A Pixinsight Learning Curve (Feat. Sh2-101) - First HSO Image

[The Lazy Astronomer]

Hello all,

A few months back, I captured a not insignificant amount of narrowband data on Sh2-101 (my first ever full narrowband image, too!), but upon processing, I wasn't getting the sort of results I was hoping for from my usual post-process software (Startools), so I decided it was finally time to give in and attempt to learn Pixinsight (something to do, at least, over the long, cloudy November I've had...).

I've spent the entirety of my free trial period so far endlessly reprocessing this same image, which is probably a good thing, as it has allowed me to directly compare my various efforts and techniques used (more on that below), and I have to say, Pixinsight is not nearly as scary and opaque as I think it's made out to be - after following a couple of Youtube videos (which netted my some limited success), I found the various tutorials and explanations from Light Vortex Astronomy, Jon Rista and Adam Block to be extremely helpful and I they got me on my way with the basics I needed, coupled with the use of some of the EZ processing suite.

Generally, I think I got on quite well with Pixinsight, but it's huge array of tools does leave me wondering whether I'm leaving something on the table when it comes to getting the most of the data, however I suppose this does leave plenty of room to grow and learn new techniques. I'm pretty much sold on it, so will shortly be purchasing a full licence, and I think a copy of Mastering Pixinsight might just be on the Christmas list this year. 

So with all that said, I thought I'd post up each of my processing attempts (6 in total) to show what I hope you'll agree is some good progress in the right direction. More than happy to take any further suggestions as well if anyone has some tips on how to improve further (+let me know your favourite ).

The total integration time 15h40m, made up of 4h05m SII, 5h10m Ha and 6h25m OIII (all 5min subs) with 6nm Astronomik filters and a 294MM. All images have been resampled to 50% original size.

ATTEMPT 1

DynamicCrop, DBE (badly done), EZ Denoise on Ha (badly done), straight HSO combine in PixelMath, HT, SCNR, various curves for colours, hue, saturation, etc., LRGBCombination (Ha as L), EZ star reduction, and some other minor tweaks. I thought it was ok at the time (and roughly on a par with what I was getting from Startools), but looking now, it's clearly not great...

 

ATTEMPT 2

Largely the same process as before, with some different curves adjustments. Some elements of this I preferred to the first one, although I ended up with these horrible yellowy-orange stars, and the colour in the Tulip itself looks worse to my eyes.

 

ATTEMPT 3

After a bit more reading up on some techniques, and playing around with different ways of stretching I ended up with this, which I felt was a significant improvement on the first 2 attempts, and revealed a lot more of the faint stuff I was looking for, but ended up far too yellow/orange/brown overall, and I still wasn't really happy with the results of EZ denoise.

 

ATTEMPT 4

After reading some Light Vortex tutorials on noise reduction, I had a stab at doing it manually instead of via EZ denoise, as well as using colour masks to selectively alter colours. I also resisted the temptation to use SCNR, to try and prevent essentially turning it into a bicolour image (plus I quite liked the greenish hues), however I subsequently decided the overall colour was far too gaudy, and I still wasn't quite happy with the noise reduction in parts, plus, I found the stars somewhat overwhelming.

 

ATTEMPT 5

Completely back to the drawing board on this one: read up one some noise reduction techniques from Jon Rista (TGV & 2 passes of MMT, followed by iterative stretching and ACDNR), played around with different PixelMath combinations to try and get me closer to the colours I wanted without having to do as much subjective adjustment later on. Also made use of Starnet to remove stars and add them back in from a less stretched version later on. This was going to be my final version, but when I started looking closer, I think I went a bit too far with the noise reduction, and although I like the stars, I don't think they sit well in the image (some dark ringing visible). Plus, I decided I wanted to try some deconvolution on the Ha layer as well.

 

ATTEMPT 6

Sixth and final attempt (well done for making it to the end, if indeed you are still reading!): backed off on the noise reduction, EZ Decon, previous attempt was used as the base for the colour layer, with some minor tweaks and stretched the star image a bit further which seems to have allowed them to sit a bit more naturally within the image.

In general for these last two attempts, I tried to resist the temptation to turn the colour up to 11, and tried for something a bit more neutral (for lack of a better word).

 

Sorry for the extremely long post, but thanks for reading, and let me know your thoughts/favourite!

Cheers!

[geeklee]

Great post and also great image(s).  A good example of how important processing is.  What may have appeared as a good start was soon eclipsed by much better versions that more knowledge allowed.  It's almost hard to believe to believe that first image is the same data as those latter ones.

I like various aspects of 3, 4, 5 and 6 - with 5 probably being my favourite.  I don't see too bad ringing but it's always easier seeing flaws in our own images! 🙂

With #4 perhaps some SCNR with the amount reduced to maybe 50-75 could have left some green in without being too much?

Super journey to read though.

[The Lazy Astronomer]

10 hours ago, geeklee said:

Great post and also great image(s).  A good example of how important processing is.  What may have appeared as a good start was soon eclipsed by much better versions that more knowledge allowed.  It's almost hard to believe to believe that first image is the same data as those latter ones.

I like various aspects of 3, 4, 5 and 6 - with 5 probably being my favourite.  I don't see too bad ringing but it's always easier seeing flaws in our own images! 🙂

With #4 perhaps some SCNR with the amount reduced to maybe 50-75 could have left some green in without being too much?

Super journey to read though.

Thanks, glad you enjoyed it ☺

You're right about number 5, I can't really see the artifacts now - I think maybe a combination of resampling and jpeg compression may be helping me there!

[Scooot]

I like using Pixinsight and enjoyed reading how it developed too. Lovely end result.

[scotty38]

Love the images and the process taken. I’ve just started looking at masks myself but so far only used the range mask. Will have to look at the colour masks too and see how they work for me. 

Edited December 7, 2021 by scotty38

[The Lazy Astronomer]

2 hours ago, scotty38 said:

Love the images and the process taken. I’ve just started looking at masks myself but so far only used the range mask. Will have to look at the colour masks too and see how they work for me. 

Thanks! For the colour masks, I used the ColorMask script and also the GAME script, which essentially allows you to draw custom masks of various shapes and sizes.

[scotty38]

Great thanks will have a look at them

[Clarkey]

I think my personal choice would be number 3. Nice images though. If you like me you end up doing multiple different versions - but still think you can do better!

I tried PI for a while and although it was quite good, I just did not have the time to learn it well enough to do it justice. I'll be sticking to APP, Affinity and Startools for now.

[The Lazy Astronomer]

38 minutes ago, Clarkey said:

If you like me you end up doing multiple different versions - but still think you can do better!

Yeah, tell me about it! Every time l got to the end of the process, I'd think "this is the one", then look again the following day and think I could do better! I am officially done with this one now though - almost sick of having to look at it 😅