Graeme Gill wrote:
Milton
Taylor wrote:
Question for you: what energy levels are
required to generate the colors that are at the extremes of this
chart? Are they dangerous? If so, then I'm not sure this chart is all
that useful...it would need to be bounded by a safe/reasonable energy
limiting shape.
The energy is approximately 1/CMF value, and since CMF's
go to zero out of visible wavelengths, the energy has to
go to infinity. I stopped at 380nm and at which the Y sensitivity
is 0.000039097450, so on the order of 26000 times the light
level at our most sensitive wavelength (555nm). At 380nm,
this may give you a really good sun tan.
Thought so!
Most
color work simply assumes that it all looks
the same, as long as it's photopic, and doesn't assume a maximum.
Only some color appearance models (ie. Hunts model) and safety
folks worry about absolute light levels once it's out of
the scotopic region.
On the other hand if it's all safe...then
does it not answer the original question about what colors the eye can
see? (Also, how do you know this is what the eye can see? i.e. what is
the basis behind this shape?)
It's all in the Colorimetry theory. Basically a monochromic source
must be able to generate the most difference in cone responses, hence
most saturated colors (or two sources along the purple boundary).
So where does that leave us? What does that gamut chart you just
plotted actually mean? That we can see virtually any emitted colour if
it's strong enough?
What happens if you redrew the gamut to also limit the maximum energy
at any hue to something like the maximum typical brightness of an LCD?
(Say 250 cd/m2) knowing full well that this would only happen in
practice if you did something like discussed below?
Now that's a very interesting suggestion.
Suppose you leave the panel's backlight at 250 cd/m2, why could we not
get the profiler to take advantage of this fact, still scaling white to
125, but allowing the other colors that the eye is less sensitive to,
to go brighter, to the max of 250? Or am I missing something here?
It should be practical to create a LUT based profile that does this
sort
of thing. The main problem is that the windowing system will sabotage
the effort. It will send direct RGB values to the screen via the RAMDAC
tables (for window decoration) making sure you're adapted to the 250
cd/m2
white point. You'd need a full screen application that used the profile
to
be able to exploit this approach.
(I'm not sure any of the OS/windowing systems actually properly send
everything through a display profile, nor is there a way of stopping
some applications bypassing such a default behaviour.)
If you went with this approach, wouldn't the more extreme colours look
brighter than white? Wouldn't that be perceptually rather confusing? Or
would this idea have to be used with some subtlety so that the
perceptual effect was preserved?
As for the technicalities, there are two issues there...firstly as you
would probably know Windows doesn't use profiles for its own rendering,
only it's video LUTs, so you'd have to also fiddle with those too in
some similar fashion. This would have to mean that White=256,256,256
input, but output is 128,128,128. Other colours could possibly
generate individual channel output values > 128.
Secondly, re the windows fullscreen thing, I'm not worried about that
because I'm assuming you'd be previewing full screen anyway, like in
Photoshop's full screen mode. Which also does use the profile
to adjust what you're seeing.
Actually, there is considerable merit in this idea from another point
of view. Most LCD monitors are too bright for color matching work. So
halving the typical brightness for white point would be a great idea.
The only real practical problem I see is the 8bit limitation.
...Milt
Graeme
Gill.
--
Milton
Taylor
Director
![]()
Level 1, 530 Lt. Collins St, Melbourne, Vic 3000
GPO Box 1278, Melbourne, Vic 3001
Direct: +61 (0)3 9326 6621
Mobile: +61 (0)418 172 511
Main: +61 (0)3 9326 6666
Email: milton.taylor@xxxxxxxxxxxxxx
|
