[argyllcms] Re: help understanding argyll dispcal and spotread output
- From: "Wire ~" <wire@xxxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sat, 29 Feb 2020 11:22:06 -0800
Graeme, just catching up here...
Delta E to reference is -0.005080 0.048635 -0.008299 (0.049599, CIE94
0.049599)
First three numbers are the delta *L, delta *a & delta *b.
First number in brackets is the total delta E (Euclidean sum
of the first three numbers, AKA delta E 76.) Last number is
the total delta E in CIE94 weighting.
Re spotread, thank you! It's the CIE76 euclidean measure that I wanted to
be sure about. I went through manpage but didn't see that. While I suppose
that context should imply it, I'm not experienced enough to be sure. Maybe
good to drop those sentences in page for n00bs.
* * *
Re Visual temperature...
The googlez rank (as I see it in my searches) no page as definitive for the
term Visual Color Temperature. Putting pieces together, it appears this is
an Argyll specific term that brings display temperature readings up to date
with CIE diff conventions.
I asked over at DisplayCal forums and was directed over here. DisplayCal
reports VDT in its Interactive display adjustment window for setting white.
It's color meters work the same way as CCT, which feels intuitive, but the
VDT number is not intuitive.
The dispcal man page has what I see as the definition of Visual Color Temp
in the form of a Zen koan:
//The *visual color temperature* values are calculated similarly to the
correlated color temperatures, but using the modern CIEDE2000 color
difference formula to calculate a better visual approximation to the
closest temperature to the displays white point. There will be no
difference between the UCS and CIEDE2000 temperatures if the display white
point actually lies on the particular locus.//
Unfortunately this definition is taxonomically a case of what the "Elvis of
Philosophy," Slavoj Zizek, refers to as a Hegelian negation: that which is
defined by what it isn't. Like the term 'the undead': which means not dead
but also not living. To paraphrase the above: VCT is defined as visually
better than CCT, where the visual is not seen, except when there is
no-difference!
Very tough for neophytes to get this... It had me wrapped around a mental
axle for a whole day. So, if I may be bold and restate the Argyll def in
more affirmative terms:
::VCT is the CCT with a MacAdam ellipse of a degree (step-number)
corresponding to the reported dE*00 offset of the measurement, where the
edge of the ellipse just touches the Planckian locus. VDT is the daylight
form.
Is this correct? I'm struggling with the terminology
If so, there's a question as to whether the ellipse is centered on the
visual temp or the CCT because ellipse sizes vary a lot with distance along
and proximity to the locus. I'm begging the question here that it matters
for anything far away from say 6500K.
The nice thing about CCT / CDT near 6500 is the correlate lines extend
pretty much due north, so when aligning a display, you adjust red/blue to
dial in a temp then tweak +/– green to hit the locus.Maybe a second round
with R/B to pull temp in a little tighter to D65. With practice, this
becomes second nature. And it goes along well precal tools, such as
DisplayCal interactive adjustment window, where you get three color bars
that help you visually understand these dynamics. Then you grasp Duv very
intuitively. VCT throws this learning for a loop as now the display
controls cause the number to bob around within a MacAdam ellipse.
As to the little bugaboo about CCT errors of omission for Duv, maybe VCT
can help rehabilitate CCT? Because you have to understand CCT to grok VCT,
and the VCT green/purple offset cannot be optional. Unlike CCT correlate
lines it's harder (for me anyway) to visualize a the pattern of ellipses
that connect a temp to the locus. ... Idk, this last point may be
unimportant because we don't want white ranging around all over the place.
Please forgive me for thinking out loud here.
I can see it makes sense to have a CIE 2000 measure for white. I would like
to know how to use this well.
As a display guy—as different from a lighting guy—I have no issue with Duv
being an inscrutable little number, because it works like a micrometer: it
tells me to fine tolerance how well I'm tracking the locus. I am most
interested in the offset when I'm getting really close, otherwise I just
look at the white and think gross, that's off. VCT certainly offers this
too. But VCT pushes ambiguity around in wonky way. With a single CCT value
stripped of its offset, at least I know the intention for white,
To me, it feels like CCT says 'I intend to hit my whitepoint to some
tolerance', while VCT says 'I'm probably not gonna hit my target so what
temp am I hitting that's visually closes? Bletch. Why do I want to know
this? When I see 6300K VCT is that a statement of intention, or the best
they can do. I pick this example specifically because of DCI-P3 / Display
P3 variability. ...Head swimming...
I personally don't struggle with the omission of the CCT offset, as the
tolerance can often well be inferred in context If a display report omits
it without qualification I can fall back to the rule of +/– 0.003 and if
that doesn't work, something is prolly going wrong in a bigger picture
(npi) so that the number is probably irrelevant. While this ain't good,
does VCT have an advantage of making things a heck of a lot clearer? I'm
not feeling it. :)
I am a cynic. I see the great boon ICC tech is that we learn how
incorrectly rendered our images are with ever greater precision. For the
diligent, this allows them to make some situations much more right, but for
ignorant and lazy, things get even more wrong, continually exceeding what's
really working. A little like trickle-down economics.
I like that dispcal tells me 7 different ways to look at white.
Regards,
/wire
On Thu, Feb 27, 2020 at 16:23 Graeme Gill <graeme@xxxxxxxxxxxxx> wrote:
Wire ~ wrote:
Hi,
Dispcal reports four different white temps
CCT, CDT, VCT, and VDT?
See <http://www.argyllcms.com/pro/cmdoc/8000_config.html>
The first two I get... but what are the "Visual" forms?
They are the closest colors on the given locus (black body
or daylight) using CIEDE2000 rather than the obsolete
perceptual CIE 1960 UCS.
How reliable is the Approx gamme result—assuming conventional display
space
TRCs?
If you mean the "Aprox. Gamma" value printed by dispcal, then
they are approximate. It's reliable in the sense that it
represents the overall impression of the display curves effect
on perceived brightness, but of course may be quite different
from the detailed display transfer curve shape.
Cheers,
Graeme Gill.
Other related posts:
