Comet 2022/E3 (ZTE) processing difficulty

[Ouroboros]

Hello, I’m mid-processing the recent comet in Pixinsight. I got this image after integrating the comet aligned images. I have obviously not rejected enough of the  trailing stars.  My pixel rejection was clip low, clip high pixels and low range, which appear to be the default settings.

I carried out the comet align process on the debayered light files, mainly because that’s what was used in two video tutorials I watched.  Is that the correct procedure?     My instinct would have been to use the registered light files. 

Any ideas how I might improve this?  

[Fegato]

I've only done this once properly, so I'm no expert on Comet processing, but I think yours is a fairly normal result after CometAlign. 

I got a pretty good result in January on this Comet by following Adam Block's method, removing stars with StarXterminator from each sub before the CometAlign (you could use StarNet2, but StarXterminator has a batch mode now which is useful). This took a long time obviously, and there were still some star trails around when I stretched hard to bring out the Comet tail, but less impact. I also found a method on Youtube to remove these residual star trails by AstroEdy from Astrobin - it's in Italian, but I managed to work it out - basically involves using ABE while masking the Comet....  a bit fiddly but it worked for me.

End result plus youtubes below...

https://astrob.in/wufh80/0/

 

 

 

Edited March 4, 2023 by Fegato
correct link

[Ouroboros]

Thanks, @Fegato.  Yes, my further searches on-line had led me to the methods you describe. I was thinking of acquiring StarXterminator anyway.  Although I do have StarNet++ I’d need to set up a bespoke batch process to do that. Not something I’m over confident in. So, StarXterminator has the edge there it seems. 

As I understand it the problem I and others are seeing occurs when a comet is moving very little between each sub. That’s certainly confirmed by the fact that a couple or so years back I processed a comet using an earlier version of Pixinsight’s comet align and it worked extremely well. But in that case the comet was moving at a fair old pace across the screen, so integration did a much better job at rejecting the stars.

An alternative home-baked alternative to star removal has occurred to me and that is to split the subs into two or more sets with larger time delays between adjacent subs. Integrate each set separately and then combine at the end.  I’ve not tried it, but it might work.  

[The Lazy Astronomer]

Same as the other guy, only ever done it once, so don't necessarily take my advice!!

I used Starnet2 to remove stars from each sub - it's actually quite simple: just create a process container with a stretch, Starnet2, and an inverse stretch*. Then create an image container, load your subs into it and then apply your process container to your image container. 

*you could forgo the stretches and use the linear mode of Starnet2, but I find better results are achieved on stretched images. 

This, along with some rejection stacking, removed pretty much all of the star trails. There were still some ghostly residuals on the larger stars, which were largely removed using masks and MMT (aka the Adam Block "special sauce" - he actually recommended MLT, but I found MMT worked better for me).

[Ouroboros]

Ah …. you’ve given me something more to contemplate there, @The Lazy Astronomer. 

[Elp]

If you're manually processing addressing something like this is straightforward. What I did was mask/delete out all the background leaving just the comet, then combine with a starfield stack. On the comet only image any residual star trails can be minimised with stringent kappa sigma rejection when stacking, and removed completely with image subtraction, anything remaining can be motion blurred out.

[Ouroboros]

3 hours ago, Elp said:

If you're manually processing addressing something like this is straightforward. What I did was mask/delete out all the background leaving just the comet, then combine with a starfield stack. On the comet only image any residual star trails can be minimised with stringent kappa sigma rejection when stacking, and removed completely with image subtraction, anything remaining can be motion blurred out.

😀  “….. something like this is straightforward …” Well, if you say so. Having played around with this data for several hours today I’m not so sure.  Thanks for your input though. 

 

[Ouroboros]

11 hours ago, The Lazy Astronomer said:

Same as the other guy, only ever done it once, so don't necessarily take my advice!!

I used Starnet2 to remove stars from each sub - it's actually quite simple: just create a process container with a stretch, Starnet2, and an inverse stretch*. Then create an image container, load your subs into it and then apply your process container to your image container. 

*you could forgo the stretches and use the linear mode of Starnet2, but I find better results are achieved on stretched images. 

This, along with some rejection stacking, removed pretty much all of the star trails. There were still some ghostly residuals on the larger stars, which were largely removed using masks and MMT (aka the Adam Block "special sauce" - he actually recommended MLT, but I found MMT worked better for me).

OK, I’ve tried this with Starnet2….. which was a bit of a learning curve …. and took my MacBook 6 hours to chunk through.  But I  learnt a bit about PixelMath and Image Containers etc. which was good. 

I think it’s a slight improvement.  I guess I need to do some masking now. 

My original version:

Using StarNet to remove stars in comet aligned subs and then integrated:

 

 

[Varavall]

On 02/03/2023 at 23:47, Ouroboros said:

Any ideas how I might improve this?  

Being a cheapskate I process only in free software and for comets I do the following process.

 

Stack twice or maybe 5 times if you are using rgb filters. 

First stack on the stars in ASTAP, then put the stacked image through Starnet++.

Second, go to GIMP open the starless and starry images as layers, then with the starless image top make it "subtract". That gives you the star field without the comet which v=can be tweaked for the stars.

Third, run all the subs through Starnet++ (I create a batch file and use the CLI version) to leave just the comet.

Fourth, stack on the starless comet subs in ASTAP with the manual alignment process. If using RGB filters, then just stack the subs for each filter separately, and save that with a name so you know which colour filter.

Fifth, open as layers all the starless comet stacked images with the images in second step above.  Align the comet images, then create a new layer from visible for just the comet and tweak for personal preference, then move that layer above the star field layer and make it "screen", then make new layer from visible.

Sixth, tweak image until satisfaction is achieved.

Simples!

The result which still needs some work on a few artefacts and gradients

[Elp]

13 hours ago, Ouroboros said:

😀  “….. something like this is straightforward …” Well, if you say so. Having played around with this data for several hours today I’m not so sure.  Thanks for your input though. 

 

Im guessing you're trying to get the comet only stack to look like something like this:

This is around 2 minutes manual work and can be improved upon further with better care taken masking sharpening layering etc. Its just a demonstration of why I prefer manual processing over automated software.

Adam Block goes over it via the first video listed above by @Fegato around 19-20 minutes in. If you've got access to PS I'll detail how to do it. My first comet image took a few tries, this process was something I was already doing to my DSO images when creating synthetic flats, altered slightly to apply to comets.

 

[Ouroboros]

Yes, Adam Block’s detailed videos are very good, and I’ve watched the relevant section in the video above through a couple of times. In this instance his purpose is to show what can be done to remove star residuals, but not actually how to do it in detail. Anyway, thanks for your input, @Elp.  

[Ouroboros]

3 hours ago, Varavall said:

Being a cheapskate I process only in free software and for comets I do the following process.

 

Stack twice or maybe 5 times if you are using rgb filters. 

First stack on the stars in ASTAP, then put the stacked image through Starnet++.

Second, go to GIMP open the starless and starry images as layers, then with the starless image top make it "subtract". That gives you the star field without the comet which v=can be tweaked for the stars.

Third, run all the subs through Starnet++ (I create a batch file and use the CLI version) to leave just the comet.

Fourth, stack on the starless comet subs in ASTAP with the manual alignment process. If using RGB filters, then just stack the subs for each filter separately, and save that with a name so you know which colour filter.

Fifth, open as layers all the starless comet stacked images with the images in second step above.  Align the comet images, then create a new layer from visible for just the comet and tweak for personal preference, then move that layer above the star field layer and make it "screen", then make new layer from visible.

Sixth, tweak image until satisfaction is achieved.

Simples!

The result which still needs some work on a few artefacts and gradients

Nice! Just shows what can be done with freebie software.  

[Ouroboros]

After a bit more fiddling I ended with this. Not fantastic. I have done much better with comet photos in the past. This one was the most problematic. Maybe I have to accept that the data was just was not the best. The comet was very much past its best 10 days ago.  Thanks to all of you and particularly  @Fegato and @The Lazy Astronomer whose method with Starnet++ I followed - well, kind of.  I’d like to make the background darker but the comet is very faint. 

[Ouroboros]

I can see more detail in the trailing stars version.