Re: Colormatch vs Adobe 98 - Don to Bruce part 2 of 2
Re: Colormatch vs Adobe 98 - Don to Bruce part 2 of 2
- Subject: Re: Colormatch vs Adobe 98 - Don to Bruce part 2 of 2
- From: Don Hutcheson <email@hidden>
- Date: Thu, 31 Jan 2002 19:38:24 -0500
(Continued from part one due to 12 k post limit)
>
> 6. All CIE spaces by default assume D50, in fact 'CIELab' should really be
>
> written 'CIELab(D50)'. CIELab(D65) is computed quite differently and there
>
> is no reliable translation between the two without going back to the
>
> original multi-spectra data, which is thrown away in most profiles.
>
>
>
> 7. Converting D65 samples or profiles to D50 uses a processes (typically a
>
> Bradford transform) which only works properly if all light sources used for
>
> capture and viewing have a full D50 spectrum with 100 CRI (Color rendering
>
> Index.) This is virtually impossible. The transform cannot solve metamerism
>
> errors or work properly for low-CRI sources. For that we would need profiles
>
> containing multi-spectral samples, not just the three L, a, b, variables.
>
>
While I agree with the stated premisses, the conclusions drawn are in
>
direct conflict with my experience, and put far too much weight on
>
dragging CIE colorimetry into an area for which it was never designed
>
and is probably ill-suited.
>
>
CIE colorimetry is based entirely on comparison of reflective
>
samples. It was never designed to handle emissive samples. And one
>
major difference between the way we view reflective and emissive is
>
that in the former, our eyes play the well-known trick usually known
>
as "discounting the illuminant." With emissive sources, we cannot do
>
so, because the image is self-luminous.
I'm a little confused. This is the first I've heard that CIE spaces apply to
reflective but not emissive samples. They certainly work for transmissive,
so why not emissive?
My experience is that, although no instrument seems to see a monitor's white
point 'correctly' (i.e. as my eye does,) once I nudge the profiling process
over that hurdle by eye, the resulting profile (assuming good software) has
a very high in-gamut correlation to a reflective original viewed alongside
the monitor.
The original experiments that led to the CIE spaces were emissive, not
reflective. Test subjects were asked to adjust the intensity of three
fixed-spectra narrow-band wavelengths (R, G & B) to match a series of
narrow-band variable-wavelength samples. The samples and the mixtures were
projected from behind onto a frosted glass screen, split so that the two
color fields were immediately adjoining. When they matched, the experimenter
recorded the relative intensities of the fixed RGB sources, which became
known as the color matching functions and were later transformed into CIEXYZ
and other spaces.
As far as I know all CIE color spaces are therefore based on the matching of
emissive (although obviously not CRT) samples.
Did I miss your point here?
>
>
You may be able to get a monitor and a light box to both produce
>
tristumulus values that add up to D50, but you can't get them to
>
produce remotely the same spectra. I firmly believe that our
>
perception detects that spectral difference at some level, in ways
>
that CIE colorimetry, by design, does not address.
I have long held the same suspicion. But if so, there's much work to be done
to improve the CIE's 'standard observer' model. And although colorimeters
may have a problem, a good spectrophotometer should theoretically not
exhibit the same fault, so long as it reads well into IR and UV and has
enough spectral resolution to accurately track the sharp spikes and troughs
of each source.
My gut tells me we are both missing something here. It may be as simple as
UV or IR, or perhaps it's the spikey spectral curve in the D50 standard -
not at all the smooth locus you'd expect of a Planckian 5000K source.
I guess I want to believe the CIE color spaces and standard observer data
are not at fault.
>
>
<snip>
>
Recalibrating the monitor to match each paper stock
>
is a recipe for madness....
That's not what I advocate. I just tweak the white point once. Here's my
abbreviated method.
(a) Put a stock for which you have a good profile in the booth
(b) Open a white CMYK window and assign that profile to it
(c) View the window with 'Simulate Paper White' turned on
(d) Tweak monitor RGB and view-booth brightness simultaneously till the
screen matches the stock
The hardware RGB settings that match the monitor to the first stock should
remain valid for any number of other stocks too, because the color of the
stock is being applied in Photoshop with 'Simulate Paper White.' Of course
this requires that you always view with Paper White on and you don't have
any nasty fluorescent errors in the profiles. If you do, you'll have to
tweak the errant profile(s) white points or tweak the hardware for that
stock only.
As a side note, (new Photoshop 'feature'), I wish Photoshop had a 'Simulate
Paper Color' function which would match the a* and b* but not attempt to
match L*. The current Simulate Paper White function just unnecessarily
darkens the screen. If this new option was available I would argue that it
should be a fixed default, as it's the only logical way to soft proof
anything accurately.
>
>
<snip>
>
>
My experience is that you'll have a much easier time creating
>
original-to-screen matches and screen-to-print matches if you
>
>
a) calibrate the monitor to D65 white, and
>
>
b) move the viewing booth so that it isn't in the same field of view
>
as the monitor.
>
>
Short-term memory is very reliable on color matching. Forcing you to
>
look from the screen to the sample allows your built-in
>
white-balancing algorithms, which are better than any man-made ones,
>
to adjust for any small discrepancies between the two.
Ah. Here's where we really differ. You wisely sweep the problem under the
carpet while I stupidly go on tilting at windmills, seeking the lazy man's
dream of a true monitor to proof match.
Your approach essentially says "We cant get the two to match so let's not
show them together." My approach says "I don't know why the stupid machine
doesn't see things my way, but we're not here to please science, we're here
to please the eye."
While I'm the first to admit that soft proofing is far from perfect in many
areas, if you follow my methods precisely with a good CRT or LCD monitor AND
properly adjusted local lighting, AND a good printer profile, AND a monitor
profile built with ICC Display set to 'Color Temperature OTHER', you should
find that the monitor really does match the proof amazingly closely (within
the monitor's gamut and contrast limits) even when you look at them both
simultaneously in the SAME visual field.
I know it's a little extra work up front, but it's worth it if it means you
truly can start to trust the monitor as a 'soft proof'.
>
>
None of this is to say that I have anything other than the utmost
>
respect and affection for my antipodean colleague.
Et Tu, Brute.
>
This is simply not
>
an area where we have definitive answers. Do what works for you.
Good advice.
Don
*************************************
Don Hutcheson
Hutcheson Consulting
(Color Management Solutions)
Phone: (908) 689 7403
Mobile: (908) 500 0341
email@hidden
*************************************