Incorrect orientation of the FITS file

I’ve shot a timelapse using my ZWO ASI533MC and a Rokinon 16mm f/2 lens. To make it easier for me while shooting it, I’ve applied mirroring for both axes in INDI’s settings (it messed with Bayer pattern, but easily fixable with manual override of the pattern in Siril’s settings). Gwenview (the image viewer of KDE) is easily handling the orientation (but colors are weird, which is expected, as mentioned before):


But Siril 1.0.1 is mirroring only on one axis:

If I do Image Processing → Geometry → Horizontal Mirror - orientation is finally the expected one:

After running pre-processing (no mirroring was done here) - Siril still behaves the same:

Gwenview is still fine though:

The question here is - why this is happening? Here are the links to both files:
https://fliker09.tk/sharedstuff/IMAGE_210.fits
https://fliker09.tk/sharedstuff/pp_light_00210.fit

Hello,
About the FITS file orientation I can provide you this webpage : Siril:FITS orientation - FreeAstro
It explains why your debayer is wrong too.

Thanks for the link! But tbh I don’t quite understand what to make of that info… From my understanding - all that info refers to demosaicing, which is not applicable when you just open a FITS file. I see that my RAW file does contain all 3 keywords in the header - are you trying to say that INDI didn’t write proper values when I’ve applied the mirroring to both axis (which I did to help with framing)? But how then Gwenview handles it just fine? Am I missing a piece of the puzzle?

Hi, Siril uses the pieces of information from the header to get the correct Bayer matrix and create images with the correct colors, not the correct orientation. It doesn’t matter which way the images are displayed or processed for deep-sky, and siril is a deep-sky image processing tool. We usually put them back in the correct orientation just before exporting the file after processing (it’s done automatically during plate-solving).

You can also rotate all your images using this tool and the mirrory command.

Good luck

Now this is getting even weirder then… The thing is - Siril has just one wrong axis, not both (which would be expected if the orientation was ignored altogether).
Regarding demosaicing - what’s wrong in my case? Did INDI write a wrong FITS header?
As for PCC - yeah, very handy feature with rotation correction, used it for DSO stuff. Sadly, PCC refused to solve the image, after hours and hours of my frustration :frowning:

I guess that resulted in wrong pattern modifier keywords, or maybe a wrong pattern, if you had to override the value something is wrong indeed