On Jun 1, 2005, at 3:38 AM, Markus Hitzler, Color Solutions wrote:
Chris, if we talk about the gamut of this color spaces
I'm
absolutely with
you.
But the main difference is not the gamut, it's the distribution of
the RGB's
inside the Gamut or in other words: the encoding efficiency.
I don't question the superiority of L* based tone response over a 2.2
image gamma, in particular for 8-bits/channel images. But I stand by
my original statement that that most people will have no idea what
source space was used for a given image. The difference this will
cause in output to a printing press of all things (which are very
forgiving when it comes to loss of levels), I think is rather small.
Those with a keen eye, and are very quality conscious very well may
notice a difference. I
Which leads me to wonder if the ECI will at some point recommend
Lstar-RGB over ECI-RGB? Given the choice, I'd rather work with L*
based tone response over that of image gamma 2.2, and certainly over
image gamma 1.8 when it comes to 8-bits/channel. You can make a case
for invertability with an image gamma of 1.8, but gamma 2.2 is a
closer (while perhaps crude) approximation of L* based tone response.
I think most people are not doing a lot of conversions from CMYK into
RGB, as compared to the reverse. Just a thought.
One little example:
Let's look to the dark tones at the grey axis of Adobe-RGB and
LStar-RGB.
Because both profiles are matrix systems and grey is build by the same
values of red, green and blue the examination of the grey axis
shows the
situation of the 3D-distribution too.
With Adobe-RGB you need 20 points (0-19) to describe the luminance
range of
L*=0 to L*=3. In other words: you use nearly 10% of your
coordinates to
encode only 3% of your color space. Moreover: very often you don't
have
image information at this luminance range but the noise of your
imaging
system.
Therefore you loose coordinates for the exact encoding of the other
areas
with image information.
No question there is a lot of inefficiency at the very low end.
However, it rapidly rectifies itself by distributing more bits in
shadows and 3/4 tones than a 1.8 image gamma space does. And the
human visual system is more sensitive to shadow detail loss than
highlight detail loss.
One day in the not too distant future we will always edit images in a
massive 32-bits/channel linear gamma color space, and we'll have
intelligent and adaptive gamut mapping algorithms so that we won't
have to have these kinds of discussions anymore. On one hand I find
it a very interesting conversation (from a color geek perspective),
but from a user perspective I think it's a really irritating concept
and I can't wait to see it go away.
Chris Murphy
Color Remedies (TM)
www.colorremedies.com/realworldcolor
-------------------------------------------------------------
Co-author "Real World Color Management, 2nd Edition"
Published by PeachPit Press (ISBN 0-321-26722-2)