On Thu, 18 Jan 2001 07:48:10 -0800 shAf ([EMAIL PROTECTED]) wrote:
> Altho I agree 8bits is enough (barely) to provide human
> discrimination, I don't agree with any of the 16M colors being beyond
> human discrimination. Perhaps you can point to an example(?) I'm
> actually saying all of the 16M colors do not exist, in human
> perception or physical reality.
What I was saying is that we lack the fineness of discrimination to be able to
separate
16.4m colours, even in a large gamut space. We can probably only manage about a tenth
of
that number. Our sensitivity to small colour variations is very easily tested by
making
some reference colour patches and comparing them on screen. EG there has been a
comparison
of 2 L126 grey patches on my site for yonks, which differ by B5, and you can easily
see the
colour difference but it gets much less certain below about B2 or B3 difference (~1%).
I simply don't understand this esoteric point that 'all of the 16M colours don't
exist'.
You may need a colorimeter, and/or you may need a better output device, but unless
they
somehow fall outside the gamut they are supposed to map to, I can't see how they
inhabit
some sort of Heisenbergian state. It appears to be like saying there are imaginary
blues
which are R0 G0 B327 in an 8 bit scheme.
> My "perfect" camera/film/scanner" analogy stated the point I wanted
> to make regarding RGB=0-0-255. Another would be a L*a*b conversion.
> That is, if you create an image of false colors ... 360 degree hue
> left-to-right and brightness vertically ... you'll have a
> representation of ALL (almost) RGB values. If you accept that Lab is
> the best model we have for human perception and the colors available
> in nature, then if you convert from RGB=>Lab and then convert back
> (highbits please), you'll end up with an image which looks a lot like
> the original, but if you diff the original with the converted, you'll
> realize the number of pixels which fell outside Lab space. It is damn
> near half of the image!
Since vanilla RGB is completely device dependent, you seem to be saying that the
monitor's
output gamut is a wider space than LAB? I wouldn't be surprised by this, and it would
seem
a valuable attribute. A monitor which only showed us a fraction of the image colours
would
be a pig to work with (try editing in 16 or 256colours:). Most of the colours in a
high-bit
file are 'imaginary' as they inhabit a wider range of values than the monitor is
capable
of, but this is a virtue we exploit when manipulating the image - the manipulation
brings
colours within a gamut that is useful and visible to us. So what?
> Of course anyone can point to some problems associated with my
> experiment ... conversion artifacts, the CM engine and its rendering
> intent ... but I've played with combinations, and all imply a similar
> number of RGB values which are outside Lab.
RGB values in what colour space? And more importantly, viewed via what device? You
seem to
be saying that if, say, you scan a film to RGB values then promptly output it onto
film,
using a properly colour-managed workflow, the image will include imaginary colours
which
were not in the original and which we cannot perceive. This appears, uh,
counterintuitive
to me.
Regards
Tony Sleep
http://www.halftone.co.uk - Online portfolio & exhibit; + film scanner info &
comparisons
