I am having trouble in understanding how Siril is handling 16-bits and 32-bits data. For instance, I feel like even when pre-processing a sequence of images that are in 16-bits, the software still allow to save the final result as a 32-bits file, which does not make sense to me… can you go from 16-bits to 32-bits without needing some interpolation (which can degrade the data) ?
Could someone explain which steps/parts of Siril are in 32-bits, those that are related to the specific choice of 16-bits or 32-bits (the panel in the top bar of the software), and those that are fixed based on the data being processed ?
In the end I would like to know that if I want to work in 32-bits and obtain the processed data in 32-bits, I did setup everything well beforehand in Siril and that no kind of interpolation/approximation will be done at the final export step.
I’ll still answer, there is no interpolation from 16 to 32, there is however precision loss if you go the other way around. There is an interpolation in Siril’s usual scripts, done on image rotation. It can be disabled when using Siril manually or by passing the -norot option to the register command.
@vinvin Thanks a lot, this is indeed the piece of documentation I waited for @lock042 Thank you for your answer, I did not get that the only operation was going from integers to float, and thought there was something else going on. So, makes sense now!
