"Floyd L. Davidson" wrote in message
...
"Peter" wrote:
"Floyd L. Davidson" wrote in message
...

Nikon's original D1 is a 2.7 MP camera, but the sensor
has twice that many sensor locations. Each pixel from
the D1 is made up of two sensor locations.

With an array of 120 sensor locations it would be
interesting to see the noise characteristics of a 12 MP
image that used 9 sensors for each pixel.

Even more interesting would be if instead of a Bayer
filter the sensor had first a lens that would defocus
light such that each group of 9 sensors would all get
the same light, and then use a color filter (similar to
the Bayer filter) that contained 4 each green, 2 each
red and 2 each blue, with the center being a clear
filter. Bingo, no Bayer interpolation, and instead
there is a direct RGB output (with an additional
independant luminance channel).

It may or may not result in improved SNR, but color
rendition should be significantly better than a 12MP
Bayer image.


Why? Asking seriously. I don't understand the logic.

With Bayer filtering the color of a single pixel depends
on the colors of all adjacent pixels. One result is that
an image produced with a Bayer system cannot have an
abrupt transition of, say, blue to red in the distance
of less than 5 or 6 pixels. If each pixel color is
determined by 3 or more sensors unique to that pixel it
would be possible to have a transition theoretically in
a distance of 1 pixel (adjacent pixels that are pure R,
G, and B). In practice it wouldn't be that sharply
defined because the filters would not be perfect, but it
should be possible to double the lines per unit of distance
that can be resolved as different colors.



Thanks, interesting idea.

I hope this doesn't start a war, but wouldn't a Foveon sensor eliminate the
problem.
I guess cost is one reason and I suspect a Foveon would create other
problems.

--
Peter