When come the lines?

Linwood

Well-known member
I have been experimenting with globular clusters as a good way to flush out issues with my optics, etc. but I have an ongoing battle with artifacts I do not understand.

They are tiny, and can be swamped by just letting the stars bloat and hide them, but I am hoping someone might share some info on where they come from, and can they be killed at the source rather than hidden.

This is a highly magnified (4;1) image of M5, preview stretched to to show more individual stars, but otherwise not processed. It is a stack of 3 nights data at a small scale (0.277"/pix). So yes, pixel peeping in the extreme. This one is not drizzled, it looks a bit worse in the drizzle version.

lines.jpg


Notice the line-like artifacts between the stars. They are artifacts.

These become grossly apparent if you run the stars through BlurXTerminator, WITHOUT non-stellar sharpening, just reduce the star size. It may LOOK like the lines were emphasized, but really they were already there, the stars just pull away from them. Blink between these and there's very little change in the lines if any.

bx.jpg


I went back to the original subs. Here is one, preview stretched only to about the same field of view as the first.

oneSub.jpg


The star halos overlap slightly, but some aspect of integration appears to be emphasizing the line, adding a gap on either side. In another thread it was pointed out that drizzle with non-square or non-drop-shrink=1 can exacerbate this, but these are not drizzled. The image integration settings are:

integration.jpg


So my question.... is this just way too far down in the pixel peeping range and I need to stop looking? Or is there something I can do to adjust how integration creates (or maybe more precisely emphasizes) this artifact?

Or... is it really not an artifact? Are the airy disks overlapping and actually creating a line there and it is just not as apparent in the lower SNR of the single sub?

Linwood
 
Linwood,

You need to show your high rejection map- with auto STF.

The large scale rejection (or even rejection method) can cause some issues if you are using highly variable (seeing, focus..etc) data.

From my experience this is usually a rejection issue.

-adam
 
I have been experimenting with globular clusters as a good way to flush out issues with my optics, etc. but I have an ongoing battle with artifacts I do not understand.

They are tiny, and can be swamped by just letting the stars bloat and hide them, but I am hoping someone might share some info on where they come from, and can they be killed at the source rather than hidden.

This is a highly magnified (4;1) image of M5, preview stretched to to show more individual stars, but otherwise not processed. It is a stack of 3 nights data at a small scale (0.277"/pix). So yes, pixel peeping in the extreme. This one is not drizzled, it looks a bit worse in the drizzle version.

View attachment 18702

Notice the line-like artifacts between the stars. They are artifacts.

These become grossly apparent if you run the stars through BlurXTerminator, WITHOUT non-stellar sharpening, just reduce the star size. It may LOOK like the lines were emphasized, but really they were already there, the stars just pull away from them. Blink between these and there's very little change in the lines if any.

View attachment 18706

I went back to the original subs. Here is one, preview stretched only to about the same field of view as the first.

View attachment 18704

The star halos overlap slightly, but some aspect of integration appears to be emphasizing the line, adding a gap on either side. In another thread it was pointed out that drizzle with non-square or non-drop-shrink=1 can exacerbate this, but these are not drizzled. The image integration settings are:

View attachment 18705

So my question.... is this just way too far down in the pixel peeping range and I need to stop looking? Or is there something I can do to adjust how integration creates (or maybe more precisely emphasizes) this artifact?

Or... is it really not an artifact? Are the airy disks overlapping and actually creating a line there and it is just not as apparent in the lower SNR of the single sub?

Linwood
What was your subexposure time? How was your seeing? On an individual sub, if you examine a profile line through a star, do you see just the central peak or is some of the Airy ring structure apparent?
 
What was your subexposure time? How was your seeing? On an individual sub, if you examine a profile line through a star, do you see just the central peak or is some of the Airy ring structure apparent?
The subs are R, G, B (all have this) and 240s at gain 100 on a ASI6200mm. I see it both on a SVX152T and a C11 (both at their native resolutions of about 1200 and 2800).

The 3rd posting above is an individual sub picked at random. Generally my seeing yields between a 1" and 2" FWHM measured by subframe selector, so pretty good. Are you referring to the 2D plot utility? If I draw it diagonally through the one circled on the top right in the first image, I get this on a random sub:

2D.jpg


If I go the other way through just one star I get only a central peak, no sign of secondary rings.

2d2.jpg


Interestingly if I do the same in the integration it doesn't show as existing, there's no hump in the middle at all. I tried numerous lines through what clearly appears to be a hump, so it must be my imagination (and sorry, on the above I had not noticed the STF check box before).

2d3.jpg



Anyway... Adam, here is the rejection high for approximately the same area, with the preview02 showing the area of the two stars in the top right circle, same as for the 2D plots, auto-stf. I worked hard to align it with the top (by hand) and it's precisely drawn around the area of the stars with a preview copy, but to be precise below is the overall area again with the two preview boxes. Nothing jumps out at me?

rejection.jpg
previewAreas.jpg


I sure seem to see a line in that last one, but unclear why it doesn't show on the 2D plot. I rescaled the 2D plot to enhance just the middle and it just doesn't show a dip. So optical illusion, made real by BX retracting the star halos?
 
Is there nebulosity in the background of the last image you posted?
It's the core of M5 so... sort of? But not real nebulousity, but lots and lots of bright stars and even more dim ones. (Not certain which is the "last" as we were posting almost the same time).
 
That leads me to wonder if BX was trained with such bumpy background...maybe its training was on more nearly isolated stars???
 
That leads me to wonder if BX was trained with such bumpy background...maybe its training was on more nearly isolated stars???
Maybe, but bear in mind I showed the BX only as an example. I see these without any processing. I'm not trying to make this about "look what BX did" but rather "where do these come from without BX".
 
So optical illusion, made real by BX retracting the star halos?
I've been keeping my mouth shut, but I've been suspecting this. I think your visual processing is detecting the radial fall-off of brightness and "over-shooting" to a percieved darkening around the star that is not really there in the image. There are loads of optical illusions based on similar effects, based on lateral inhibition in the retina.
 
The subs are R, G, B (all have this) and 240s at gain 100 on a ASI6200mm. I see it both on a SVX152T and a C11 (both at their native resolutions of about 1200 and 2800).

The 3rd posting above is an individual sub picked at random. Generally my seeing yields between a 1" and 2" FWHM measured by subframe selector, so pretty good. Are you referring to the 2D plot utility? If I draw it diagonally through the one circled on the top right in the first image, I get this on a random sub:

View attachment 18708

If I go the other way through just one star I get only a central peak, no sign of secondary rings.

View attachment 18709

Interestingly if I do the same in the integration it doesn't show as existing, there's no hump in the middle at all. I tried numerous lines through what clearly appears to be a hump, so it must be my imagination (and sorry, on the above I had not noticed the STF check box before).

View attachment 18710


Anyway... Adam, here is the rejection high for approximately the same area, with the preview02 showing the area of the two stars in the top right circle, same as for the 2D plots, auto-stf. I worked hard to align it with the top (by hand) and it's precisely drawn around the area of the stars with a preview copy, but to be precise below is the overall area again with the two preview boxes. Nothing jumps out at me?

View attachment 18712View attachment 18713

I sure seem to see a line in that last one, but unclear why it doesn't show on the 2D plot. I rescaled the 2D plot to enhance just the middle and it just doesn't show a dip. So optical illusion, made real by BX retracting the star halos?
I'd ask Russ Croman about this. It could be an issue of the training data for the ML model. This is obviously going to be a challenging image for deconvolution. Have you compared the results with using an automatic PSF generation as opposed to entering a value manually?
 
I've been keeping my mouth shut, but I've been suspecting this. I think your visual processing is detecting the radial fall-off of brightness and "over-shooting" to a percieved darkening around the star that is not really there in the image. There are loads of optical illusions based on similar effects, based on lateral inhibition in the retina.

Well, it wouldn't be the first time my imagination got me in trouble. Just tell me you see it also. o_O
 
I'd ask Russ Croman about this. It could be an issue of the training data for the ML model. This is obviously going to be a challenging image for deconvolution. Have you compared the results with using an automatic PSF generation as opposed to entering a value manually?
Maybe, but I'm really not trying to make this thread about BX. Let's pretend for the moment I never mentioned BX.

Do you see the wall?

Maybe it's an alien megastructure that can't be measured, but can be seen. Some sort of weird quantum effect.
 
Maybe, but I'm really not trying to make this thread about BX. Let's pretend for the moment I never mentioned BX.

Do you see the wall?

Maybe it's an alien megastructure that can't be measured, but can be seen. Some sort of weird quantum effect.
It's perfectly measurable. Use any utility that gives the color value under the cursor and the existence of brighter pixels is seen to be real and quantifiable.
 
Maybe, but I'm really not trying to make this thread about BX. Let's pretend for the moment I never mentioned BX.

Do you see the wall?

Maybe it's an alien megastructure that can't be measured, but can be seen. Some sort of weird quantum effect.

There's definitely something there, and it is measurable. You can see the bump in your 2D plot. The middle grouping of stars you circled in red would probably show what you're expecting (a peak in-between the two star profiles).

Do you see the line in any of your registered images? Do they only begin to appear in your integrated image? Since BXT helps show the lines better, you can apply it to calibrated images to help figure out where the lines first begin to appear.
Star Profile.jpg
 
Did you try to integrate the images without rejection?
These are some easy tests...
-adam
 
Now I'm totally baffled. I pulled up a sub and saw it clearly, but it was a registered sub, and I had been looking at raw, so I went back through the evolution. Left to right with manual preview adjustment to look similar is raw, calibrated, cosmetically corrected, and registered.

I must have something really bogus in my cosmetic correction, I use it blindly every wbpp run, and haven't looked at it in a year. It has auto-detect (only) set and both at 3 sigma, for hot and cold.

Here's where it really gets weird -- if I dial the hot/cold to zero, which i would think means "do nothing" it still causes it.

I tried a master dark but got an access viloation, so restarting everything now, but...

Cosmetic correct?!?

(Adam, just saw your note, but I think next run I'll take off cosmetic correction, but I want to experiment manually a bit more first)
cc.jpg
 
Huh... I wonder if you are not getting *enough* rejection. Is this thing in a single frame? Is this just a cosmic ray?
-adam
 
Back
Top