Re: Monitor calibration software/hardware
Re: Monitor calibration software/hardware
- Subject: Re: Monitor calibration software/hardware
- From: Chris Murphy <email@hidden>
- Date: Fri, 21 Sep 2007 19:53:46 -0400
A few points.
In many instances it is not possible or even desirable to get a 1:1
reproduction of scene colorimetry. It may not be possible due to
device limitations, and it might not be desirable because it would
look wrong because of the difference between scene and print dynamic
range. Dynamic range affects scene color appearance but not scene
colorimetry.
So the idea that we want to have a completely reversible transform in
all instances doesn't make sense. In the case of CLUT based display
profiles for an LCD, and using a gamut compression option or CAM when
building the profile will get you a display profile that won't round
trip well. That's kindof the point. You give that profile LAB values,
and get RGB values that intentionally cause different LAB values to
be displayed than what you asked for.
And since we're talking about display profiles, there is in effect
only one rendering intent because even if the display profile
contained different table data for at least two intents (say
perceptual and colorimetric), we don't have a mechanism to select
which intent is used. Therefore in practice there is only one set of
table data and all intents point to that single set of table data.
With regard to white and black luminance, it's not really applicable
to talk about delta E. The real target should not be a specific white
and black luminance, but rather a dynamic range or contrast ratio.
And that would be something close to print contrast ratio because we
don't really have 10000:1 contrast ratio displays that would then
reproduce the scene color appearance of scene colorimetry. But what
our display's contrast ratio is depends highly on ambient lighting
conditions, and our typical display measuring devices are not at all
equipped for doing this correctly.
As for validation I think it's OK to use the same measurement device
for calibration validation. This is essentially pass/fail, and uses a
behind the scenes set of metrics to determine that the expected
behavior is what's occurring.
That is, if the starting LAB value, passed through the profile,
resulting in RGB value to display, which should yield a DIFFERENT LAB
value. An *expected* LAB value. And compare the expected LAB to the
actual measured LAB value. Are measured and expected LAB values the
same and within tolerance? Usually what's compared are just starting
LAB values as though the profile's job is to guarantee that LAB value
is reproduced, which often is not the mandate.
Even if you expect a full roundtrip of LAB (LAB goes through the CMS
and that LAB value is displayed as far as the measurement device of
choice is concerned) it's still a problem to use delta E because that
implies some level of actual performance or accuracy. If you have a
display calibration validation process that goes through 20 LAB
values, passes them through the CMS and then for the display to
display them, and they get measured and are exactly dead nuts on and
the display gets deltaE=0 for all 20 patches, this translates into
"it's working correctly, you have accurate color." But that's
somewhere between misleading and false because it does not at all
account for dynamic range or the ambient lighting conditions.
I do like the idea of verifying that the display is behaving in a
manner consistent with its behavior when it was profiled, and
tracking the device's behavior as it ages. Quite useful. The use of
deltaE I think is distracting and isn't really helping us make better
choices. It's not even clear to me which deltaE various products are
using, and that too makes a big difference. Even deltaE00 is far from
perfect.
Chris Murphy
Chris Murphy
Color Remedies (TM)
New York, NY
----------------------------------------------------------------------
Co-author "Real World Color Management, 2nd Ed"
_______________________________________________
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