Brush stoke noise - what's happening?

Started May 23, 2017 | Discussions
Lyle Aldridge Senior Member • Posts: 1,151
Brush stoke noise - what's happening?
1

This forum has definitely piqued my interest in using camera lenses instead of my telescope. After admiring the Rho Ophi shots posted by others, I gave it a try a few nights ago, using my unmodified Canon T2i and 300mm/F4L lens, the only true high-quality lens I own. I shot 28 unguided subs of 90 seconds each, using my LXD-55 mount, plus 24 darks. No flats or bias.

My immediate problem is the diagonal brush stroke-like pattern that appears in the 100% crop from the stacked image below. This particular stack is from Sequator, but I get the same pattern when stacking with DSS. The DSS stacks - median or average - just show less color. This first has been stretched just a bit in PS levels to make the pattern a bit more visible, but not processed beyond that. I seem to get the same pattern whether I use the darks or not. In fact, the pattern is just slightly less evident when I stack without them, again in either stacking program.

100% crop from Stack of 28x90sec, 1600 ISO

I obviously see noise in a typical 1600 ISO sub, but I don't see the same diagonal pattern.

100% crop from single converted RAW, 90 sec, 1600 ISO

This crop also has been stretched as the stack was. Anyone know what's going on here?

Canon EOS 550D (EOS Rebel T2i / EOS Kiss X4)
If you believe there are incorrect tags, please send us this post using our feedback form.
rnclark Veteran Member • Posts: 3,957
Re: Brush stoke noise - what's happening?
1

Lyle Aldridge wrote:

This forum has definitely piqued my interest in using camera lenses instead of my telescope. After admiring the Rho Ophi shots posted by others, I gave it a try a few nights ago, using my unmodified Canon T2i and 300mm/F4L lens, the only true high-quality lens I own. I shot 28 unguided subs of 90 seconds each, using my LXD-55 mount, plus 24 darks. No flats or bias.

My immediate problem is the diagonal brush stroke-like pattern that appears in the 100% crop from the stacked image below. This particular stack is from Sequator, but I get the same pattern when stacking with DSS. The DSS stacks - median or average - just show less color. This first has been stretched just a bit in PS levels to make the pattern a bit more visible, but not processed beyond that. I seem to get the same pattern whether I use the darks or not. In fact, the pattern is just slightly less evident when I stack without them, again in either stacking program.

100% crop from Stack of 28x90sec, 1600 ISO

I obviously see noise in a typical 1600 ISO sub, but I don't see the same diagonal pattern.

100% crop from single converted RAW, 90 sec, 1600 ISO

This crop also has been stretched as the stack was. Anyone know what's going on here?

It is pattern noise.  The T2i is an old generation sensor without on-sensor dark current suppression, and with high pattern noise.  You need to do dark frame subtraction to reduce this problem, or better get a newer camera.

The directional nature is due to drift from frame to frame from less than ideal polar alignment, so the pattern drifts in one direction.  Dithering would help too: offset the frame a few pixels in random directions every few frames.

Roger

OP Lyle Aldridge Senior Member • Posts: 1,151
Re: Brush stoke noise - what's happening?

rnclark wrote:

The directional nature is due to drift from frame to frame from less than ideal polar alignment, so the pattern drifts in one direction. Dithering would help too: offset the frame a few pixels in random directions every few frames.

Roger

Thanks for the input, Roger.

I'm really questioning, however, whether much of this is due to polar alignment. This mount is bolted to a concrete pier with the best PA I've ever been able to achieve, more effort than most people seem to think an LXD mount is really worth. Here is a 100% crop of the last image in the sequence overlaid on the first, with the top (last image) set to "difference." As I flip through the images in my viewer, the change from frame to frame is completely imperceptible without doing this kind of overlay, and these two (as I presume would be expected) show the most drift of the several combinations I tried.

That's obviously not perfect, but is it enough to produce the directional artifact I'm seeing in the stack? Is it possible that I need to dither a bit to get more drift?

Also, is there some reason I'm not understanding that causes the 24 dark frames in the stack to make the effect worse? My T2i is certainly getting long in the tooth, but it's never done this kind of thing before.

Matt Fulghum
Matt Fulghum Senior Member • Posts: 1,024
Re: Brush stoke noise - what's happening?

In my (limited, take what I say with a large grain of salt) experience, the answer to both those questions is yes.  the diagonal striping is pattern noise being reinforced by mount drift.  dithering will help immensely, especially when coupled with calibration frames (darks and bias at the least) and sigma rejection stacking.  In DSS, try using the Sigma-Kappa stacking; in Siril I use Windsor sigma.

 Matt Fulghum's gear list:Matt Fulghum's gear list
Fujifilm X-M1 Fujifilm X-E2S Fujifilm X-T20 Fujifilm X-T3 Fujifilm X-T4 +6 more
OP Lyle Aldridge Senior Member • Posts: 1,151
Re: Brush stoke noise - what's happening?

Matt Fulghum wrote:

In my (limited, take what I say with a large grain of salt) experience, the answer to both those questions is yes. the diagonal striping is pattern noise being reinforced by mount drift. dithering will help immensely, especially when coupled with calibration frames (darks and bias at the least) and sigma rejection stacking. In DSS, try using the Sigma-Kappa stacking; in Siril I use Windsor sigma.

I guess I'm getting progressively confused. Are you saying that a lack of drift causes the same pattern noise to appear on the same pixels, so that it gets reinforced by the stacking, rather than randomized? It would seem ironic to me that I've worked so hard to get this mount to the point where it tracks with less drift than it ever has, and that this is causing a new set of problems.

Also, as indicated before, I've noticed that adding two dozen 90-sec dark frames, shot right after the lights, seemed to make things worse, rather than better. I did not include bias or flats, so I'll give it another try with the additional frames. At least those are easy to get.

Finally, FWIW, these were 90 second subs. I have no idea why the DPR mouse-over shows 30 for the frame in my post.

Matt Fulghum
Matt Fulghum Senior Member • Posts: 1,024
Re: Brush stoke noise - what's happening?

pretty much.  because the same pattern repeats so very close to itself, frame after frame, it doesn't trigger the pixel rejection in the stacking algorithms.  Dithering would spread the pattern noise out, and should really help out the stack.

 Matt Fulghum's gear list:Matt Fulghum's gear list
Fujifilm X-M1 Fujifilm X-E2S Fujifilm X-T20 Fujifilm X-T3 Fujifilm X-T4 +6 more
rnclark Veteran Member • Posts: 3,957
Re: Brush stoke noise - what's happening?

Lyle Aldridge wrote:

Matt Fulghum wrote:

In my (limited, take what I say with a large grain of salt) experience, the answer to both those questions is yes. the diagonal striping is pattern noise being reinforced by mount drift. dithering will help immensely, especially when coupled with calibration frames (darks and bias at the least) and sigma rejection stacking. In DSS, try using the Sigma-Kappa stacking; in Siril I use Windsor sigma.

I guess I'm getting progressively confused. Are you saying that a lack of drift causes the same pattern noise to appear on the same pixels,

It looks like you still had some drift, even if not much.  It is that small drift that caused the "brush stroke" noise.  An offset in the other direction would help. So during imaging, every few frames move the view up, down, left, or right just a tiny amount.

so that it gets reinforced by the stacking, rather than randomized? It would seem ironic to me that I've worked so hard to get this mount to the point where it tracks with less drift than it ever has, and that this is causing a new set of problems.

Also, as indicated before, I've noticed that adding two dozen 90-sec dark frames, shot right after the lights, seemed to make things worse, rather than better.

That may be due to the darks being at a different temperature than the lights, so the pattern noise is different.  Dark current doubles for every 5 to 6 C increase in temperature.

I did not include bias or flats, so I'll give it another try with the additional frames. At least those are easy to get.

Finally, FWIW, these were 90 second subs. I have no idea why the DPR mouse-over shows 30 for the frame in my post.

That is a dpreview bug.

Roger

swimswithtrout Veteran Member • Posts: 3,758
Re: Brush stoke noise - what's happening?

Lyle Aldridge wrote:

This forum has definitely piqued my interest in using camera lenses instead of my telescope. After admiring the Rho Ophi shots posted by others, I gave it a try a few nights ago, using my unmodified Canon T2i and 300mm/F4L lens, the only true high-quality lens I own. I shot 28 unguided subs of 90 seconds each, using my LXD-55 mount, plus 24 darks. No flats or bias.

My immediate problem is the diagonal brush stroke-like pattern that appears in the 100% crop from the stacked image below. This particular stack is from Sequator, but I get the same pattern when stacking with DSS. The DSS stacks - median or average - just show less color. This first has been stretched just a bit in PS levels to make the pattern a bit more visible, but not processed beyond that. I seem to get the same pattern whether I use the darks or not. In fact, the pattern is just slightly less evident when I stack without them, again in either stacking program.

100% crop from Stack of 28x90sec, 1600 ISO

I obviously see noise in a typical 1600 ISO sub, but I don't see the same diagonal pattern.

100% crop from single converted RAW, 90 sec, 1600 ISO

This crop also has been stretched as the stack was. Anyone know what's going on here?

That's a text book case of Correlated Noise/ aka Walking Noise. Your mount is drifting, even if slightly, in the same direction. Your camera suffers from FPN (fixed pattern noise). The drift is so small from frame to frame that it can't be eliminate by stacking, but is instead amplified into what you're seeing in your photo.

Here's a close crop of one of my old photos, from a shot of the Owl Nebula using a 300mm telephoto lens on a "camera tracker" showing the same thing.

Darks and bias frames will help a little, but dithering will completely eliminate it. ( As will a better camera)

OP Lyle Aldridge Senior Member • Posts: 1,151
Re: Brush stoke noise - what's happening?

OK, it takes a while, but the lights are coming on. Thanks much to Roger, Matt and SWT.

It's too late for me to dither, but I'm hatching a plan to re-shoot the darks and collect the other calibration frames to see if I can salvage anything from this series. My usual practice has been to gather the darks after the lights. With our rapidly falling temps after sundown in southern Arizona, I guess I could be getting darks shot at much cooler temps than the lights.

Sir Canon
Sir Canon Senior Member • Posts: 1,572
Re: Brush stoke noise - what's happening?

rnclark wrote:

Lyle Aldridge wrote:

This forum has definitely piqued my interest in using camera lenses instead of my telescope. After admiring the Rho Ophi shots posted by others, I gave it a try a few nights ago, using my unmodified Canon T2i and 300mm/F4L lens, the only true high-quality lens I own. I shot 28 unguided subs of 90 seconds each, using my LXD-55 mount, plus 24 darks. No flats or bias.

My immediate problem is the diagonal brush stroke-like pattern that appears in the 100% crop from the stacked image below. This particular stack is from Sequator, but I get the same pattern when stacking with DSS. The DSS stacks - median or average - just show less color. This first has been stretched just a bit in PS levels to make the pattern a bit more visible, but not processed beyond that. I seem to get the same pattern whether I use the darks or not. In fact, the pattern is just slightly less evident when I stack without them, again in either stacking program.

100% crop from Stack of 28x90sec, 1600 ISO

I obviously see noise in a typical 1600 ISO sub, but I don't see the same diagonal pattern.

100% crop from single converted RAW, 90 sec, 1600 ISO

This crop also has been stretched as the stack was. Anyone know what's going on here?

It is pattern noise. The T2i is an old generation sensor without on-sensor dark current suppression, and with high pattern noise. You need to do dark frame subtraction to reduce this problem, or better get a newer camera.

The directional nature is due to drift from frame to frame from less than ideal polar alignment, so the pattern drifts in one direction. Dithering would help too: offset the frame a few pixels in random directions every few frames.

Roger

i had the same problem with my t2i

-- hide signature --

I tend to overdo things

 Sir Canon's gear list:Sir Canon's gear list
Panasonic Lumix DMC-TS1 Canon EOS 350D Canon EOS 550D Canon EOS 700D Canon EF 50mm F1.8 II +6 more
Matt Fulghum
Matt Fulghum Senior Member • Posts: 1,024
Re: Brush stoke noise - what's happening?

yeah, you really want to temperature match the darks as closely as possible to the lights; that's where cooled astro cams really have a leg up.

Even those suffer from FPN, though; I'll stick a shot of mine down below for illustrative purposes. The best way to cure it, short of a new camera, is definitely dithering and good calibration files; I bought a copy of APT specifically for the ability to program dithering into my capture sessions more easily.

This is a pretty extreme example; something like 700 5s exposures, unguided, on an AVX.  you can clearly see the walking pattern at the bottom of the image.

 Matt Fulghum's gear list:Matt Fulghum's gear list
Fujifilm X-M1 Fujifilm X-E2S Fujifilm X-T20 Fujifilm X-T3 Fujifilm X-T4 +6 more
rnclark Veteran Member • Posts: 3,957
Re: Brush stoke noise - what's happening?

Matt Fulghum wrote:

yeah, you really want to temperature match the darks as closely as possible to the lights; that's where cooled astro cams really have a leg up.

A cooled camera certainly has advantages, but now that cmos sensors have excellent on-sensor dark current suppression, and lower dark current at a given temperature than traditional CCDs, the advantage of the cooled CCD is much less, except in hot environments.

Keyboard shortcuts:
FForum MMy threads