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

Grid Pattern in 1600MM-C.

  • Please log in to reply
13 replies to this topic

#1 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 11:30 AM

I've noticed that sometimes a grid pattern shows up in my images.   This isn't the end of the world as it tends to only show up at extreme magnification like 500% but its interesting to me as it seems to be a grid of like sized "blocks" that seem to alternate between sharp and slightly softer.  

 

I tried to capture it as best as I could.   I pushed these images a bit to enhance what I'm trying to show.  

 

7r6nI4w.jpg

DbtrG5a.jpg

a0DzqCC.jpg

 

Do you see it too or have I completely lost my mind?  laugh.gif

 

I only happened upon this because I was looking in close at a certain area, has any of you ever seen this before?    I only pushed these images hard to help show off the grid a bit more but it stands out fairly well without pushing it.   The S2 filters shows it better than an HA.

 

Possible artifacts from calibration?   These images both have flats and darks applied, no bias.  



#2 jdupton

jdupton

    Apollo

  • *****
  • Posts: 1311
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 11 October 2018 - 11:44 AM

ciraxis,

 

   I am not seeing anything at all. Is the third image supposed to be a clue as to where to look for the grid boundaries? I'm not sure what I should be looking for.

 

   Is it possible you are seeing an aliasing effect (like a Moire pattern) between the resolution of the display screen and the background noise in the image?

 

 

John


Edited by jdupton, 11 October 2018 - 11:46 AM.


#3 jerahian

jerahian

    Sputnik

  • *****
  • Posts: 27
  • Joined: 02 Aug 2018
  • Loc: Maine

Posted 11 October 2018 - 12:06 PM

ciraxis, I see it too, although it is very, very subtle (i.e. personally, I wouldn't be worried about it too much).

 

That said, I believe this is one of the primary issues with CMOS sensors.  CMOS sensors are prone to pattern noise due to unevenness between the individual pixel cells and multiple A/D circuits in the readout.  In CMOS sensors, light striking the pixel creates a voltage proportional to its intensity.  The voltage is then sampled directly *at the pixel* and digitized on the imager.  This makes readout faster and consumes less power, but the downside is the pattern noise.

 

In comparison, CCD sensors are traditionally more expensive due to their fidelity in the way pixels are read out through a common circuit and their use of microlenses at each pixel for increased sensitivity (sensitivity is quickly becoming or has become a non-issue as CMOS sensors have improved greatly over the years).

 

So, in the end, I don't think you've lost your mind.  However, you shouldn't be staring at 500% magnified image pixels for too long, just in case. :p

 

Regards,

Ara



#4 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 12:16 PM

ciraxis, I see it too, although it is very, very subtle (i.e. personally, I wouldn't be worried about it too much).

 

That said, I believe this is one of the primary issues with CMOS sensors.  CMOS sensors are prone to pattern noise due to unevenness between the individual pixel cells and multiple A/D circuits in the readout.  In CMOS sensors, light striking the pixel creates a voltage proportional to its intensity.  The voltage is then sampled directly *at the pixel* and digitized on the imager.  This makes readout faster and consumes less power, but the downside is the pattern noise.

 

In comparison, CCD sensors are traditionally more expensive due to their fidelity in the way pixels are read out through a common circuit and their use of microlenses at each pixel for increased sensitivity (sensitivity is quickly becoming or has become a non-issue as CMOS sensors have improved greatly over the years).

 

So, in the end, I don't think you've lost your mind.  However, you shouldn't be staring at 500% magnified image pixels for too long, just in case. tongue2.gif

 

Regards,

Ara

I'm not worried about it at all, you can't even see it once you zoom out past 400%.    It was just an odd pattern I noticed, I know this s kind of a pointless topic but its cloudy and raining out here.   smile.gif



#5 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 12:19 PM

ciraxis,

 

   I am not seeing anything at all. Is the third image supposed to be a clue as to where to look for the grid boundaries? I'm not sure what I should be looking for.

 

   Is it possible you are seeing an aliasing effect (like a Moire pattern) between the resolution of the display screen and the background noise in the image?

 

 

John

 

 

The pattern is there, it is not a moire or anything similar.   This isn't a big deal to me, I just noticed the pattern and wanted to see if anyone else has seen it with their 1600


  • jerahian likes this

#6 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 21193
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 11 October 2018 - 12:26 PM

This is a registration artifact. It occurs usually when you are doing distortion correction, but may also occur if your data is particularly shallow with very small rotations.

 

There are a few things to try. FIRST, my assumption is your data is pretty shallow, and that can lead to clipped pixels. You need to use a pedestal when calibrating, ideally 800 DN (which matches a 50 ADU camera offset). This is the output pedestal in ImageCalibration.

 

That should fix the issue, however, if it does not...

 

Noisier data can be subject to registration artifacts like this, especially when using distortion correction, due to the very small changes made in the data. This leads to non-uniform interpolation artifacts in the data. One thing you can do to avoid this, is to get deeper subs. That requires getting longer subs, and if necessary, lowering gain, optimizing your tracking, and getting subs twice as long or more, might help swamp the read noise more, improve your signal offset, and minimize how obvious these artifacts look, if it doesn't eliminate them entirely.

 

Finally, if nothing else works, you can use a different interpolation algorithm. Cubic spline will soften the data a bit and bloat the stars a bit more, but it will usually completely eliminate any registration artifacts. You can compensate for the loss in detail a bit by drizzling, deconvolving, and downsampling again:

 

JjVm5kd.gif



#7 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 12:34 PM

This is a registration artifact. It occurs usually when you are doing distortion correction, but may also occur if your data is particularly shallow with very small rotations.

 

There are a few things to try. FIRST, my assumption is your data is pretty shallow, and that can lead to clipped pixels. You need to use a pedestal when calibrating, ideally 800 DN (which matches a 50 ADU camera offset). This is the output pedestal in ImageCalibration.

 

That should fix the issue, however, if it does not...

 

Noisier data can be subject to registration artifacts like this, especially when using distortion correction, due to the very small changes made in the data. This leads to non-uniform interpolation artifacts in the data. One thing you can do to avoid this, is to get deeper subs. That requires getting longer subs, and if necessary, lowering gain, optimizing your tracking, and getting subs twice as long or more, might help swamp the read noise more, improve your signal offset, and minimize how obvious these artifacts look, if it doesn't eliminate them entirely.

 

Finally, if nothing else works, you can use a different interpolation algorithm. Cubic spline will soften the data a bit and bloat the stars a bit more, but it will usually completely eliminate any registration artifacts. You can compensate for the loss in detail a bit by drizzling, deconvolving, and downsampling again:

 

JjVm5kd.gif

Interesting.   

 

The data above is only two hours worth of imaging so its not a lot of information.   I only notice these artifacts at 400%, 300% if the image is really pushed which I've been trying not to do.   Really no one is going to look at these at that size.   My current mount has just been awful and will rarely give me good subs past 3 min.   I've been sticking with 2min subs 200/50 settings on the camera, probably not ideal but it is what it is.   I'm happy with the results I've been getting.   I've been working on shooting at least 4 hours per filter

 

Also I only recently purchased PI and have yet to stack all of my images in it, I've been using Nebulosity for all of my image stacking, I use translation plus rotation when doing so.  I have been reading a bit here and there to learn PI but my time unfortunately has been short, which sucks because I fully realize the potential that I am missing out on.  


Edited by ciraxis, 11 October 2018 - 12:37 PM.


#8 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 12:39 PM

This is the HA image that the original pics are from, but this image isn't pushed to show the grid pattern.

 

uph0Eh6.jpg



#9 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 21193
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 11 October 2018 - 01:19 PM

The total integration time does not actually matter much here with regards to these registration artifacts. You register each sub individually, so each and every sub will get the artifacts. Depending on how large the artifacts are vs. how much you dither, even integrating the stack may not eliminate the artifacts (I've had that happen before).

 

Using Cubic Spline interpolation in StarAlignment in PI would eliminate the problem, at the cost of some resolution. Depending on how well sampled you are, that cost may be imperceptible...and looking at your data, that might be the case. So I would try cubic spline registration in PI if you can (and, yes, I would most definitely take the time to learn PI's pre-processing tools...all you really need is ImageCalibration, StarAlignment and ImageIntegration).

 

You should also definitely try using PI's ImageCalibration tool and add an 800 DN (16-bit) Output Pedestal. That itself may eliminate the artifacts. If not, use of a pedestal and cubic spline interpolation should do the trick. 



#10 happylimpet

happylimpet

    Mercury-Atlas

  • *****
  • Posts: 2934
  • Joined: 29 Sep 2013
  • Loc: Southampton, UK

Posted 11 October 2018 - 02:07 PM

The ASI1600MM has a chequerboard pattern of sensitivity. Usually darks and flats will take this out, but if for some reason this remains, it will show.

 

IN this case, I suspect that (as Jon R suggested above) you have a Moire effect from slight rotation between stacked subs, such that in some areas the chequerboards align, and in others they dont. This will lead to blocks of varying pixel-to-pixel level, as the chequerboards go in and out of phase.

 

Stack more subs or do more careful calibration, and i expect it will go.



#11 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 21193
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 11 October 2018 - 03:31 PM

The ASI1600MM has a chequerboard pattern of sensitivity. Usually darks and flats will take this out, but if for some reason this remains, it will show.

 

IN this case, I suspect that (as Jon R suggested above) you have a Moire effect from slight rotation between stacked subs, such that in some areas the chequerboards align, and in others they dont. This will lead to blocks of varying pixel-to-pixel level, as the chequerboards go in and out of phase.

 

Stack more subs or do more careful calibration, and i expect it will go.

I don't know that it is the checkerboard. I've found that flats correct that very well. 

 

I have found that if enough pixels are clipping to black (which can and will happen if you calibrate shallower data without a pedestal), this is all but guaranteed to happen. It also happens more with Lanczos interpolation than with some of the other interpolation algorithms.

 

A well constructed flat though, should effectively neutralize the checkerboard as it is part of the FPN the flats model and correct.


  • happylimpet likes this

#12 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 08:04 PM

If I get a chance this weekend I'll read up on pre processing in PI, it will be interesting to see the difference. I might try it on something that I have more time into. I just finished Pacman at around 12 hours.

I guess I have a lot of reading to do, I dont even know where to start. All part of the fun though, I enjoy learning

#13 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 7633
  • Joined: 12 Oct 2007

Posted 11 October 2018 - 08:23 PM

This looks like an interpolation artifact in stacking due to slight scaling and shifting of exposures. Try a different interpolation method.

If the problem only shows in zoomed out views it is probably just an aliasing artifact in the display and not in the data. But if it shows up when you zoom way in as it does here, then it is probably really there in the stack. But it isn’t due to the camera. Just the way the align and stack is done.

Frank

#14 ciraxis

ciraxis

    Vostok 1

  • -----
  • topic starter
  • Posts: 134
  • Joined: 12 Feb 2014

Posted 11 October 2018 - 08:26 PM

I have to really look hard to see it at 300%

Its really not that much of an issue for me but it is going to change the way I stack.


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