Re: The DeltaE 2000 color difference formula [was: Evaluation of Color Difference Formulas]
Re: The DeltaE 2000 color difference formula [was: Evaluation of Color Difference Formulas]
- Subject: Re: The DeltaE 2000 color difference formula [was: Evaluation of Color Difference Formulas]
- From: "dpascale" <email@hidden>
- Date: Fri, 20 Jul 2007 00:08:18 -0400
Marco,
This is the idea!
Now, in practice, if you would ask your customer to judge three color pairs,
each with, lets say, a 2 deltaE difference, but each from a different
parameter (i.e. 2 deltaE due to lightness, or 2 deltaE due to saturation
(chroma), or 2 deltaE due to hue) will he/she recognize that these are
differences of the same amount?
They may well first notice the lightness difference and think the other two
are less in error. They may not realize that they should look at steps in
discernable differences of individual parameters (lightness, chroma, hue).
Expressed otherwise, they probably "expect" (wrongly) that a 2 deltaE
difference in lightness "looks" similar to a 2 deltaE in hue and, because
they do not look similar, say that these two differences are not the same.
Of course, only our customers would do such mistakes, and not us, since we
know better...
On the other hand, we need more than "knowing better" to explain to a
customer that What You Think You See Is Not What You Really See!
(WYTYSINWYRS)
Danny Pascale
----- Original Message -----
From: "Marco Ugolini" <email@hidden>
To: "ColorSync Users Mailing List" <email@hidden>
Sent: Thursday, July 19, 2007 10:35 PM
Subject: The DeltaE 2000 color difference formula [was: Evaluation of Color
Difference Formulas]
In a message dated 7/19/07 12:48 PM, Klaus Karcher wrote:
dpascale wrote:
[...] there is a concensus in the scientific community that the
CIEDE2000 color difference formula is the best... ...on average.
I'm not a color expert of the caliber of Danny, but, since we are all more
or less confused about the issue of calculating color differences (me
included), I feel like sharing my understanding of the DeltaE 2000 color
difference equation -- though I apologize if my contribution may be found
lacking in scientific precision.
Often, when this issue of color difference equations is raised, the
language
soon becomes very obscure, a conversation for the cognoscenti only, and
mine
is just an attempt to bring this subject a little closer to earth,
possibly
without introducing erroneous information into the mix.
My understanding of the DeltaE 2000 (a.k.a. DE00) equation is that it
accounts for the non-linearities of the CIELAB space more effectively than
the other available color difference equations. ("Non-linearities"
indicates
the *varying* distances -- some shorter, some longer, depending on hue,
saturation and luminosity -- between colors in CIELAB space whose visual
differences are instead perceived by humans as more or less of *equal*
magnitude.)
When we calculate DeltaE difference values, we do it with the intention of
using those DeltaE values in a linear fashion -- meaning that we imply
that
a 5 DeltaE difference between 2 samples, say, in the green region of the
visible spectrum describes a visual impact closely comparable to that
produced by a 5 DeltaE difference between 2 samples, say, in the yellow
region. But this is often not the case.
Color differences between samples are perceived differently by the human
eye
depending on the region of the spectrum in which these sample pairs are
located, and, depending on how accurately it is calculated, a 5 DeltaE
difference in the yellows may *not* have a visual impact as strong as that
of a 5 DeltaE difference in the greens. That's because our ability to
discern hue and saturation differences in the yellow portion of the
spectrum
(or in the deep blue or deep red regions) is less keen than it is in the
green region. (The peak of human vision is reached around wavelengths of
550-570nm, in the green region of the visible spectrum.)
So, a good color difference equation will have to compensate for those
imbalances and irregularities in human vision that simple linear math does
not adequately account for. The DeltaE 2000 equation uses scaling
functions
to that effect, which attempt to match the impact of the differences
*measured* by the formula more closely to the visual differences actually
*perceived* by the average healthy human eye. In other words, a difference
of 5 DeltaE 2000 hopefully *feels* visually like 5 DeltaE no matter the
region of the spectrum in which it is being measured. One should not have
to
make mental calculations to judge whether a 5 DeltaE difference is
acceptable or not depending on which colors it applies to (e.g., OK for
greens, but not for yellows, etc.).
It appears to me that such is the aim of an "ideal" color difference
equation, as this quote seems to confirm:
"When instrumental measurements are made, the numbers should predict what
observers see. The calculated color differences should predict visual
color
differences."
[Roy S. Berns, "Billmeyer and Saltzman's Principles of Color Technology",
3rd Edition, Wiley-Interscience, 2000, page 108.]
DeltaE 2000 is neither "ideal" nor "perfect", and I'm sure that future
color
difference equations (perhaps based on color appearance models like
CIECAM)
will tackle the task even more effectively. But I understand that the
color
scientists at RIT's Munsell Color Science Lab (Roy Berns himself among
them,
if memory serves me well -- please let me know if I should be mistaken)
view
it as the most accurate to date among those available for use by the
public.
If I made any factual errors, please be kind. I'm trying... :-)
Thank you.
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