Star Appearance Issues When Using AutoDev

[juno16]

Hi Ivo,

I have inquired about similar issues before and continue to struggle with my star appearance when using Autodev vs FilmDev.

I recently shot this image of M33. I somewhat like the blueish overall appearance of the galaxy, but if you have any recommendations of what color balance would look better, please let me know.

M33 All Data from DSS 1st pass through ST + sm.jpg (498.54 KiB) Viewed 6704 times

My issue is with the stars. As you can see, if I zoom in on any bright star, I see halo’s and sometimes multiple colored halo’s. I haven’t used FilmDev on this image yet to see how the stars look, but usually they are somewhat more tamed than when I use AutoDev.

Capture.JPG (27.28 KiB) Viewed 6704 times

Please let me know if you feel that the appearance is acceptable to you, or please recommend a Startools process that might help. I have tried the Repair module without success, but possibly I am not using the right settings.

Could be that I am trying to chase down a non-existent or minor issue. Or, possibly it might be equipment/exposure related and not remediable.

Startools (1.7.438a) has done a really nice job with my image using mostly defaults. I need to run through another process to tweak things.

As always, thanks for your fine software and your help!

Jim

EDIT: Exposure info
EXplore Scientific ED102
SV 0.8 FF/FR
ASI 533 MC Pro
IDAS LPS D-1 filter
About 9.5 total hours of integration using 90 second exposures at unity gain (100).

Link to tiff file save after process in ST:
https://drive.google.com/file/d/1BcL76O ... sp=sharing

Link to stacked fit file from DSS:
https://drive.google.com/file/d/1PCyVUt ... sp=sharing

[admin]

Hi Jim,

Please bear with me as I finish up some things here. I will respond shortly.

Thank you!

[juno16]

Absolutely Ivo, no rush!

Thanks,
Jim

[admin]

Hi Jim,

Thank you for bearing with me.
I finally had a good look at the dataset.

The "issue" is that AutoDev is just really good at brining out any detail - even if it is artificial or the result of an anomaly.

When looking at your the stellar profiles in your dataset, something - quite literally - doesn't add up.

For example, looking at the dataset in the luminance domain, this is what your stars look like when optimizing a stretch for them;

Selection_279.jpg (52.06 KiB) Viewed 6355 times

It is the same across all channels, so it is not a chromatic aberration issue.

It's almost as if the dataset consists of different exposure times, or as if the dataset was shot under vastly different atmospheric conditions. Similar issues can also occur when not letting mirrors cool to ambient temperature before commencing imaging, but there are no mirrors in your optical train, correct?

As a result, not only do the stellar profiles look rather odd in the luminance domain, but when it comes to combining the color channels, chrominance is also all over the place, causing significant color artifacts too;

Selection_280.jpg (151.73 KiB) Viewed 6355 times

I can understand not being happy with the stars as they are. However, what AutoDev and the Color module(!) are showing you, is something that should be corrected in your data and not something that should be "hidden" during post-processing (though you certainly can do that if that is what you really want in ST). E.g. it is not "standard fare" or something you should (want to) put up with by just using a custom stretch and desaturating your highlights.

Just for a quick test, would you be able to stack your dataset with, for example, ASTAP to see if it is maybe a stacking issue?

I believe some people with the same sort of issue got better results when they ensured hot pixel removal is not selected on the Cosmetics tab in DSS...

Do let me know your thoughts/findings!

[juno16]

Hi Ivo,

Thanks a bunch for looking at this issue. Sound kind of messy!

If you don't mind a slight delay, I will get to this asap. I will try stacking using astap and reprocessing in Startools.

Don't have much time today, so please bear with me! I need to get this resolved obviously, but I have been having issues like this for awhile.

I imaged over two nights using the same exposure settings. Sky quality seemed worse the second night, but not much different from normal.

I did reprocess using the FilmDev module and got much better results on the stars. Still some guiding issues, but more uniform star color.

https://drive.google.com/file/d/19LiMEa ... sp=sharing

I will let you know and thanks a bunch for your fabulous support!

[juno16]

Hi Ivo,

I tried ASTAP using the settings outlined, but I get a horribly noisy image. I didn't finish processing in Startools. I'm sure that something is off.
The file is about the same size (fit) as the DSS staked fit file (~106mb), but not sure why there is so much noise.
Here is a snip of my Stacking tab in ASTAP.

Capture.JPG (169.17 KiB) Viewed 6292 times

I know that you are very busy, but here is a link to the fit file stacked in ASTAP. Maybe if you get a chance to take a look, you will see what I mean and set me straight on the settings.

Thanks for your help and if you don't get time to work with this I understand.

Thanks again!

https://drive.google.com/file/d/111T8wW ... sp=sharing

[admin]

I imaged over two nights using the same exposure settings. Sky quality seemed worse the second night, but not much different from normal.

Ok. So if you compare one frame (say the last frame) from the 2nd night, and another frame (say the first frame) from the 1st night, the stars look roughly the same?

I did reprocess using the FilmDev module and got much better results on the stars. Still some guiding issues, but more uniform star color.

Indeed, this is pretty much "hiding" the issue (the Digital Development response curve just makes the stars bloat) but it does not fix the issue. If it can be fixed then bloating the stars (or otherwise trying to repair the star cores in post) is not necessary and a more optimal curve can be used.
Getting to a point where AutoDev (and Color module's Color Constancy mode - even if you don't prefer to use it for your final coloring) can be used without any issues is greatly preferable.
The strange way the cores look will also cause issues for, for example, deconvolution; it will impact the way the point spread function looks (it will make it differ depending on brightness, whereas decon needs/assumes a uniform PSF to be able to work correctly).

Can you confirm for me that "hot pixel removal" was not selected on the Cosmetics tab in DSS?

Thank you also for stacking in ASTAP. I was hopeful this would have yielded some sort of clue, but I'm thinking something has gone wrong; it almost looks like the resulting stack is a single frame (perhaps the rejection algorithm rejected everything?).
A latent pattern (ST's AP-friendly scaling algorithm FTW!)
is also visible at various zoom levels, making me think something maybe wasn't configured right in ASTAP with regards to the used Bayer Matrix? Don't spend too much time on getting ASTAP to work if you think we can find some clues in another way though...

That said, in the ASTAP stack too, I can see very strange stellar profile...

Selection_282.jpg (33.1 KiB) Viewed 6285 times

Perhaps it is just the result of the tracking error (the stars not very round), combined with the rotation of the field? Was the orientation in the 2nd stack markedly different to the orientation in the 1st? Is the problem less noticeable when just stacking one stack without adding the other night's frames?

[juno16]

Hi Ivo,

Thank you for spending so much time on this issue.

Definitely a data issue, but I have to get to work troubleshooting. I will look at the possibilities you mentioned.

I did confirm that hot pixel removal is not checked in the Cosmetics tab (unfortunately!)

I do have star trailing issues especially the closer to the meridian that my target is. My mount (AVX) is a big part of that and I will have to live with it for now. Total rms error in PHD2 was under 1.2 arcsec which isn’t horrible (for me), But adding to the tracking issues is that I just changed from a dslr to the ASI 533 that has a much smaller fov and amplifies theses issues.

As poor as a choice that it is, I will probably have to go the filmdev route for now. Even with all of the problems with my data, I really love this hobby and have managed to exceed my original goals in producing successful images (of course, relatively speaking).

The star issues are disappointing, but with the 9.5 hours of integration on this target (my longest so far), I was happy with the details in the image (especially the Ha areas).

As you know, I have been dabbling with Photoshop too and the Startools process of this data (even with the issues) completely blew away what I could do in Photoshop (of course, skills limited with both applications!).

I will dig around and post any findings.

Thanks again Ivo for your time, support, and excellent product!

Jim

[admin]

I will dig around and post any findings.

Please do - I'm always interested in hearing solutions to problems that other may well experience too.

Worst case, you can still totally process these issues out in StarTools, after using AutoDev - there is no need to use FilmDev if you are not really looking for a photographic film response.

First of all, in the Color module, bump up the Highlight Repair. It was made precisely for aberrant color information in the highlights like we're seeing in this dataset. Cap Green will further remove any spurious green.

From there, you can use the Shrink (and maybe Repair?) module. The Shrink module has a feature similar to the Highlight Repair if you need to apply it do other/more stars. Just put the stars(+halos) in a mask, go into the Shrink module, set Iterations to 0 (so nothing gets shrunken), De-ringing off too, and bump up the Color Taming parameter. This makes stars assume the color of their surroundings like little chameleons.

Of course, you can also decide to shrink the stars some, which is probably not a bad idea.

So, after a Bin/Crop/Wipe and this AutoDev;

StarTools_283.jpg (174.11 KiB) Viewed 6269 times

You should get something like this in the Color module with just those two parameters set (Cap Green and Highlight Repair) and default color balance;

StarTools_284.jpg (179.29 KiB) Viewed 6269 times

The smaller stars now have minimal color aberration left, which you can fix in the Shrink module like so;

StarTools_286.jpg (144.16 KiB) Viewed 6269 times

After that you can do a second pass to shrink the stars (or do it all in one hit).

You should end up with something rather decent;

M33 All Data from DSS.jpg (281.85 KiB) Viewed 6269 times

And that's just a really quick Shrink w/o deconvolution, sharpening, etc. IMHO, it definitely beats the bloated FilmDev stars (tweak to taste obviously).

Hope this helps!

EDIT: I know I sound like some sort of evangelist at this point, singing the praises of AutoDev, but particularly M33 is a good example of its strengths; you should be able to see the extremely faint, intricate darkest dust lanes in its outer edges and that's with just a single global stretch. All natural, real detail, no local contrast enhancement.

[juno16]

Wow! Outstanding! This is excellent Ivo!

I will get to spend some time working with the process next week and I am really excited about getting into it.

You really did improve the star profiles and the coloration. I really do like Autodev, and I am looking forward to applying your settings and recommendations.

I personally am leaning toward some of the star coloration/halo issues are related to tracking/field rotation. Unfortunately, I did push things a bit with the new camera fov. I know that it is not ideal, but it is really good to hear that most of the errors are repairable to a large degree.

Your process is wonderful. Thanks for showing what can be done in Startools to bring out the best image from this dataset!

How do you feel about possibly using the lens module (curvature linked) to round out the stars a bit?

I cannot thank you enough Ivo for the outstanding help and the time that you spent to working with me. I will post any findings/learnings.

A big thank you sir!