Star artifacts - optical or processing

[ajh499]

Hello all

I hope someone can help me with a problem that I seem to have always had with my images - I seem to (almost) always have these odd artifacts around stars.

The attached images shows what I mean.
There's two issues - the first is the bright saturated blue highlight at the top of the brighter stars in this 300% crop. Is that a processing artefact or an optical problem? I see it in many images, it can be a different colour depending on the star, and can be in different directions but always has that "pasted on" look to it.

The second issue is the grey halo around the stars, where is that coming from? It seems to be part of the stellar profile of the stars, but why so monochrome? Again, it looks "pasted on".

This image is an LRGB composite using Ha for the L, but the same effects are present in eveything I try to colour whether it's LRGB or just RGB.
There was only basic processing done for this example, just autodev, crop, wipe and colour.

Thanks for any help, it will be most appreciated

Alex

[happy-kat]

If you stretch a single light frame do you see something similar?

[ajh499]

No, but then this is from a mono camera, each filter on it's own looks OK

The problem seems to appear in the colour module, up until then with the image only showing luminance, it looks fine.

I'm fairly sure that the coloured highlights are an optical problem, the red frames are sharper than the green and blue, but why is that resulting in odd spikes rather than a halo, I'm not sure.

And where does the gray halo come from?

[ajh499]

Stacking seems to be playing a role here.

My original image, and the way I usually stack images is done in Siril using SirilIC to drive it.

As an experiment I tried using ASTAP for stacking and got the following using the default settings - A colour image composited by ASTAP and only displayed using autodev in ST

astap-colour.jpg (542.12 KiB) Viewed 6424 times

Setting ASTAP to not create a final image and output the individual filter stacks, then align them in a second pass (I can't find a way to do that in one go) results in the following when then composed in ST.

astap-compInST.jpg (420.08 KiB) Viewed 6424 times

That's still worse than my original attempt, but at least it shows that the stacking software is contributing in some way. Maybe not aligning the filters properly?

I'm still not sure of the source of the grey though

[happy-kat]

just reread the first post, there's is no second autodev which is needed.
autodevv / bin / crop / wipe / autodev / process with other modules / colour / denoise

[ajh499]

Yes there was a second autodev after wipe on both versions composed in ST.

The version that was composed in ASTAP wasn't wiped, as even though I have turned all the cosmetic options off in ASTAP it still seems to be messing with the background of the image, and attempting to wipe it just made it worse. That's not the point of that image though, I just wanted to see what it did to the star highlights.

Alex

[ajh499]

I'm still stumpted with this!

This example is not a great data set, it wasn't possible to get much time on it, but I see artifacts like this in practically all of my images and I'd like to get to the bottom of what's going on.

It may just be that the focus is different between channels (the scope is only a doublet - SW ED 80), and I've had problems refocussing on filter changes, but I do see a similar problem in my first (and so far, only) image from an Askar FMA135 which should be better.

Stacking might be a part of this, but I've tried three different programs for stacking and while results change slightly, they don't make a big difference.

There was a post on here somewhere, where someone had a similar, but not as severe problem, but there was never really an explanation of the issue, just a bit of a workaround to reduce the worst of the effects, but I can't find the post again.

Composing in L + Synthetic L from RGB mode seems to mask the problem a bit, but in L, RGB it's terrible. Using Legacy mode in the Colour module doesn't work either, it just totally desaturates the stars.


Please could someone more knowledgeable than me take a look at the stacks, I'm confused.

https://www.dropbox.com/s/zs2kzik8edh4rpi/M16.zip?dl=0


Thanks

Alex

[Mike in Rancho]

Hi Alex,

I took a look at the M31 set. You are right, the RGB is maybe a little noisy. Ha seemed pretty cool though!

At first I just composed the RGB in the normal fashion and processed through. Presumed same integration time per color, so left defaults. Indeed, when I got to color the stars seemed quite out of kilter as far as color was concerned. Particularly due to your experiment with ASTAP combining the channels, I wondered if the registration was failing slightly, or perhaps you utilized an RGB channel align (similar to the one in DSS that is handy for atmospheric dispersion) that centered things better.

Still, it could be reasonably addressed in ST. I used all the standard workflow modules. Strong Wipe, reasonable stretch for the limits of the RGB data using IFD and ROI, compression of the shadows in Contrast, then pretty basic HDR, Sharp 200%, and SVD with increased deringing parameters, 4 sample stars. I then used Shrink, no iterations, but 2 pixels deringing and default halo/color taming. Star sampling for color still left quite a bit of scattered green (more on that later), but instead of balancing it out and possibly turning M31 too purple, I just green capped. An SS followed by denoise with lowered scale finished it off, and it wasn't too bad I think.

To investigate the star problem more fully though, I opened all the FITs as layers in Gimp and just flipped back and forth with visibility. (I had previously done something similar with the RGB channels after saving from ST right after my first test of the color module). I didn't even bother stretching, just zoomed in on various stars in full linear. Your blue and especially green channels stars all have what I would almost call coma. Though I don't see any diffraction spikes, so perhaps something to do with a reducer or field flattener, perhaps in conjunction with certain of your filters?

While flipping the layers it did seem that the central spots of the stars mostly stayed in place, so registration and alignment may not have been an issue and perhaps the stackers did pretty well on that.

You may want to try out some star testing with your various filters and see if you can discern anything through experimentation as to why there is a difference. And if that is something that is happening all the time (you did mention this issue exists in several datasets).

I then tried using your Ha set via NBA for a second go-through. It was fair, but I had to go back and pull some saturation at the end due to the way it blew out the highlights. Multiple ways of combining NB and RGB, just depends what your plan is and how you are intending to utilize it. If I did that again I would probably use the color module on the RGB in anticipation of the NBA contribution to come. As it was though, it just added some nice extra nebulosity away from the target, and increased the detail on target. I didn't try Ha as L, as that can take a wrecking ball to colors and this was sort of a color experiment anyway.

Anyway, that's all I can think of for now, is that something maybe isn't matching up equally during filtered acquisition.

[ajh499]

Mike

Thank you for all of your input, it's very helpful having someone else looking at the problems.

I'll try your suggested settings for the workflow and see what it looks like.

I have usually been trying Ha as L but just to have a look at the colour I'll do what you did and just use the RGB. I take it you just used the RGB alone, without the Ha at all?

This is a particularly poor data set (by the way it's M16 not M31 ), I took this in the summer and I only get a couple of hours when I can make it out between neighbouring houses. It does seem to be an especially bad example of this problem that I do see with other images too.

I'm not sure what the conditions were while capturing this, but it's possible that there was some cloud around. There were a lot of days around that time that were forecast to be clear, and looked clear to the naked eye, but showed up fuzzy in images. Maybe that's the issue.

Next time I get a clear night, I'll check the focus with each filter and see if there's any difference or detectable problems

Thank you, again for taking a look


Alex

[Mike in Rancho]

Ha, yes M16. I was thinking Eagle whilst typing Andromeda, I guess.

Also forgot to mention the most critical part of the workflow with regard to the star color flaws -- highlight repair in Color! I think it was 5 pixels. Silly me. And then shrink came after that of course.