Sigma stacking of images before/after meridian flip

Image processing, astrometry, photometry, etc.
Post Reply
Stefan_Lilge
Posts: 7
Joined: 01 Dec 2019, 13:30

Sigma stacking of images before/after meridian flip

Post by Stefan_Lilge » 28 Aug 2020, 19:29

Hallo,

I have never been able to successfully stack images that are rotated by 180 degrees (images before and after meridian flip) with sigma stacking in AstroArt7. I _think_ that it worked in AstroArt6.
I can get preprocessing to register the images by aligning with "star pattern + rotation" and setting maximum rotation to 180 degrees. But sigma stacking still seems to "see" a 180 degrees rotated image, which means that signal/noise ist awful, probably because most pixels are rejected. That does not happen if I "average" the images.
So if I have 10 images before meridian flip an 10 after, a stack of the first 10 images will have a much better S/N ratio than stacking all 20 images (if sigma is used).

Is this a bug or a feature? Or a matter of getting the settings right? As I mentioned I am pretty sure that this problem was not present in AstroArt6.

Thanks and greetings,
Stefan

fabdev
Posts: 461
Joined: 03 Dec 2018, 21:43

Re: Sigma stacking of images before/after meridian flip

Post by fabdev » 28 Aug 2020, 21:42

Hello,
Sigma (or Average / Sum / Median) is always performed after registration, so if there's a problem with rotation you will see it with or without Sigma.
As a test, you could enable one time the option "Create Log file", or "Confirm each image" or even "Save each image" to verify that all images rotated 180 degrees were processed correctly.
If possible, post a jpeg with Average and one with Sigma.
Greetings,
Fabio.

Stefan_Lilge
Posts: 7
Joined: 01 Dec 2019, 13:30

Re: Sigma stacking of images before/after meridian flip

Post by Stefan_Lilge » 30 Aug 2020, 21:42

Hello Fabio,

for alignement I use the settings from the first attached image ("automatic" does not work as I get an error message that AstroArt cannot register the 180 degrees rotated images). Sigma is set to 3.0.
The second image shows a stack of 18 images, of which 9 are taken before meridian flip and 9 after the flip (in a different night). Average is on the left, Sigma on the right. Everything else was left unchanged.
Average is clearly better than Sigma. The camera is an ASI2600MC, but I have had the same problem with other cameras (also mono chips) in AstroArt 7 (not noticed in AstroArt 6).
Actually each of the groups of 9 pictures (before/after flip) sigma stacked has much better S/N than all 18 sigma stacked.
Attachments
Settings.jpg
TestAverageSigma.jpg

fabdev
Posts: 461
Joined: 03 Dec 2018, 21:43

Re: Sigma stacking of images before/after meridian flip

Post by fabdev » 31 Aug 2020, 01:28

Hello Stefan,
it seems to me that the result of Sigma is just a single image (maybe it's just the first image) and Preprocessing was aborted during Sigma.

Are you using the 64 bit version of Astroart?

Rotation + Sigma of 18 big color images (20+ megapixels) may require up to 10 GBytes of extra disk space, so verify that there's enough disk space for the "Temporary folder" of Preprocessing.

A further test could be selecting "Binning 2x2" in Demosaic: if it works, it's just shortage of resources, which was not notified.

Stefan_Lilge
Posts: 7
Joined: 01 Dec 2019, 13:30

Re: Sigma stacking of images before/after meridian flip

Post by Stefan_Lilge » 31 Aug 2020, 23:44

Hello Fabio,

I have tried the 64 bit version of AA (so far I have used the LAA version) on my computer with 16GB RAM and results were the same.
The SSD where the temp path points to has 95GB free.

As a workaround I have split preprocessing into two steps, first step was dark/flat/debayering und registration and saving each image instead of averaging/sigma stacking. This gave me all images in the same orientation. In a second step I averaged and sigma stacked these registered images and this time sigma was as good as average.
What amazed me was that even the average version in the split preprocessing was better than the average version of the one-step preprocessing. I may have wasted some S/N with some of my recent images as some of them included some flipped frames after meridian.

In case you would be interested to look at the date I have uploaded a ZIP file (600MB), it contains the raw lights and a master dark and master flat (flat darks already subtracted). The dark temperature is lower than the temperature of the lights, but that shouldn't matter as the camera doesn't need darks, a bias would be sufficient.

http://ccd-astronomy.de/Teilen/M16Data.zip

Yours, Stefan

fabdev
Posts: 461
Joined: 03 Dec 2018, 21:43

Re: Sigma stacking of images before/after meridian flip

Post by fabdev » 01 Sep 2020, 01:18

Hello Stefan,
I downloaded your sequence. It seems to me that the problem is triggered by the option "Copy missing borders". Disabling it your files are processed correctly with Sigma.

The function "Copy missing borders" is a cosmetic function which tries to rebuild the "darker borders" caused by the shift between each image.

For some reason, (probably rotation of field) it ruins the processing. I think that it affects both Average and Sigma, but on Sigma is more evident. I will investigate more in the next days.

Let me know if this is solves,
Thanks, Fabio.

Stefan_Lilge
Posts: 7
Joined: 01 Dec 2019, 13:30

Re: Sigma stacking of images before/after meridian flip

Post by Stefan_Lilge » 01 Sep 2020, 10:52

Hello Fabio,

unchecking "Copy missing borders" did the trick. Now Sigma looks as good as Average (and both look good).
Who would have thought that such a "small" option could be the cause for this.

Thanks a lot,
Stefan

Post Reply