"Calibration and Alignment Only" introduces grid artifact

Stacking program, astrometric (plate) solver, fits file viewer

Brought to you by: han59

"Calibration and Alignment Only" introduces grid artifact

Forum: General discussions ASTAP

Created: 2025-05-26

Updated: 2025-06-06

1 2 > >> (Page 1 of 2)

cytan - 2025-05-26

Hi Han,
I wanted to align 2 fits files that were generated on different nights using ASTAP "Calibration and Alignment Only". In ASTAP, I did not give it flats, dark flats or darks because I only wanted an aligned image. Anyway, this file was rotated by about 180 deg to align with the reference and a periodic grid was introduced. The problem file is called "grid artifact" and the original file is called "original".

Here's the link to the fits files before and after alignment with ASTAP 2025.05.21 ARM64 bit:

https://1drv.ms/f/c/30d77764ca14119b/Ev2aGEnKn1ZJrpqHhUnmTSEBYmHlaFViWboWs8IQN081iA?e=KfYoWk

Please have a look. Or is there an option I need to enable in ASTAP to get rid of the grid pattern.

Thanks!

cytan

grid artifact.jpg

original.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Han - 2025-05-26

The lines are artifacts if there is a small angle between the two to align images. Normally it is filtered away but becuase the colour are so extreme it doesn't work. Try to first to bring a balance in the two input images by the menu below:

Cheers, Han

Untitled.png

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-05-28

Hi Han,
I'm probably doing something wrong here:

Loaded the 2 files:

M94, 2022-05-08, 26x300L, AP130GT, (IDAS_D2), ZWO ASI2600MC Pro_stacked.fits

M94, 2025-05-25, 155x30L, AP130GT, (None), ZWO ASI2600MC Pro_stacked.fits

Ran Calibration and Alignment only

Went into pixel maths and did what you instructed above.

Saved what is onscreen on ASTAP which has the grids displayed. See image capture below: grids_ASTAP

And the "corrected" saved fits file. It definitely is better, but the grid pattern is still there :( : grids_result

Here's what auto colour correction found. Image: correction

Any other thing I should do?

cytan

correction.jpg

grids_astap.jpg

grids_result.jpg
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Han - 2025-05-28

I see it in the current Mac version 2025.05.21. I made a few days ago a change in the "calibration & alignment" and do not see here any longer. I will release an updated Mac version tonight.

cs Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-05-28

Thanks!

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-05-28

I have uploaded a new version 2025.05.28. Unfortunately I do not think it is solved. It seems something specific to the Mac version. I very busy with modifying the solver but I hope to look to it tomorrow.

cs, Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-05-29

No problem. Thanks for working on this.

cytan

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-05-30

Fixed in the last version 2025.05.30. Tell me if this works for you.

Cheers, Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-05-31

I just tried 2025.05.30 version out, and still no joy :(. Attached are 3 images:
1. Pixel math: pixel math.jpg
2. Image before pixel math: without pixel math.jpg
3. After applying pixel math: with pixel math.jpg

cytan

pixel math.jpg

with pixel math.jpg

without pixel math.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-05-31

I noticed these artifacts as well. Initially I tought these where created by the display zoom factor but closer inspection reveals you can see them at any zoom factor and also in other FITS viewers. But also in the orginal stacked images prior to aligning. So the must by created by stacking or in the original images. Stacking is more likely due to the angle of the pattern. Can you share the original lights, master dark and master flat?

I'm guessing but it looks like one or more images where scaled differently. Which alignment method did you use, star alignment or astrometric alignment?

Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-05-31

Hi Han,
Alignment is with star alignment.
I don't understand your statement about seeing the artifact in the original stack. Attached is the original image before "Calibration & alignment" where I don't see the grid pattern. FYI, I aligned the two fits files in Nebulosity and they do not have the grid pattern either.

I have 165x30s exposure subframes (Note: I have no choice but to use short subframes because of my location) that generated the fit file. Unfortunately my "free" OneDrive Account won't be able to accommodate so many files. If you have a place where I can upload, I can certainly do it.

cytan

Before Calibration & alignment.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-05-31

Attached the pattern what see in the original stack when I play with the display and zoom settings. It looks like a pattern created by the display stretch algorithm. It disappears when you zoom in.

Does it disappear when you save it as PNG or JPEG and view it in an other viewer?

At the moment I do not think looking to your stack will help. I will install StarTools and have an other look.

Untitled.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-05-31

Okay I have StarTools installed. I press the AutoDev. The original image looks normal. However when I load the aligned image (one of them is aligned to the first image) , press AutoDev I see the same checker pattern under an angle of about 20 degrees. This doesn't make sense. The two image where aligned within maybe one degree. So why does the checker pattern has an angle of about 20 degrees? Secondly I do not see it in ASTAP or CCDCiel or PixInsight. No checker pattern.

Then as a test I cropped the aligned image in ASTAP. Loaded again in StarTools, pressed AutoDev. No avail, still the checker pattern.

At the moment I have the impression the StarTools AutoDev picks up something in the aligned image and creates the checker pattern. Can you ask the StarTools developer for advice?

Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-06-01

OK. I'll ask on the StarTools forum page.

Here's the image from Nebulosity after alignment without the grid pattern for comparison.

I'll keep you posted about what is found/not found.

Thanks for looking into this.

cytan

Last edit: cytan 2025-06-01

Nebulosity aligned.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-06-01

Hi Han,
Sorry to bother you again, but when I stretch the ASTAP aligned image in Nebulosity I see the green grid lines! See attached.

cytan

Nebulosity grid lines visible.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

cytan - 2025-06-01

For comparison, here is the original image (before "Calibration & alignment") after stretching the same way in Nebulosity which does not show the grid lines.

Last edit: cytan 2025-06-01

Original, same stretch.jpg

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
- han.k - 2025-06-01
  
  I do not see the grid lines in your screenshot your but I believe they are visible.
  
  StarTools must pick up the information of the aligned image. I'm sure there is some information in the image. As an explanation the following. As soon you align so rotate an image mathematically a few degrees you will notice that it will create some lines. At some places there are zero pixels and next to the zero pixels you will have two pixels. You have to correct that. What ASTAP does is it keeps a record of the number of pixels placed at each new position. When there are zero pixels it will take the nearby pixels as a replacement. When there are two pixels placed, it will divide the values by two. But there is a catch. Astronomical image have a pedestal/back ground value. If you stack several images (especially in case of Sigma Clip) you only want correct the number of pixels added by first subtracting the background. Currently ASTAP only aligns with values above the background.
  
  In Cytan image the background has a strong slope so the align algoritmh will not work properly since it works with a single background value and I assume these vague artifacts are picked up by StarTools.
  
  So that should could explains the artifacts. Only the about 20 degrees angle of the artifacts is weird.
  
  For pure the alignment of images (so not sigma clip stacking), I could consider an other algorithm. Where two pixel land, I could just divide the value by two. Maybe half the value could be used for the zero value pixels. Then after the rotation the background could be equalised. (This is not possible for sigma clip stacking).
  
  So this is an explanation and it gives also a possible solution. Note that there was never much thought on the calibration & alignment functionality. All attention was on sigma clip stacking.
  
  I will modify the current algorithm.
  
  If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
  - cytan - 2025-06-01
    
    Thanks Han for describing what ASTAP does. I think you're absolutely right that the artifacts were introduced because of sigma clip stacking. In Nebulosity, the "alignment" stacking was done with "Average" and I cannot select any other setting like "Standard Deviation Filter" setting.
    
    Anyway, I've zoomed in and marked on the the ASTAP aligned image that has been stretched in Nebulosity which I hope will clearly show the grid lines.
    
    cytan
    
    Last edit: cytan 2025-06-01
    
    zoomed_in.jpg
    
    If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Martin Dowd - 2025-06-01

As a long time user of Startools just some advice about the preparation of your data set ( or frames ) before you load , open and stretch ( AutoDev or Optidev ) your stacked and aligned image.

https://www.startools.org/links--tutorials/starting-with-a-good-dataset/general-pre-processing-dos-and-donts

I did have an issue with a checkered pattern in my early Mono LRGB data sets which was stacked in Deep Sky Stacker. As soon as I used ASTAP as my stacker the problem did not occur again and its been faultless and seamless for over 2 years now . My checker pattern was not the same as your problem though.

Martin

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
- cytan - 2025-06-01
  
  HI Martin,
  Thanks for the link. I've known about StarTools required practise ever since I've started using it. In the past I've used Nebulosity (which unfortunately has become abandonware) to align stacked images taken on different days, months or years apart and I've never seen this problem.
  FYI, this is actually the first time that I've used "Calibration and Alignment" in ASTAP. My apologies to Han for opening a whole can of worms.
  
  cytan
  
  If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-06-01

Hi Martin,

Nice long check list.

Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Martin Dowd - 2025-06-01

Startools was developed by Ivo Jager and is a unique processing engine completely different from traditional processing engines like Pi and PS . It actively tracks your data from start to finish.
Hence your stacked data set must be as raw as possible
The checklist is long but easy to achieve and will give Startools the best opportunity to provide optimum results for your images.

Apologies for giving Startools a plug in this post

Martin

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-06-01

Using the images from Cytan, in the code routine "calibration and alignment" I have replace the image values temporary by a fixed value. The routine to correct the number of pixels added was switched off. Below the result. There are no zero value, only double and quadrable values at the line crossings. The difference between the two aligned images is 0.29 degrees. So it surprising to see the angles of the lines.

But I'm confident it can be fixed.

M94, 2025-05-25, 155x30L, AP130GT, (None), ZWO ASI2600MC Pro_aligned.png

Untitled.png

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-06-01

The solution will be simple, but after a chat with AI, I think I can even make it even better using a technique called "Inverse Mapping (a.k.a. Backward Mapping)" . That could also improve the standard stacking method. :)

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

han.k - 2025-06-01

This new idea looks good. I have only prepared the MacOS executables for this moment. Could you try this one? Open the ASTAP app with the right mouse button and move the executable to
/Contents/MacOS

astap_mac_X86_64.zip (v2025.06.01)
http://www.hnsky.org/astap_mac_X86_64.zip

MacOS M1 astap_mac_M1.zip (v2025.06.01)
http://www.hnsky.org/astap_mac_M1.zip

Executable only. Move the executable in the application at /Contents/MacOS

Han

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

1 2 > >> (Page 1 of 2)

Log in to post a comment.