Thanks Bart for the in-depth and clear explanation.
Concerning the -somewhat 'softened'- resulting image, this is very much visible when one compares the same image taken in 4-shot and in 16-shot modes.
Hi Thierry,
You're welcome. Many moons ago when I was immersed in film scanning issues, I made the following
animated explanation to explain the benefits of staggered sensors (6 scan lines, 3 of which offset by half a sensel) as they were introduced by Epson. Maybe it'll help to further illustrate the principle, but then applied in multiple directions.
As for the luminance resolution improving a bit less than the chromaticity one, I guess this is due to way the information is sampled: the 2-times more GREEN pixels provide the "Luminance" information to/for the neighbor RED and BLUE pixels while each R, G and B pixel provides its own "Chrominance" information.
Is that correct Bart?
It's part of the explanation. Not only is the Green (close to Luminance) information sampled at twice as many locations as Red or Blue, but also the demosaicing makes very good use of the (lower levels) of luminance being contributed by the Red and Blue sensels. So in a way, at least some luminance is always recorded at
all sampling positions. Only when there are color differences but little or no luminance differences, then the Lumnance (most important for human vision) resolution will drop (but those are rare cases). That's why Bayer CFA demosaicing works so well.
Therefore color resolution will benefit more than Luminance resolution from microstepping due to the higher sampling density of all colors. Overall resolution will benefit from the additional in-between samples.
Cheers,
Bart