2023-09-28: APP 2.0.0-beta24 has been released !
Improved application startup, fixed application startup issues, upgraded development platform to Oracle GraalVM JDK21
We are very close now to releasing APP 2.0.0 stable with a complete printable manual...
Geometric shapes in integration results
There is probably a simple explanation for this, but I apologize that I couldn't find it by searching the forums.
Last night I shot Supernova SN2023ixf in M101. The previous night I shot 3 subs each of R, G and B and it worked fine. This time, when I weeded out the subs with poor quality scores, I ended up with some strange geometric shapes around the image -- almost like a picture frame bevel.
That was with 4 subs. However, if I add back in the lower quality subs (low SNR, clouds, etc) I get a better result without this geometric artifact:
I suspect this has to do with one of the processing algorithms and the number of subs, but I can't explain why this works for 3 subs, but not 4 or 5 subs (I didn't try more than that). Any ideas?
That looks really weird, can you share the fits headers with me? The info is shown in the metadata/header window to the left of the image viewer 😉
This has been reproducible with other targets as well. It never happens when I stack lots of frames. It seems to be worse with small numbers, like:
2 frames - strong effect
3 frames - no effect
4 frames - effect
5 frames - no effect
6 frames - less effect
7 frames - no effect
8 frames - no effect
9 frames - no effect
10 frames - no effect
Here is what two stacked frames look like -- the effect is even stronger:
I have attached the FITS headers from the stacked result as well as the 2 individual subs that went into this.
Also, I am adding dynamic distortion correction to 4/registration, changing mode to advanced in 5/normalization and enabling 1st Degree LNC/3 tier and 5% MBB in 6/integration. Those are the only changes. Master flats, master dark flats, master darks and BPM used.
Note: This is still occurring as of Beta 23. It's not a killer, given it's rare to stack so few frames, but that's all I could get out of the supernova in NGC 1097 last night.
@mabula-admin I have uploaded the subs I used, the integrated results, and the RGB-combined result to /AstroPixelProcessorUpload/MountainAir-geometricartifact on FileStation.
As I mentioned, this is likely lower priority. In both these cases I tried to sneak in targets between other targets and ended up with a small number of subs -- in this case a ridiculous two for B and G and only one for R.
If this isn't worth fixing (it seems to go away entirely when you get to 7 subs per integration or more), perhaps just pop up a warning to users who try to generate integrations from a small, even number of integrated frames.
Hello all. I have put my integration in the folder in File Station. If you look at the header of the fit file, you can see my integration settings. The artifacts come from the Automatic mode in Tab 6. With so few images, Median is selected as the Rejection method and that creates the Geometric artifacts. You have to select Manuel Average in Integrate Tab 6, then everything works as expected. The hotpixels can be removed with Cosmetic Correction, at least most of them.
Hope this helps.
Thank you very much for reporting this and your testing. It is something that definitely should not occur for sure, so I will test this and fix it 😉
I guess the combination here of both median and MBB causes this. Or does it also happen with MBB turned off? ( I will test myself of course 😉
Hi Mabula, I can confirm that when MBB is deactivated the edges look fine.
With best regards.