Re: perceptual differences in Lab deltaE
Re: perceptual differences in Lab deltaE
- Subject: Re: perceptual differences in Lab deltaE
- From: Wire ~ via colorsync-users <email@hidden>
- Date: Tue, 7 Jan 2020 11:31:02 -0800
On Tue, Jan 7, 2020 at 04:14 Roger Breton via colorsync-users <
email@hidden> wrote:
> Hi Simon,
>
> [...] I'm curious how were you able to print chart patches with as little
> as 0.3 deltaE (which flavor?) color differences... Suppose I was to
> replicate your findings with my students?
>
> Best / Roger
>
> -----Original Message-----
> From: colorsync-users <colorsync-users-bounces+graxx=
> email@hidden> On Behalf Of simon--- via colorsync-users
> Sent: Tuesday, January 7, 2020 5:48 AM
> To: Roger Breton via colorsync-users <email@hidden>
> Subject: perceptual differences in Lab deltaE
>
> Hi Andrew,
>
> I did some tests a few years ago by printing charts patches to exhibit
> small deltaE variations.
Ah Roger, I see why you posed the original query re billions of colors:
you're an instructor!
And I found the gap with Andrew...
Roger begged the question "only so many colors are useful," which Andrew
seized as hinging on a definition of the term color. Seems fair.
The discussion fractured between Andrew's reframing "color" to mean a
uniformly perceptually distinguishable difference under CIE vs my point
that the gear industry brags about capability to produce a range of
stimuli.
The question Roger posed was directly in context of assessing display
performance, so by my reckoning the claim of "millions" or "billions" is
understood in that context as a device capability. The perceptual factors
are much more complex and subsumed by the device physics under ICC.
To drill into this a little further: We know that the device is built to
the tri-stimulus model, and varies its output in a roughly linear response
to data on its inputs. Whether anyone can or cares to see the output
needn't be considered given the electro-mechanical basis of the device:
three primaries, intensity controlled as individual channels (guns) and
mixed through an aperture in direct proportion to the input channel data
values. As it is well-known with such devices to be useful to go higher
than 8 bpc to avoid certain rendering artifacts — regardless of other
aspects of waste in the data format — and because the system physics are
linear WRT inputs, aaannd as the inputs are numerical supporting millions
or billions of combinations, we arrive at a euphemism of a count of colors,
which seems to me wholly appropriate and fair. If you want to know what
color and count really mean, further study the device.
Andrew seized on the term "color" and re-contextualized it for reasons I
don't quite follow, reframing device color as a count of discernible
regions within the CIE spectrum locus according to some dE threshold
criteria. Sure it seems like an interesting way to count colors. And it's
the way ICC colorimetry works! So ok fair enough. But what does it have to
do with the original question in context of display vendor claims? Don't we
use tools like DisplayCal to just run gamut plots normalizing device
numbers are to generic device spaces. Andrew likes ProPhoto, the web likes
sRGB, the device does Adobe, etc. The plots tell us how well the specific
display fits as a simplified percentage or let us look at detailed
differences. And so on... Well, to the extent that the device adheres to a
standard color space, which is now common, the input data format can be
considered to be a literal count of colors within that space, according to
explicit CIE criteria! Andrew's deal about qualifying the device response
to his personal definition of what can be considered a color seems
tyrannical. The science is clearly agnostic to the concern of the meaning
of qualia. The CIE model is an approximation. Andrew is free to pick what
he feels is a useful distinction of color, but the system avoids this; a
distinction is not required. In fact, the point of the system is to
provide the tools to create whatever distinctions work for you.
Then I began to notice something happening here to the dialog: the whole
thing has become circular.
Andrew issued an edict that any count of colors is only significant
according to his personal CIE dE criteria which he is claiming as the
philosophically one-true and real definition of color without questioning
his methods or the limits of the CIE model and the fact that the CIE
experiments were carried out using devices similar in physics to the ones
he thinks over-count colors.
Roger and Simon are replicating the color matching experiments that are the
basis of the CIE model in terms of the model's measures of discernible
differences and asking each other if you can apply these measures to create
a discernible difference! And they're not sure, but suspect they can!
The web may explode at any moment.
Andrew, I want to give you the benefit of my doubt for what seems a
well-intentioned point of enlightenment about how to think about color in a
way that's vigilant to gear-industry hyperbole. But your arguments read to
me as mansplanation.
I had wondered if Roger's original question was intended as a red herring?
Now I see you're after teachable moments. But this makes for a rabbit hole.
(insert pithy Lewis Carrol Alice quote)
I've nothing more to add, so over / out for now and carry on.
Dell rulez!!
_______________________________________________
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