Have a look at the reviews for
various image sharpening tools, and how they struggle to deal with the
problems that arise when an "edge" (such as the line between a watch's
black hands and the watch's white face) falls in the center of a pixel.
Most of us can certainly see the difference between
RGB(255,255,255) and RGB(255,255,254). Those differences cause
"banding". Look at The Luminous Landscape's excellent review of Kodak's
monochrome professional camera (the 760m, if I recall). 12-bit color is
better than 8-bit, but it's not enough. Neither is 16-bit color.
There's a difference between "precision" and
equations. An equation can define the number "pi", but no amount of
decmial precision can define it accurately.
JPEGs look blocky because their cosine formulae
cannot capture the rich subtleties of the real world. Fractal images
formats are somewhat better, because the equations they use to compress
data can represent more of the chaos detail inherent in reality.
DPReview uses photographs of a watch face to
illustrate advances in photographic sensors. Look closely, and the loss
of information due to "precision" is always apparent. Find a real watch
face, and study the edge of one of the hands. Examine with a magnifying
glass the shadow that slender hand casts on the face. No matter how
closely you look, you will not see pixels, or banding, or differences
in shades of color. Put the same scene under ever increasing powers of
magnification, and you will unveil ever richer amounts of detail.
This is not to say that computers will never capture
and render reality. Far from it. Ray Kurzweil writes, in his book "The
Singularity is Near", that computers may become smarter than humans,
and in fact may completely model and emulate the human mind, in the
next 30 years.
http://www.kurzweilai.net/articles/art0134.html?printable=1 But virtual reality won't be achieved through
precision. It will be reached by understanding and interpreting the
"rules" of how reality appears to our mind through our senses, and
modelling that ever more closely.
You can recognize a friend from a single line
describing her profile drawn by a sketch artist. MIT has demonstrated
we can recognize age, gender, mood, and even individuals, if we are
shown only 13 points on that individual's body in motion as they walk
(dots at the shoulders, elbows, hands, hips, knees, and feet, and one
more for the head).
The endeavour to capture and convey reality with an
imperfect abstract sounds to me like "art". I hope this art is Nikon's
goal, not some quixotic quest for megapixel precision.
** Originally posted at DPReview.com at 7:37 PM, Sunday, October 16, 2005 (GMT-5)