Re: Monitor profile verification [was: Eye One Pro for monitor calibration? [was: Re: NEC 2690 SpectraView]]
Re: Monitor profile verification [was: Eye One Pro for monitor calibration? [was: Re: NEC 2690 SpectraView]]
- Subject: Re: Monitor profile verification [was: Eye One Pro for monitor calibration? [was: Re: NEC 2690 SpectraView]]
- From: Marco Ugolini <email@hidden>
- Date: Sun, 02 Dec 2007 01:05:45 -0800
- Thread-topic: Monitor profile verification [was: Eye One Pro for monitor calibration? [was: Re: NEC 2690 SpectraView]]
In a message dated 12/1/07 1:58 AM, Klaus Karcher wrote:
> Well ... you could /use/ another instrument (e.g. one of those cited
> 35k$ reference spectroradiometers) and compare the results ;-)
Hi Klaus.
That argument makes me smile too, and wonder how practical such advice is
for the average user, even the professional user. :-)
> I's somewhat misleading to use a instrument to verify it's own results:
Not if one's purpose is to verify the *internal consistency* of a monitor
profile -- meaning its ability to produce the values it was created to
produce by the selfsame measurement device -- rather than its *absolute
accuracy*. That's a fine point, but one worth pondering nonetheless.
There are two aspects to monitor profiling, in my view:
1) Internal consistency: is the profile producing values close to the ones
expected according to the measurement device used to create it?
2) Absolute accuracy: how closely do the colorimetric values predicted by
the profile match the colorimetric values actually measured by a
*reference-grade* instrument?
Point 1 is within the reach of the "common mortal" type of user. Point 2 is
for the lucky ones (like Roger) who have access to a $25K+ reference
instrument.
> It will only show weather the measurements are repeatable and the
> calibration and profiling process is consistent in itself.
Agreed.
> The test won't reveal systematic discrepancies of the instrument.
> E.g. if your instrument's measurements are always too blue, your monitor
> will show a yellow cast, but the verification won't reveal a color cast.
To be practical about it, a color cast will be detectable even by way of
visual inspection. If the device is *clearly inaccurate*, the discerning
user will detect it. If the inaccuracies are minor and hardly detectable
visually, then I wonder how important they are, ultimately.
In my opinion, an expert user learns fairly quickly whether or not one's
measurement devices are trustworthy or not. There is the mathematical side
of the accuracy question, and then there's the practical side of it too,
which includes not-quantifiable aspects as well, such as an individual
user's tolerance of the results, if they should drift somewhat from an
absolute ideal. That is completely up to the individual's preference.
> What you did with BasICColor Display's verification is to compare your
> display with your display. The results show how precisely your display
> can reproduce colors that fall within it's own gamut.
I think there is widespread confusion about this issue, so please allow me
to describe it as I understand it.
The verification procedure starts with a list of RGB numbers that are being
sent to the display: at their origin, these RGB numbers are simply numbers,
and have no *inherent* colorimetric appearance. It's only by virtue of the
active monitor profile that they assume a definite colorimetric appearance.
Next, the verification procedure takes measurements with the colorimeter;
after that, it normalizes and adapts the measurements chromatically to the
monitor profile's white point.
Finally, it produces two lists:
(1) A list of the L*a*b* numbers that are derived from the initial RGB
numbers when the active monitor profile is assigned to them (reference
values);
(2) The colorimetric values *actually* measured off the display (normalized
and chromatically adapted).
The resulting DeltaE numbers are calculated from the comparison of these two
lists of colorimetric values.
So, it seems clear to me that the verification process is merely an
*internal consistency check* (how close are the measured numbers to the
expected ones?). It's *not* an absolute accuracy check.
This is far from being pointless, or limited to tracking device drift, as
Andrew Rodney says, which is only partially true, in my opinion -- no slight
intended, Andrew. There is more value in the verification procedure than
that, because a profile that produces off the display colorimetric values
that are significantly divergent from the reference values -- when measured
with the *same* device that created the profile -- is inherently useless,
since such a discrepancy would constitute evidence that its internal
architecture is clearly faulty and mathematically unsound.
Best regards.
Marco Ugolini
_______________________________________________
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