well for me too. I think the recommendation is to use StarX on the linear image, but I find it performs a little better (less artefacts on big stars, little to no lost detail) with the approach you describe.

[SiD the Turtle]

On 30/07/2024 at 07:56, Fegato said:

When do you take the stars out?  

I tend to stretch the whole image lightly, to the point where the stars are about where I think I want them.  I then remove the stars and work on the subject, stretching further, noise reduction etc. Then when I'm happy with that, I'll add the stars back. You can also brighten or reduce intensity of the stars only image, as well as increase colour saturation, before you add back. If you follow this sort of approach, there is no need to do any star reduction on the combined image.

BTW - what was your intention in using HDRMT?  I only use this in extreme cases of bright areas completely masking details, and will often use masks if I do, to protect other areas.  I think it's added a lot of artifacts, blotching etc. to your image, and also makes it look less "natural" (whatever that means!). As a result, I prefer your third image above to the fourth one...

 

Early on, after dynamic background extraction, while the image is linear, as StarXterminator can handle it in a linear state, apparently. Indeed, I found it much easier to just use stretching on the stars on their own and add them back, rather than damaging the nebulosity trying to deal with them on the main image.

Edit: on HDRMT, I used it for an image of Bode's and the Cigar Galaxy as I am using far too wide a FoV scope for them, I wondered if it'd help here but honestly it looked like those horrendous sunset shots people do where the clouds end up way too... sharp is probably the best way to describe it.

Edited August 7 by SiD the Turtle

[SiD the Turtle]

On 30/07/2024 at 22:45, Clarkey said:

I normally use starX, but in images such as this I find it destroys a lot of detail. With careful use of curves you can stretch the darker areas of the image without overdoing the star stretch. Sometimes it give a better result.

If you do have to remove the stars, try using the image blend script to recombine, rather than the simple pixelmath statement. There are a number of options that allow the stars to be returned in a less intrusive way. You can also alter the intensity of the blended layers.

Thanks, I'll have a play with that for the next image.

[Fegato]

38 minutes ago, SiD the Turtle said:

Early on, after dynamic background extraction, while the image is linear, as StarXterminator can handle it in a linear state, apparently. Indeed, I found it much easier to just use stretching on the stars on their own and add them back, rather than damaging the nebulosity trying to deal with them on the main image.

 

Yes StarX does handle linear images - but in the background it does a stretch to achieve it, so I guess there's not going to be much difference as long as any stretch before star removal is a simple one. Having said that, RC does recommend using on a linear image, so I guess you're doing the right thing!

Quote from RC's site:

StarXTerminator is trained on images stretched using a simple midtones transfer function (MTF). When processing linear images, StarXTerminator internally performs such a stretch automatically, then precisely reverses it after processing to return the image to a linear state.

[ollypenrice]

Keep it simple.

You have your finally stretched starless where you want it. It looks pretty convincing to me, though I suspect the lower quarter of the background maybe just a bit bright but, whatever...

You now want to add the stars. Only the stars. Not the glow around the stars. Not the background around the stars. Just the stars. You can use Pixelmath, try, repeat, try again, etc etc. Or you could use Photoshop on the star layer and tweak it in Levels or in Curves and see what you were doing in real time. So easy. So nice.

Sorry, just me banging on again...

Olly