Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

How do you fix "walking noise" in processing?

  • Please log in to reply
27 replies to this topic

#1 Phillip Creed

Phillip Creed

    Soyuz

  • *****
  • topic starter
  • Posts: 3,888
  • Joined: 25 Jul 2006
  • Loc: Canal Fulton, OH

Posted 13 August 2019 - 09:56 AM

I got a 64-minute integration of M31 from earlier this month.  Only issue with it is this weird diagonal banding.  In reading other posts, it's often referred to as, "walking noise" and its best preventative measure is dithering.

That's fine for fixing it BEFORE it happens.  But if it's already there, what can you do to reduce/eliminate it in PixInsight?  I just downloaded the trial version.

Clear Skies,

Phil

Attached Thumbnails

  • M31-Stack-J2-CN.jpg


#2 fmeschia

fmeschia

    Surveyor 1

  • *****
  • Posts: 1,631
  • Joined: 20 May 2016
  • Loc: Mountain View, CA

Posted 13 August 2019 - 10:04 AM

You can try to remove/reduce it with bias and dark calibration. if it doesn’t work, you may be out of luck, I don’t think there’s any process to remove walking noise in PixInsight.

Francesco



#3 james7ca

james7ca

    Voyager 1

  • *****
  • Posts: 10,031
  • Joined: 21 May 2011
  • Loc: San Diego, CA

Posted 13 August 2019 - 10:15 AM

Sixty-four minutes isn't much integration time and it's unlikely that you can do much to eliminate the walking noise (other than with better calibration). However, more integration time will probably help, since any new session is unlikely to show exactly the same pattern of noise (so, to some degree it will average out over time).

 

However, you might try a round of cosmetic correction on your calibrated light frames before they are registered and integrated. That may or may not help, but if you are going to be limited to this one data set it might be worth a try.


Edited by james7ca, 13 August 2019 - 10:37 AM.

  • pfile likes this

#4 SilverLitz

SilverLitz

    Surveyor 1

  • -----
  • Posts: 1,700
  • Joined: 17 Feb 2018
  • Loc: Louisville, KY

Posted 13 August 2019 - 10:26 AM

Last year I had walking noise with some DSS stacked exposures, using a Dark library that had temps over 10C range.  I then sorted/stacked sub-groups in 1C sets (lights/darks) and the walking noise went away, on the same data.  I then stacked the 1C sub-group stacks.


Edited by SilverLitz, 13 August 2019 - 10:27 AM.

  • Jerry Lodriguss likes this

#5 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 14,653
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 13 August 2019 - 10:43 AM

The reason people always suggest dithering is that it is very difficult to do much with it after the fact.  I dither all the time now so I never have to deal with it any more.  But unfortunately, I don't have any suggestions other than the usual noise reduction techniques (which don't work too well on it), or just burying the noise by clipping the blacks.

 

-Dan



#6 jerahian

jerahian

    Surveyor 1

  • *****
  • Moderators
  • Posts: 1,500
  • Joined: 02 Aug 2018
  • Loc: Maine

Posted 13 August 2019 - 11:26 AM

Embrace dithering.  I have come to accept it as an essential noise reduction element to imaging, and now dither every frame in RA & Dec.  Sadly, I don't know how to approach the removal of walking noise in post, but Mark's experience in post #4 regarding wide ranging darks is interesting.



#7 spokeshave

spokeshave

    Mercury-Atlas

  • *****
  • Posts: 2,608
  • Joined: 08 Apr 2015

Posted 13 August 2019 - 12:29 PM

Whenever the topic of walking noise comes up, there is a chorus of people who say that dithering is the solution. Dithering certainly helps and should always be done. However even dithered images can show walking noise depending on the scales of the correlated noise and the dither. I try to jump in where I can to add that dithering treats the symptoms, but it is better to cure the disease - field drift. Find out what is causing the drift that is producing the walking noise, cure it, and the problem goes away (you'll still want to dither though). Drift is usually due to one of two things - polar alignment error and differential flexure. Find out which one it is, and cure the disease instead of just treating the symptoms.

 

As for reducing the correlated noise after the fact, other than traditional noise reduction techniques, there unfortunately isn't much that can be done.

 

Tim


  • dswtan, bobzeq25, tkottary and 1 other like this

#8 sharkmelley

sharkmelley

    Aurora

  • *****
  • Posts: 4,825
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 13 August 2019 - 01:15 PM

I don't think you indicated what camera you are using.  It could be important but for now I'll assume you are using a one shot colour camera.

 

For most one shot cameras you need to do all the following:

  • Use dark optimisation (assuming it is available in your processing suite)
  • Use bayer drizzle stacking with sigma rejection and play with the pixel rejection parameters (again assuming it is available)
  • Use good noise reduction on the final stack

A couple of further explanations:

  • Dark subtraction and/or dark optimisation is not very effective on some Nikon DSLR cameras.  This is an unfortunate side-effect of the spatial filtering that's applied to the raw data but changes from model to model.
  • I recommend Bayer drizzle stacking because it leads to finer granularity noise in the stack.  Finer granularity noise is easier to remove with noise reduction.

Mark


Edited by sharkmelley, 13 August 2019 - 01:21 PM.

  • bobzeq25 likes this

#9 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,610
  • Joined: 27 Oct 2014

Posted 13 August 2019 - 01:23 PM

I'm not actually recommending this, there's a lot more to learn about PI first.  But...

 

You can make it less noticeable by desaturating the background.  Create a starless version using various PI tools.  Use range mask to create a mask over everything but the background.  It will not be perfect.

 

Mask the image. Use ColorSaturation to desaturate the background.

 

Your eyes will respond less to the noise if it's desaturated.

 

I pretty much agree with Spokeshave.  But get the dither right in amount and randomness, and the correlation between the noise and the dither will be small.

 

Another way to reduce the issue is owning a camera with low fixed noise, and doing RGB imaging.  That camera is mono, of course.  <smile>

 

Pretty much agree with Sharkmelley also.  But if the camera is CMOS, and has amp glow, optimization dooesn't work well.


Edited by bobzeq25, 13 August 2019 - 01:24 PM.

  • Monsy likes this

#10 Phillip Creed

Phillip Creed

    Soyuz

  • *****
  • topic starter
  • Posts: 3,888
  • Joined: 25 Jul 2006
  • Loc: Canal Fulton, OH

Posted 13 August 2019 - 03:24 PM

I don't think you indicated what camera you are using.  It could be important but for now I'll assume you are using a one shot colour camera.

 

For most one shot cameras you need to do all the following:

  • Use dark optimisation (assuming it is available in your processing suite)
  • Use bayer drizzle stacking with sigma rejection and play with the pixel rejection parameters (again assuming it is available)
  • Use good noise reduction on the final stack

A couple of further explanations:

  • Dark subtraction and/or dark optimisation is not very effective on some Nikon DSLR cameras.  This is an unfortunate side-effect of the spatial filtering that's applied to the raw data but changes from model to model.
  • I recommend Bayer drizzle stacking because it leads to finer granularity noise in the stack.  Finer granularity noise is easier to remove with noise reduction.

Mark

Mark,

The camera is a Canon T6 (EOS 1300D).

For the image, I used:

 

128 x 30-sec @ 1600 ISO for lights
30 darks
30 flats (histogram just past 1/2)
100 bias

 

Mark,  Deep Sky Stacker has a 2x2 drizzle option for drizzle stacking.  It takes a span of geological time and results in a file that's about half a gigabyte, but might be worth it.

Clear Skies,

Phil



#11 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 10,778
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 13 August 2019 - 06:06 PM

Reducing field drift will make the pattern noise less noticeable but it will still be there. Dithering will actually reduce the noise and should always be done - and it will help.

I don’t know what post processing techniques would work best. I would just dither next time.

Frank

#12 Phillip Creed

Phillip Creed

    Soyuz

  • *****
  • topic starter
  • Posts: 3,888
  • Joined: 25 Jul 2006
  • Loc: Canal Fulton, OH

Posted 14 August 2019 - 10:26 AM

I did an Automatic Background Extraction last night and that seemed to help a little bit.  I'm sure there are other measures I could take, but it is really tough learning the myriad of functions PixInsight has, as well the proper settings for each.

PixInsight is a powerful tool, but it's not intuitive.  It's like reading through an obscure IRS tax form or the lyrics to Beck's, "Loser".

I'm convinced this software was created by aliens.  Benevolent aliens that want Earthlings to improve their astrophotos, perhaps.  But aliens, nonetheless.

Aliens, I tell you.

Aliens.

Clear Skies,

Phil


  • AstroVagabond likes this

#13 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 14,653
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 August 2019 - 10:36 AM

Whenever the topic of walking noise comes up, there is a chorus of people who say that dithering is the solution. Dithering certainly helps and should always be done. However even dithered images can show walking noise depending on the scales of the correlated noise and the dither. I try to jump in where I can to add that dithering treats the symptoms, but it is better to cure the disease - field drift. Find out what is causing the drift that is producing the walking noise, cure it, and the problem goes away (you'll still want to dither though). Drift is usually due to one of two things - polar alignment error and differential flexure. Find out which one it is, and cure the disease instead of just treating the symptoms.

 

As for reducing the correlated noise after the fact, other than traditional noise reduction techniques, there unfortunately isn't much that can be done.

 

Tim

I've also seen a chorus of people saying that the problem is field drift. :-)

 

To be honest, I have seen no data to back up that claim, and have seen no one who has fixed this problem by fixing polar alignment or differential flex.  And in fact, when I used to get this, I did not have either of those problems.  My scope is in an observatory and the polar alignment is spot on.  And if I had differential flex, I'd have guiding problems which I do not.

 

If you look at the noise closely, you'll see that the streaks can be quite long, which would require a very large amount of field drift - something I've never seen.  In addition, the noise tends to be wavy, with differently waviness in each streak.  It looks nothing like the star streaks you see from field drift.  

 

I understand that the hypothesis is that guiding is bringing field drift back into place, which make the noise "walk" across the image relative to the star locations.  But again, that does not square with other clear indications in the image.

 

I have no idea what causes this noise pattern, but I don't buy the field drift hypothesis.  Fact is, whatever causes it, dithering fixes it.

 

-Dan



#14 spokeshave

spokeshave

    Mercury-Atlas

  • *****
  • Posts: 2,608
  • Joined: 08 Apr 2015

Posted 14 August 2019 - 11:24 AM

I've also seen a chorus of people saying that the problem is field drift. :-)

 

To be honest, I have seen no data to back up that claim, and have seen no one who has fixed this problem by fixing polar alignment or differential flex.  And in fact, when I used to get this, I did not have either of those problems.  My scope is in an observatory and the polar alignment is spot on.  And if I had differential flex, I'd have guiding problems which I do not.

 

If you look at the noise closely, you'll see that the streaks can be quite long, which would require a very large amount of field drift - something I've never seen.  In addition, the noise tends to be wavy, with differently waviness in each streak.  It looks nothing like the star streaks you see from field drift.  

 

I understand that the hypothesis is that guiding is bringing field drift back into place, which make the noise "walk" across the image relative to the star locations.  But again, that does not square with other clear indications in the image.

 

I have no idea what causes this noise pattern, but I don't buy the field drift hypothesis.  Fact is, whatever causes it, dithering fixes it.

 

-Dan

Flexure can cause slow drift that is not obvious over the course of minutes in individual subs but can become apparent over the course of hours for an entire imaging session. Additionally, you can have fairly serious flexure and see no evidence of it at all when guiding. Guiding cannot correct flexure - by definition, it happens independent of guiding. I also don't consider field drift to be a hypothesis. I thought it was fairly well-established fact. There is one simple, foolproof way to prove to yourself whether field drift cause walking noise or not. Stack the subs without registering them. If the target and stars drift across the frame in the same direction as the "walking noise" streaks, then it is utterly undeniable that the field is drifting and that is causing the streak pattern in the noise structures. 

 

Dithering reduces, but does not eliminate, fixed pattern noise. Any remaining FPN will still create walking noise - albeit of lower amplitude - if there is field drift. I have seen numerous examples of walking noise even when dithering is being done. I'm not advocating against dithering - I thought I made that pretty clear. I am advocating that in addition to dithering, one should look for and correct any sources of field drift. In addition to creating walking noise, field drift can cause several other problems. I think correcting it is sound advice and I'm not sure I understand why anyone would oppose it.

 

Tim


  • Tim, jdupton and bmhjr like this

#15 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,668
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 14 August 2019 - 12:32 PM

Dan,

 

I've also seen a chorus of people saying that the problem is field drift. :-)

 

To be honest, I have seen no data to back up that claim, and have seen no one who has fixed this problem by fixing polar alignment or differential flex.  And in fact, when I used to get this, I did not have either of those problems.  My scope is in an observatory and the polar alignment is spot on.  And if I had differential flex, I'd have guiding problems which I do not.

 

If you look at the noise closely, you'll see that the streaks can be quite long, which would require a very large amount of field drift - something I've never seen.  In addition, the noise tends to be wavy, with differently waviness in each streak.  It looks nothing like the star streaks you see from field drift.  

 

I understand that the hypothesis is that guiding is bringing field drift back into place, which make the noise "walk" across the image relative to the star locations.  But again, that does not square with other clear indications in the image.

 

I have no idea what causes this noise pattern, but I don't buy the field drift hypothesis.  Fact is, whatever causes it, dithering fixes it.

 

-Dan

 

   If you ever see the walking noise problem in any data you have full access to, you can verify the field drift and differential flexure hypothesis for yourself. It is very easy to do.

 

   To see if your field is drifting or not, stack all of your raw subs for an evening's imaging session. If you have four hours of data, just stack the raw subs. Do not calibrate them, do not align them. Simply feed the saved images right out of the camera into a stacking program.

 

   Look at the result. You will see how much drift in position you had over the full evening of imaging. Measure the length of the star trails in the unaligned integrated image in term of pixels. Now look at your properly calibrated, aligned, and stacked image with the walking noise and measure the length of the faint streaks you see there. If the two measurements agree and the streaks follow the same directional paths, then that shows that the noise is due to the field drift over the course of a session. (By the way, the streaks and field drift will not always be linear. Polar alignment error and differential gravitational pull in different directions as the mount tracks will often lead to curved star trails.)

 

   Long ago, I used this method to find out where my own differential flexure was coming from. I would stack my unaligned images after each session and note whether I was seeing RA, DEC, or changing ALT-AZ drift. I would make a change to my optical and guiding system attachments and repeat after the next session. I was able to figure out which components were most contributing to the flexure and associated field drift.

 

   The main reason most folks don't think the two issues (field drift and walking noise) are related, is that their images may show no appreciable ill effects from the drift. As an example let's say that your image scale is such that your drift during a 3 minute exposure is only 1/2 of a pixel when the star images show a FWHM of, say, 3+ pixels. You are likely not able to easily see the very slight elongation buried under the seeing changes and guiding error. However, at the end of four hours of imaging, you will find that the stars actually trailed by 40 pixels. All the individual subs look good but stacking everything unaligned shows 40 pixel long star trails. Likely as not, the walking noise streaks will also be shown to be 40 pixels long.

 

 

John


  • TrustyChords likes this

#16 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 14,653
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 August 2019 - 01:36 PM

>> All the individual subs look good but stacking everything unaligned shows 40 pixel long star trails. Likely as not, the walking noise streaks will also be shown to be 40 pixels long.

 

If your stars have moved 40 pixels from the first image to the last, either you're not guiding or your guiding is doing a horrible job.  And in fact, the length of the streaks in "walking noise" is far more than that.  For field drift to be the issue, you'd have to have a significant, noticeable change in position of the stars from the first image to the last.  And when you're done stacking, you'd have to see a darker edge to the image where the stacked frames did not align, which would be equal to the length of the noise streaks.  Never happens.

 

I don't believe I have any data from the days when I dealt with this kind of noise so I can't run the exact test you suggest.  But the fact is, you'd see other, obvious indications in the image, and your guiding would have to be nearly dead, for that much drift to occur.

 

-Dan



#17 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,668
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 14 August 2019 - 02:54 PM

Dan,

 

If your stars have moved 40 pixels from the first image to the last, either you're not guiding or your guiding is doing a horrible job.  And in fact, the length of the streaks in "walking noise" is far more than that.  For field drift to be the issue, you'd have to have a significant, noticeable change in position of the stars from the first image to the last.  And when you're done stacking, you'd have to see a darker edge to the image where the stacked frames did not align, which would be equal to the length of the noise streaks.  Never happens.

 

   That doesn't have to be the case. You can have field drift even in the presence of perfect guiding. That is why this is so often confusing and hard to wrap your mind around.

 

   Differential flexure has nothing to do with guiding, per se. It is the movement of one optical axis with respect to another during imaging. That is the very definition of differential flexure. Any time you are guiding with one optical system and imaging with another, you can have this field drift due to the differential flexure. 

 

   Imagine that could save each of your guide exposures over a four hour session in addition to saving your images gathered by the tethered main scope. At the end of the four hours, you stack all of the guide exposures. Because you are guiding, they will always line up and do so perfectly if guiding is perfect. Now consider the other main optical system which is tethered to your guide scope and is also imaging. 

 

   Over the four hours of your imaging system use, the optical axis of the main scope can slowly diverge from the optical axis of the guide scope. It could be that one of the scopes has a movable mirror such as in an SCT allowing it to tilt during movement across the sky. It could be that the mounting rings on the guide scope can allow a slight flexure under gravity as the mount tracks across the sky. It may even be due to using a long refractor whose tube can bend ever so slightly under its own weight. There are also cases where the focuser can droop a few ten-thousandths of an inch due to gravity during that four hours. In all cases, the optical axis of one scope does not point to exactly the same place when compared to the optical axis of the other scope over time.

 

   That divergence of optical axes over time is the definition of differential flexure. Even though the guiding can be perfect from the viewpoint of the guiding system, the field in the other scope can move a number of arc-seconds over a period of hours as their axes diverge. That shows up as field drift in the main scope's images.

 

   The usual cure for field drift due to differential flexure is to use an off-axis guider. That way, both the guide camera and the imaging camera see the exact same field drift and guiding prevents star drift over time. (Except in the case of field rotation due to polar misalignment which can also cause field drift and walking noise even with an OAG. In such a case, though, it is much, much less severe and not usually very noticeable.)

 

 

John


Edited by jdupton, 14 August 2019 - 03:06 PM.

  • schmeah, sharkmelley and bmhjr like this

#18 spokeshave

spokeshave

    Mercury-Atlas

  • *****
  • Posts: 2,608
  • Joined: 08 Apr 2015

Posted 14 August 2019 - 02:59 PM

If your stars have moved 40 pixels from the first image to the last, either you're not guiding or your guiding is doing a horrible job. 

That's just not true. If one is using a guide scope and the system has flexure, guiding can appear perfect, yet there can be significant drift. When guiding with a guide scope, the guide software controls where the guide scope is pointed, not the imaging scope. If there is differential flexure in the system, the imaging scope can change its pointing position relative to the guide scope as the mount moves. It is a very well-documented and well-understood effect. Your issues may have been different, but that doesn't mean that the effect doesn't exist and shouldn't be corrected.

 

Tim


  • schmeah, jdupton and sharkmelley like this

#19 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,668
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 14 August 2019 - 03:46 PM

Phil,

 

 

That's fine for fixing it BEFORE it happens.  But if it's already there, what can you do to reduce/eliminate it in PixInsight?  I just downloaded the trial version.

 

   Back to your original question now that this thread is on my radar. I sort of hate to do this since you are going to be trying to learn PixInsight but you can address the issue to some extent in image processing. I can describe the areas to investigate but it will likely seem like Greek (or Geek since this is PixInsight) to you for a while.

 

   You cannot completely eradicate the walking noise in processing but can knock it down so that it is much less noticeable. I don't expect you to follow this like a tutorial but instead memorize it for now and play with it as you learn to use PixInsight with the many written and video tutorials that are available.

 

   Now some good news -- since you can see the walking noise in your image, that means you had field movement during the imaging session. That is normally bad but the movement is sort of like dithering -- the movement of the field of view on a frame to frame basis over time. Unlike dithering which is random, your field movement is linear. That means it can be eye-catching in the image. But movement of any kind can help a little in this case.

 

   There are two main opportunities to reduce the effects of the walking noise in your images while you process them as below. Both of these will slightly degrade some of your image data but will help with reducing the visibility of the walking noise.

  • During Cosmetic Correction
    Normally, after calibrating your images with your Dark Frames and Flat Frames, you will want to apply Cosmetic Correction to them. The opportunity here is to be a bit more aggressive with the treatment of Hot and Cold pixels from your sensor. When you open the Cosmetic Correction Process, you will use both the Master Dark and Auto Detect sections of the tool. Load your Master Dark Frame in the tool. Enable both the Hot Pixels and Cold Pixels portions of the Master Dark section. Be sure the CFA check box is selected in the top section of the process tool.

    With your EOS 1300D, I would suggest using a heavy-handed 12,000 hot pixel count in the Hot Pixel Qty entry box and 3,000 in the Cold Pixel Qty box.

    In the Auto Detect section, enable both Hot and Cold options. Try 2.5 sigma units for each. 

    Run the Cosmetic Correction Process Tool against all your calibrated light frames. If this is too restrictive, you can retry later using a lower number pixels for the Qty numbers in the Master Dark section and higher Sigma units (like 3.0, the default) in the Auto Detect section.
     
  • During Image Integration
    Once you have converted the raw images to color and are ready to run the Image Integration Process, you only need to change the Sigma rejection settings in the Pixel Rejection (2) section of the tool. Here, use a lower number for both Low and High. Set Sigma Low from its default of 4.0 to maybe something like 3.0. Change the Sigma High from its default of 3.0 to something like 2.5. You can experiment with these number to see their effect. For changing Sigma High (which directly pertains to helping hide the walking noise), lower numbers will reduce the visibility but will lower the Signal-To-Noise ratio (~quality) of the image. Higher numbers for Sigma High will keep more of your data but also allow more of the walking noise to become visible.

   I hope this turns out to be of some help to you. The best solution is prevention with random dithering during capture, good polar alignment, and good control of differential flexure during image capture. Once you have the data though, careful calibration of the images with Dark Frames and Flats along with heavier application of Cosmetic Correction and tighter rejection settings during integration can help save your existing data.

 

 

John


Edited by jdupton, 14 August 2019 - 04:01 PM.

  • bobzeq25 and zoltankotg like this

#20 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 14,653
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 August 2019 - 04:42 PM

>> The usual cure for field drift due to differential flexure is to use an off-axis guider.

 

>> If one is using a guide scope and the system has flexure, guiding can appear perfect, yet there can be significant drift. When guiding with a guide scope...

 

Yep, I'm very familiar with what differential flex is.  I've had "walking noise" when using an off axis guider with my SCT.

 

Again, I've seen all these explanations before.  What I haven't seen is someone who has actually had "walking noise" and fixed it by using an off-axis guider, by correcting their polar alignment, or by fixing differential flex.

 

All these explanations kinda sound reasonable and kinda make sense.  But sounding reasonable is not the same as evidence that it's the cause.  Especially when there is evidence to the contrary.  My own personal experience and what I've seen in my images casts serious doubts on these explanations.  I'm not saying they're wrong.  But I'm definitely not convinced they're right either.  I'd need to see something more definitive to be convinced.

 

-Dan



#21 spokeshave

spokeshave

    Mercury-Atlas

  • *****
  • Posts: 2,608
  • Joined: 08 Apr 2015

Posted 14 August 2019 - 07:05 PM

Well, I demonstrated to myself to my own satisfaction that walking noise resulted from field drift. I had the same problem, particularly when I was still using a DSLR. Dithering helped - a lot. But all it really did was let me stretch the background more before the correlated noise became evident again. It became painfully evident that field drift was the cause when I stacked the unregistered images. The stars made the same trails that the noise made in the registered stack. That was unequivocal proof to me that the streaks in my case were caused by field drift. I fixed it by going to an OAG. So now you do know someone who had walking noise and fixed it with an OAG. The only mechanism that I am aware of that can cause FPN to produce a linear pattern in a stack is field drift. Over several years in the dozens of threads on this topic in this forum and others, I have never seen anyone present an alternate explanation whereas I have seen several people cure the problem by eliminating drift.

 

I have no explanation for your case. It's a shame that you no longer have any subs that show the walking noise. I would have like to study them to try to understand why they don't fit the widely-accepted explanation.

 

Tim


  • jdupton likes this

#22 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 10,778
  • Joined: 12 Oct 2007
  • Loc: Melbourne, Australia

Posted 14 August 2019 - 07:13 PM

If you have pattern noise and there is drift then the resulting pattern noise will decrease as you increase the drift. The worst case in terms of noise is no drift or dithering at all.

But the drift has a huge effect perceptually because it produces structure the eye immediately picks up on. But it doesn’t pick up on randomish noise with little structure that would be present strongly with no drift.

Dithering randomizes it so it evens out in the stack and is greatly reduced - not just perceptually but as a noise term also.

For the op - since it is streaky you may get benefit from a smoothing kernel that is asymmetric and has greater smoothing perpendicular to the streaks.

Frank

Edited by freestar8n, 14 August 2019 - 07:14 PM.

  • jdupton likes this

#23 jdupton

jdupton

    Soyuz

  • *****
  • Posts: 3,668
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 14 August 2019 - 07:17 PM

Tim,

 

   Your experience here agrees with my own. Back in 2008, I struggled hard fighting the problem and figured out that it was indeed coming from the field drift using unaligned stacking just as you did. I struggled on for two more years and then did an in-depth analysis of the problem and where in my imaging system it originated. Finally, in 2013, I bit the bullet and went with an OAG. At that point, walking noise was no longer a problem for me. I actually didn't start consistently dithering for three more years but the causes and contributors were clear me.

 

 

John



#24 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 14,653
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 August 2019 - 09:17 PM

John & Tim:

 

Thanks for posting!  You are the first I've seen that have posted with actual experiences that fixed the  walking noise by eliminating field drift.  Definitely makes me reconsider my position. :-)

 

-Dan


  • spokeshave likes this

#25 schmeah

schmeah

    Fly Me to the Moon

  • *****
  • Posts: 6,156
  • Joined: 26 Jul 2005
  • Loc: Morristown, NJ

Posted 15 August 2019 - 05:33 AM

>> The usual cure for field drift due to differential flexure is to use an off-axis guider.

 

>> If one is using a guide scope and the system has flexure, guiding can appear perfect, yet there can be significant drift. When guiding with a guide scope...

 

Yep, I'm very familiar with what differential flex is.  I've had "walking noise" when using an off axis guider with my SCT.

 

Again, I've seen all these explanations before.  What I haven't seen is someone who has actually had "walking noise" and fixed it by using an off-axis guider, by correcting their polar alignment, or by fixing differential flex.

 

All these explanations kinda sound reasonable and kinda make sense.  But sounding reasonable is not the same as evidence that it's the cause.  Especially when there is evidence to the contrary.  My own personal experience and what I've seen in my images casts serious doubts on these explanations.  I'm not saying they're wrong.  But I'm definitely not convinced they're right either.  I'd need to see something more definitive to be convinced.

 

-Dan

I can tell you with absolute certainty that I had consistent walking noise despite excellent guiding. Added an OAG. Gone. Differential flexure. Unfortunately I can’t go back 10 years to find the data.

 

Derek




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics