2022-05-11: APP 2.0.0-beta 1 has been released !!!
Download links per platform:
[Solved] Should OSC users use dithering to reduce debayering noise?
Debayering OSC images introduces noise because it uses interpolation instead of the true value. This noise is fixed if the image is not moved in the frame so stacking won't average it out. My questions are specific to the APP debayering algorithms (but feel free to chip in with others).
Will dithering reduce this noise? If so, what is the optimum number of pixels to dither each frame by? Presumably we should dither by an odd number of pixels right? Do these answers change if we don't use the Adaptive Airy Disk option?
Thank you for your excellent question 😉
For sure, dithering will reduce the debayering artefacts very nicely. As it would all other noise factors as well.
Debayering means demosaicing a Bayer Color Filter Array (CFA) pattern of 2x2 pixels (eg. RGGB), so we can simply assume that we need to get rid of a 2x2 pattern. Now from sampling theorem you would greatly reduce the 2x2 pattern by sampling this 2x. So that would mean to dither in a range of 4 pixels to get quickly rid of the demosaicing artefacts 😉
Since there are, no doubt, other noise patterns on your sensor, there is not an optimale dither range. Simply put, the bigger dither step that you can take, the better. Overall noise will drop faster with integrating frames with bigger dither steps until you reach the theoretical limit for noise reduction.
In my experience, it is very good and normal to dither with dither steps of 10-20 pixels on most sensors. If you would only dither a couple of pixels, noise reduction will not be that optimal 😉
This answer will not change if you use another algortihm than Adaptive Airy Disc.
Dithering odd or even amount of pixels does not matter, the dither algorithm randomizes the size and direction of the dither steps.