Hello. New to siril and these forums. Currently using siril to average terrestrial images for noise reduction. Wondering why the converstion of cr2>Fits and then exported to Tiff generates output images that are 2-3 stops darker than the original file.
Could it be that when converting to TIFF from FITS, you are not properly adding an ICC profile that marks the TIFF as being linear data as opposed to an sRGB TRC (which is assumed if an ICC profile is missing)
The ICC profile has been implemented in TIFF export in the last version only, make sure you have it. @lock042 is it even in the 0.99.6 or was it after the release?
It also depends how you view the original, the FITS and the TIFF and at what stage you involve demosaicing.
This is because siril FITS image are converted into linear data.
This is perfectly normal and fine to have darker image.
Is FITS linear different from a traditional Linear converstion ? (Using Raw therapee/Photoshop/Capture One). I know that linear conversions typically get a little darker and obviously more flat in appearance but this is exceeding that by much more.
Again, it depends on whether or not the TIFF has a proper ICC profile that indicates it is linear data, or if the ICC profile is missing.
Linear TIFF data that is missing the ICC profile and as a result misinterpreted as being aSRGB is not ever “a little dark” - it’s always extremely dark.
RawTherapee will always embed an ICC profile that matches your output profile. The exception being if you use the “output reference image” function - see https://github.com/Beep6581/RawTherapee/issues/5575
Using imagemagick to convert a FITS into TIFF will ALWAYS result in a TIFF that is missing the ICC profile unless you explicitly add the ICC profile with additional commandline options.
Based on @vinvin’s post, it sounds like older versions of siril will always be missing the ICC profile on export also.
And new version save in sRGB (as we generally use the TIFF export after histogram stretch). Dev version ask the user if he wants sRGB or no profile.