Re: Evaluation of Color Difference Formulas
Re: Evaluation of Color Difference Formulas
- Subject: Re: Evaluation of Color Difference Formulas
- From: "dpascale" <email@hidden>
- Date: Thu, 19 Jul 2007 17:35:51 -0400
Klauss,
[...] there is a concensus in the scientific community that the CIEDE2000
color difference formula is the best... ...on average.
... on average color differences? on average viewing conditions? or both?
The formulas only compare coordinates. The newest ones twist the CIELAB
space to make it look more uniform. The various appearance models also twist
the space to take into account viewing conditions, so...
Do you think color difference formulas with additional parameters that
account for different viewing conditions (like CAMs) could be useful in
this context?
... potentially yes, but do not forget that, for all intent and purposes, a
color-difference formula is only to judge the quality of a match,
irrespective of the viewing conditions (apart from chromatic adaptation).
Users want a fixed number, not something that varies according to
illumination intensity and its color.
Can you tell anything about
<http://www.imaging.org/store/epub.cfm?abstrid=22169> or
<http://www.iscc.org/ISCC2007/abstracts/T11IG1_Xue.pdf> ?
Anyway -- the first time I looked on a image with color pairs ordered by
DE00, I was very astonished how bad this sequence corresponded with my
perception -- see <http://digitalproof.info/FOGRA39/fo27vs39.tif> (1,9MB
Lab-TIFF). I assumed it was because of 8-bit quantization errors, gamut
issues, simultaneous contrast or any other of these countless color
appearance mysteries.
But now I see a new aspect: Maybe Phenomena like the Hunt Effect
(colorfulness increases with luminance), the Stevens Effect (contrast
increases with luminance) and the Helmholtz-Kohlrausch Effect
(brightness depends on cromaticity) affect the validity of Color
Difference Formulas in different viewing conditions more than I expected.
Are there any studies about it? (I guess there are).
I have not read these articles but I have read this very recent article:
----------------------------------
Proposal of a New Concept for Color-Appearance Modeling
Yoshinobu Nayatani, Hideki Sakai
COLOR research and application
Volume 32, Number 2, April 2007
Abstract: A new type of color-appearance model (CAM) is
proposed together with its concept and flow of formulations.
The topics described are: (1) The existence of two kinds of
color-appearance models, CAMs previously used and CAMs
newly proposed. (2) All the CAMs, previously developed and
used, do not predict color-appearance attribute of perceived
lightness of object colors under any illuminations. They may
be adequately called ''the model for predicting color-appearance
match between object colors under different adapting
conditions.'' (3) Newly improved CAMs take the Helmholtz-
Kohlrausch effect in the VCC method into account. They can
determine object colors with the same Tone (equi-perceived
lightness, equi-whiteness-blackness, and equi-perceived
chroma) irrespective of hues under reference illuminant. The
newly improved models can be named Integrated CAMs.
Their applicable fields are described in detail.
----------------------------------
Nayatani is the guy who formulated the Helmholtz-Kohlrausch Effect, and this
article is a high flying view of the concepts and things to look at in
describing color appearance.
I would not say that the lack of a perfect fit between what we see and the
formula numbers comes from different viewing conditions (not to say it has
no effect), but mostly from an inadequate model on how to describe color.
Take all the examples of your color list and look at them in isolation; with
a D50 white background to be properly adapted, and you will arrive to the
same conclusion that the perceived differences do not fit one-to-one the
number order. However, globally, the perceptual differences for the higher
calculated numbers (the first column) are also more visible than for the
second column, and the second collumn differences more visible than the
third, etc.
Danny Pascale
email@hidden
_______________________________________________
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