Re: MS Color Control Applet
Re: MS Color Control Applet
- Subject: Re: MS Color Control Applet
- From: Graeme Gill <email@hidden>
- Date: Wed, 05 Jul 2006 13:56:59 +1000
Chris Murphy wrote:
On Jul 1, 2006, at 12:53 AM, Graeme Gill wrote:
It's more complicated that that. All (useful) V2 profiles adapt the
monitor white point
to be D50 already. What's changed in V4 is that now even absolute
colorimetic intent
in display profiles have to adapt the display white point to D50,
because the ICC
have (wrongly IMHO) declared that an emissive devices white is to be
regarded
as the illuminant, rather than the devices media white point.
The idea is to prevent chromatic adaptation from occurring twice.
I'm unaware of any such problem with V2 profiles.
> When
the human has fully chromatically adapted to display white, and the
assumption on the part of the display profile is that the human has not
fully chromatically adapted, the profile building does some chromatic
adaptation when they should do none and leave any chromatic adaptation
to the CMS.
Would you care to explain this in some more detail ? While a CMS could
do anything, I'm not aware of any that make the assumption that the
viewer isn't fully chromatically adapted to the white point of the display
for Relative Colorimetric (ie. the PCS) Intent.
In an ICC v2 world, the end user's relative chromatic adaptation is not
defined and each vendor would assume something different. So they
fudged their profiles in a manner that meant you really weren't getting
actual measurement data in them anyway and it wasn't reversible.
Right, but this has no direct connection with the change to the
interpretation of the white point tag. The above issue could
have been dealt with by supplementing the white point tag with
a tag that describes the chromatic adaptation actually used,
and the problem is solved (ie. like the 'chad' tag). It would have
simultaneously solved the problem for reflection profiles where
the chromatic adaptation prescribed is the "wrong Von-Kries" transform.
CMS's would then have been free to (officially) use a better chromatic
transform for this.
In an ICC v4 world, they are chromatically adapted to D50 so that by
default there isn't going to be double chromatic adaptation.
You will have to explain this. I haven't come across any problem
with "double chromatic adaptation".
> But the
CHAD tag being required if the media white is NOT D50 means that we can
reverse the adaptation to actually get measurement data and adapt as
needed - anywhere from none to full adaptation. The results are
actually better, both when converting from display profile to output
device as well as output device to display.
But the V4 'chad' tag is for recording the chromatic adaptation used
to transform the illuminant, not the media white. See page
30, Section 9.2.11 chromaticAdaptationTag in the V4.2 Spec.
For reflection prints there is a distinction between three whites:
1) The illuminant
2) The media illuminated by the illuminant
3) The PCS Relative Colorimetric White (D50)
The 'wtpt' white point tag defines 2) under the standard D50 illuminant.
If the measurement illuminant wasn't actually D50, then it is made
as if it was using a chromatic transform, and this is recorded in
the 'chad' tag. ie. the 'chad' tag defines how 1) is transformed to D50.
For reflective media this is all clear. The transformation of 2)
under D50 into 3) is done using a "wrong Von-Kries", which I
think is a mistake, but at least it is clear what's going on.
What was unclear under V2 was how to handle emissive media.
Now what happened for many profiles (ie. sRGB) is that
they took a (what I think) is sensible approach, and said
"For reflective media, the assumption is that the viewer
almost completely adapts to the media white. For a display we'll
assume the same", so by putting the actual display
white in the 'wtpt' tag, you get a PCS and Relative Colorimetric
response that is set at D50, and by applying the usual Absolute
Colorimetric transform using the 'wtpt' tag, you can recover
the absolute white point of the emissive display. Sort of.
The only difficulty was the one caused by the use of the
"wrong Von-Kries" transform, which is innocuous for
most reflective media where the paper is almost white,
but has problems when you're transforming from D65 to D50,
or where the paper is distinctly tinted.
A display profile really should use a better chromatic
transform than this, and most did (ie. sRGB). This left
a CMM that followed the spec. slightly out of kilter.
Introduce the 'chad' tag in ICC V2 to specify how the
white point transform should be undone, rather than
use "wrong Von-Kries", and problem solved. It all works.
But in V4, all this has been altered. Now the monitor
white is to be interpreted as an "illuminant", and
now the 'chad' tag is to only apply to the chromatic
transform used to transform the illuminant to D50.
So now a displays absolute intent no longer undoes
the display white point to D50 transform, and
functionality has been lost.
[Note that the 'chad' specifications have subtly changed
between V2 and V4. In V2 is was
"This tag converts the input XYZ color, measured at a device's
specific viewing conditions, to the output XYZ color in the PCS
viewing conditions after complete adaptation."
In V4 it is:
"This tag, which must be invertible, converts an XYZ colour,
measured at a device's specific illumination conditions, to
an XYZ colour in the PCS illumination conditions after complete
adaptation."
So in V2 the 'chad' tag could be interpreted to cover the
Relative to Absolute white point transform whereas in V4
it cannot.]
What this means in practice is a reduction in ICC functionality.
Using V4 profiles
and the ICC intents, it is no longer possible to do absolute colorimetic
matching on monitors. You're stuck with relative colorimetric.
Umm, no. The assumption is that the human has chromatically adapted to
their display's white point, regardless of what it is.
I don't see how you can claim that functionality hasn't been lost.
With V2 it was possible using Absolute Colorimetric to get the absolute
display white (ie. D65). Under V4 you will get D50. The 'chad' tag
does contain the information needed to reverse this, but there
is no official intent that makes use of it. (The only official
use of the 'chad' tag in V4 is to use it to transform data in
a profile into a state as if a different illuminant chromatic
adaptation had been used). So if you try and do proofing using a
display with V4 and ICC Absolute Colorimetric, it won't work, where
under V2 it did work.
> further chromatic adaptation for that white point is unnecessary when making
> that assumption.
The point is not "Further chromatic adaptation", it is undoing the
adaptation the CMM has assumed, to be able to do real absolute intent.
A proofing situation is one in which you can't assume the user is
100% adapted to the display, since that's not the only thing they
are looking it.
> BUT this doesn't mean that you don't get paper white
> simulation to the display anymore with AbsCol (i.e. paper white
> simulation in Adobe apps). All you have to do is try it. It still works
> and in my view it's working a lot better than it used to.
Looking at the V4 mechanisms, and reports recently from people using
V4 applications, I think my understanding is correct, and in
my last discussions with Marti Maria we both agreed this was the case.
If you can point to something specific in the specification that
I've missed, I'd be grateful.
If you aren't seeing this effect, perhaps the profiles or application you
are using isn't ICC V4 compliant ?
It's also possible for a V4 application to implement a non-ICC standard
Absolute intent, which makes use of the 'chad' tag to restore the V2
functionality.
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden