Chromatic aberation free demosiacing.

[CharlesRamsey]

With a Bayer pattern it is possible to select 4 pixels bin them and create an image free of chromatic aberation caused by demosiacing algorithims. This is all obvious but I'm surprised all cameras don't have this option. However for any given pixel it is possible to bin it in 4 different ways. You can call these top left, top right, bottom left and bottom right. This gives 4 chromatic aberation free images offset by 1 pixel. Then it is possible to use these images to interpolate an image with the same resolution of the origional pixel count. This of course has it's advantages and digadvantages.
Resolution would be smeared over an area equivalent of a 9 pixel binning. It's just a matter of taste do you want the chromatic aberation ot the reduced resolution?

[hjulenissen]

With a Bayer pattern it is possible to select 4 pixels bin them and create an image free of chromatic aberation caused by demosiacing algorithims. This is all obvious but I'm surprised all cameras don't have this option.

One Bayer cell can e.g. be:
[r g1
g2 b]

If you generate one full-color (co-sited) rgb triplet as :
ro = r
go = mean(g1, g2)
bo = b

Then you get a 1/2 x 1/2 resolution image where the red color plane is shifted slightly towards upper left, and the blue plane is shifted slightly towards lower right.

I think it makes more sense (for a simple algorithm) to interpolate/filter each plane separately (so as to generate full-resolution color planes). This still would not generate the highest possible luminance resolution, and would do nothing to reduce the impact of aliasing at the recording stage.

-h

[Bart_van_der_Wolf]

With a Bayer pattern it is possible to select 4 pixels bin them and create an image free of chromatic aberation caused by demosiacing algorithims.

Hi,

Chromatic aberration is caused by the lens, not by the demosaicing. So your assumption is wrong, and therefore the 'solution' is also wrong. Maybe you are referring to false color artifacts caused by different sampling densities between Green and Red/Blue, but that requires a different solution than simple binning.

Cheers,
Bart