M 81 and M 82

User images created with StarTools.
decay
Posts: 46
Joined: Sat Apr 10, 2021 12:28 pm

Re: M 81 and M 82

Post by decay »

Hi Stefan,

sorry, please ignore my last post and my question!
Meanwhile I recognized, you used the "Spatially Variant Mode", with 15 PSF samples. I missed that while reading the log.

Thanks & best regards, Dietmar.
Stefan B
Posts: 194
Joined: Sat Oct 31, 2020 8:59 pm

Re: M 81 and M 82

Post by Stefan B »

Hi Dietmar, yes, I used the spatially variant mode, great tool! Happy holidays to you, too!
Regards
Stefan
decay
Posts: 46
Joined: Sat Apr 10, 2021 12:28 pm

Re: M 81 and M 82

Post by decay »

Hi Stefan,

last days, I had a go (or maybe some more :) ) with your ST session log and your remarks. I would like to give feedback step by step and – of course – there a some new questions :lol:
The data had a very homogeneous gradient which was very easy to get rid of in Wipe. That's nice ... recently I had problems with several gradients in my images which had different directions. A real pain.
Yes, some months ago I built a flat field box DIY and this was a really big improvement. I never managed taking flats during twilight. Nevertheless, if there are gradients due to light pollution they result together with earth rotation in really funny shapes in the stacked image.
I binned to 50%. In my eyes the image and especially the stars got blocky with 40% or below.
Great, that matches the outcome of our discussion earlier.
In Wipe I used a high 'Dark Anomaly FIlter' value of 11 pixels. This may seem high, but I am used to this from my own data. This optimizes the wipe of the color data. It's striking to have a look at the color data with a value of 1 and afterwards around 10. Most background pixels are red with a value of 1 pixel and you get a homogeneous background with a value of 10. Makes the work in the Color module much easier.
OK, this is really interesting. I never paid any attention to the colour outcome of the wipe module. But this may explain struggles I sometimes had finding a reasonable colour balance later on :( So, what does the predominant existence of red pixels mean? I guess, that’s stretched noise with a predominant red fraction? This would mean, that the background has a red colour cast and indeed, it has (which is not that surprisingly). I did the processing up to the color module twice, first with a 'Dark Anomaly Filter' value of 1 and then with a value of 11. In color module I applied the same settings in both cases and the difference is obvious:
M81-DarkAnomalyFilter-Diff.jpg
M81-DarkAnomalyFilter-Diff.jpg (128.44 KiB) Viewed 212 times
But I do not understand, in which way is the colour calibration of the background influenced by the 'Dark Anomaly Filter'? Is there some kind of “coloured blotches”, which extend up to 11 pixels? But I cannot see something like this. Your advice setting this value at such a high level is very helpful and I certainly will do that in future – but I would feel even better if I would understand this behaviour. :?: Do you have any explanation?

Thanks & best regards, Dietmar.
User avatar
admin
Site Admin
Posts: 3108
Joined: Thu Dec 02, 2010 10:51 pm
Location: Melbourne
Contact:

Re: M 81 and M 82

Post by admin »

decay wrote: Thu Apr 28, 2022 7:59 pm But I do not understand, in which way is the colour calibration of the background influenced by the 'Dark Anomaly Filter'? Is there some kind of “coloured blotches”, which extend up to 11 pixels? But I cannot see something like this. Your advice setting this value at such a high level is very helpful and I certainly will do that in future – but I would feel even better if I would understand this behaviour. :?: Do you have any explanation?
Dark anomalies prevent Wipe from detecting the true background levels for each "channel" individually (red, green, blue, luminance, and NB accent channels). If not mitigated (e.g. either masked out. or filtered out by the DAF) Wipe will think any dark anomaly is the background. Now it will leave the true background sitting on top. Datasets should indeed rarely require such high Dark Anomaly Filter settings, however 1 pixel may not be enough, particularly if the dataset is noisy.

See Wipe docs for more information.
Ivo Jager
StarTools creator and astronomy enthusiast
decay
Posts: 46
Joined: Sat Apr 10, 2021 12:28 pm

Re: M 81 and M 82

Post by decay »

Thank you, Ivo.
Datasets should indeed rarely require such high Dark Anomaly Filter settings, however 1 pixel may not be enough, particularly if the dataset is noisy.
Reading this, I revisited the dataset and indeed this "effect" stops at about 4 to 5 pixels. That’s good news, I should haven taken a closer look before. But luminance wipe seems to be very good even if DAF is only set to 1; this "effect" only affects colour. So my assumption is, that these "dark anomalies" are coloured in some way (not just "dark"). Is that right? I cannot see such anomalies in the dataset ...
Stefan B
Posts: 194
Joined: Sat Oct 31, 2020 8:59 pm

Re: M 81 and M 82

Post by Stefan B »

I didn't evaluate exactly how much DAF is needed, but a value of around 10 fixes the issues in my images so I went with that ;)

I guess the anomalies are color mottles of some sort but you will probably have a hard time finding them. At least they are pretty dark and small and therefore the color mottling will be hard to notice. I think.

With most of other people's images DAFs of around 3 or so resulted in good color wipes in my eperience. But I have the impression with images made with Canon DSLRs you have to push it a bit. There were some data sets on the internet where I saw something similar. Maybe a sensor issue. But strange that this doesn't get away with dithering. If your luminance noise looks like random single pixel noise this should apply also to color. But don't know...

Regards
Stefan
decay
Posts: 46
Joined: Sat Apr 10, 2021 12:28 pm

Re: M 81 and M 82

Post by decay »

Hi Stefan, that‘ s even more interesting than I thought! Thanks a lot for sharing your findings and experience.

I thought about your note that the “problem” persists even with dithering. Really interesting. For now, I have no idea for an explanation of this effect. Perhaps Ivo or someone else has … ?

Nevertheless, lessons I learned:
1. Always check colour result of Wipe Module.
2. Always use Dark Anomaly Filter to exclude dark anomalies whether they are visible or not and whether they are coloured or not :) And a setting of about 10 may take some time but seems to be enough for all of us :lol:

I would like to go on with some feedback concerning the remaining workflow you shared, but more for completeness ...

And yes, best at last …
I guess you stacked in DSS? At least it looks like my stacks in DSS. I recommend to just try one of the other free stackers like ASTAP or Siril. I know that this means getting to know another software and stuff but in my experience it's really worth it. In my hands other stackers always resulted in better stacks with less grainy noise and a better SNR in general. I think there's more in the data than your stack shows
And of course, you are right, I used DSS. Last days, I teased my notebook and myself and finally I managed to stack with both ASTAP and Siril. And I have a lot of notes … and two new stacks … and some more questions :mrgreen:

Please, let me know if you would like to go on with this thread, as I there is little to nothing that I could give back to you.

Best regards, Dietmar.
Stefan B
Posts: 194
Joined: Sat Oct 31, 2020 8:59 pm

Re: M 81 and M 82

Post by Stefan B »

Hi Dietmar,

No problem, let's go ;) Just don't expect that I have all the answers :lol:

Regards
Stefan
decay
Posts: 46
Joined: Sat Apr 10, 2021 12:28 pm

Re: M 81 and M 82

Post by decay »

Hi Stefan,

sorry … better late than never:

AutoDev (redo global stretch)
As said, I think the most important point is to gain experience setting the right ROI. I often chose square shaped ROIs, but these rectangular shape like you did, may be better suited. The most difficult point in my eyes is to find the right brightness and noise level for the background, as it has severe impact on the resulting image. Changing display brightness levels and ambience lights make it even more tricky. I will take a reference screenshot from the outcome of AutoDev and will use it later on for comparison when doing AutoDev on other images. Just a try … ;)

Contrast, HDR, Wavelet Sharpen
OK, nothing special about these steps. (I am not sure, if HDR had a stronger impact in the past. Maybe my old data sets had other characteristics or previous versions of ST behaved not likewise.)

Spatially Variant PSF Deconvolution
Don't know what to say. Wow :bow-yellow: . As you said, a great tool. It must be some kind of magic.

Color
I noticed, that you set Cap Green to 100 % and Highlight Repair to 10 px, but with little (?) impact.

Shrink
I never used this module before, but I surely will do that in future. Maybe cosmetic, but it really helps making the image looking more convenient.

Super Structure
Another magic tool. As I took my first noisy images, this module made my day. I read that it is possible to use it twice, but I never did it. In my eyes, this is the most striking difference between your and my first rendition. It is important to use it with care, you did it by reducing the Strength parameter for 'DimSmall' and adjusting the Saturation parameter for the 'Saturate' run. Thanks for pointing out, this will help me to improve my next renditions.

De-Noise
Thanks to tracking, this module ‘knows’ where and how the noise has to be reduced. Fascinating. I have the impression, that it works best, when the Superstructure module has been used previously in the workflow.

FilmDev
Have you noticed, that the image gets slightly darker, when the module it started, without setting Digitial Development or other parameters? I wonder why? I think, this effect, together with setting Skyglow to 2 % afterwards, seems to mitigate the mottling of the background a little. That’s quite nice.

All in all it’s the sum of multiple points, which makes your rendition look that fine. Thanks again, this will surely help to improve my renditions.

With my next post, I would like to share my ASTAP stack and maybe you could take a look … but as said, only if you like ...

Best regards, Dietmar.
Mike in Rancho
Posts: 467
Joined: Sun Jun 20, 2021 10:05 pm
Location: Alta Loma, CA

Re: M 81 and M 82

Post by Mike in Rancho »

Hi Dietmar,

I'll let Stefan handle some of the specifics of his workflow, but --

The new HDR should be both more capable and produce better all-around images than old-HDR, but may take a bit of practice. Some of the stuff buried in the previous preset algorithm choices are now within our control. Probably the most common issue is how to get something like "reveal core" in 1.8 (an important thing in galaxy season!), which can be done with a low context and increased shadow on the left-side sliders.

In SS, occasionally Strength can have a counterintuitive result. I more often adjust the gamma to moderate the effects of DimSmall and Isolate. Usually Isolate, as .50 is pretty heavy. DimSmall used to have a default setting of .50 but is now .75 and I often leave it alone.

For a post-tracking FilmDev, note that there are likely default settings in the Dark Anomaly that probably cause that minor initial effect, even without starting to change the other controls. I will sometimes drop the gamma also, which can help suppress a dirty sky and also increases the contrasty appearance. That can be balanced off with Skyglow, if desired. Every image may need a little something different to show what it is you want to show.

Intended display matters too. I've realized I usually have to make those final adjustments to make a (darker) version for my phone, even when the original may look great on a calibrated computer monitor. Probably because the phone has a jacked-up gamma that I am counteracting?
Post Reply