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 14:52:34 +1000
Roger Breton wrote:
Graeme,
Allow me to dissect the MWP tag section and comment along.
I always felt that the MWP was more designed for printed media than for
self-luminous sources. But I'll admit I see nothing in this last sentence
that "binds" the definition of MWP to a printed sheet or a monitor display
since both can be assigned XYZ tristimulus values. And EyeOnePro or a Spyder
or a DTP94 or a SpectroCam can all return XYZ from a hardcopy and the
faceplate of a CRT or LCD.
I think you are quite right to get this impression. The ICC V2 had
the feeling of a simpler matrix monitor standard (ColorSync 1.0 ?) with
a more sophisticated print Lut based standard added on, but not fully
integrated.
If the media is measured under an
illumination source
which has a chromaticity other than D50, the measured values must be adjusted
to D50 using the
chromaticAdaptationTag matrix before recording in the tag.
In the previous specs, in the Annex, I remember there was reference to
adapting the colorimetry to D50 but it was left to the profiling package to
decide on which method of adaptation to use, XYZ scaling or some other
methods. In v4 specs, there is no choice.
The choices are the same. What has changed is than in V2 the interpretation
of an emissive devices white point as being an "illuminant" or a "media
white" was unspecified. In V4, it has been specified that an emissive
display is assumed to have a media white that is the perfect diffuser illuminated
by a (possibly) non D50 illuminant. Many V2 packages and standard profiles (ie. sRGB)
had made the opposite assumption.
The reason why the V4 choice is the wrong one in my opinion, is
that for reflective media the actual observed media white point
depends on an independent thing called the illuminant which
is likely to be shared for all prints viewed as a group,
whereas for emissive media, the white point (hence
V4 "illuminant") is a self contained property of the media,
and can't be shared with other prints or displays.
In practice most usual viewing situations are that the viewer
is viewing just the output medium, and the image will dominate
their field of view. It is therefore reasonable to assume
that they are mostly adapted to the image. If the images
neutral axis is in alignment with the media white point,
this means the they are mostly adapted to the media white point.
My own experience, the experience of many others, and the
default use of media relative colorimetry in the ICC itself
supports this assumption.
(It is a mistake I think for the ICC V4 spec. to claim
the contrary - Page 78 ICC V4.2 "Note 1").
As soon as we switch to a situation where more than one
media is being viewed, this adaptation assumption is
broken, and we need to recover the absolute colorimetry
to be able to deal with the situation.
Now for comparing two reflective prints, it is quite difficult
to set up a viewing situation where each print is illuminated
with a different illuminant, so in most situations the same
illuminant will be used to view two prints side by side.
The recovery of the reflective media's absolute media
white is all that is required to be able to make a visual
match between two prints under the same illuminant.
Comparing two displays, or a reflective print and a display,
the situation is different. If you treat the
display as a media with a perfect diffuser white and a non-D50
illuminant, the effective illuminants are different,
because the display "illuminant" is always inseparable
from the display. You can't easily "illuminate" your
print and your display with the same "illuminant".
If instead you treat the display's white point as
a media white point, then using Absolute Colorimetric
allows you to achieve a match between your display and
a reflection print, or between two displays, just like
the purely reflective situation.
> But regardless of the specs
version, adaptation is only be required when measuring self-luminous sources
since all reflectance measurements always use D50. Unless someone is crazy
enough to use D65 and adapts the values to D50 through a CAT!
The ICC specs use of a CAT was specifically to allow for the situation where
measurements are taken under non-D50 illumination, and only that situation.
It's non use of "wrong Von-Kries" transform for the media white point
relative transform is what made the treatment of an emissive display
white as "media" or "illuminant" such an issue. For reflective prints
most media is closed to being white (ie. close to being the perfect
diffuser), so the difference between the perfect diffuser and the
media white is usually pretty small (1 - 5 delta E). The difference
between a "wrong Von-Kries" and a perceptual CAT is very small. As
soon as you deal with D65 to D50, or really tinted paper, this
error can no longer be overlooked. Since printing on tinted paper
is not something that surfaced as an issue (the assumption
being I imagine that noone would expect accurate colorimetry
for such a situation), the display problem put it in focus.
Rather than deal with the use of "non Von-Kries", the display
problem was pushed of to the section of the ICC that did
deal with using a real CAT, allowing for a non-D50 illuminant.
In V2 this wasn't done in a "pure" way, and things like sRGB
sit in a situation in which the media relative transform
is being done using a Bradford CAT, rather than the
"wrong Von-Kries", where in V4 it has been made "pure" again,
the result being that Absolute Colorimetric intent has
been lost for display profiles.
So this says to me, nothing new, as far as measuring paper is concerned in
v4 specs. But very different, I'd venture to say better, for measuring
displays. Suppose I'm a die-hard for D65 calibration, this means that all
measurements of my screen have to be chromatically adapted to D50 through
Bradford before they find their way in the profile. So, here, I don't yet
see anything suspect or questionnable.
See above. The problem is that the ICC allows Bradford etc. for transforming
the illuminant, but not the perfect diffuser to media white transform, and
the latter is the one that is undone when Absolute Colorimetric is selected.
Again, here, nothing new. All colorimetry is specified relative to the
perfect reflecting diffuser (I never thought I'd see a "perfect reflecting
diffuser" in my life but BabelColor new white standard comes as close as is
physically possible to this as can be).
For displays, the values specified must be those of
D50 normalized such
that Y = 1,0 (i.e. 0,9642 1,0 0,8249).
Meaning that (absolute) display measurements have to be adapted (CAT02) so
that the XYZ of the white point is encoded as D50 in the profile.
No, the "Absolute Colorimetric white point" (the 'wtpt' tag we are discussing)
must be D50, not just the display measurements. This is the crux of
the difference between V2 and V4.
But, how is the new definition of the MWP "absolute" is wreaking havoc for
screen colorimetry? It did not strike me as such in the above.
See above.
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