[Tig] color table: system gives name?

[Richard Kirk]

> On Sat, 21 Jun 2008, Rob Lingelbach wrote:
>>
>> Are there any grading systems in use by colorists using film and/or
>> video that can output a color name, input the values from the table
>> at http://www-swiss.ai.mit.edu/~jaffer/Color/cne-2007-rgb.txt ?

Sort-of been there, sort-of done it. I remember the phrase 'putting the
Cray in Crayola' from the nineties. Back then, I worked for Canon. One of
the things I looked at was a natural language interface for colors.

Truelight has this in the command line utilities and in the library, but
it has never been used in a product. When we handle test strips, the
patches are labelled "0000" "0001" and so on. These strings are all
numbers, but any string is supported, and I have experimented with using
tables of named colours.

There are some 'natural' colours that people with normal eyes can detect
unambiguously - the invariant yellow and the invariant blue-green /
sea-green / cyan can be picked out by eye to a nm or so. Deep reds, which
cover a large wavelength range also have a consistent appearance. Violets
are harder to judge but I think they look the same. However, if you have a
monochromator and you are trying to find a typical 'green' or 'blue' you
will find there is a whole range of possible colors, and the best you can
do is to find the middle of the range which isn't obviously something
else. Oddly enough, I was doing this exact thing a couple of weeks ago,
using Thomas Young's wavelengths from 1802-3. It would seem that his 497nm
'blue' is what we would call 'cyan', and his 'indigo' is what we would
call 'blue'. Since then, I got a copy of Steve Shevell's 'The Science of
Color', and to my surprise there in chapter 1 (by J.D.Mollon) is more or
less the same argument, only that was written in 2003 (bah!).

Things get worse when naming colours like 'cream' which can be anything
from pink, through orange to yellow, but are mostly white. Yellow can be
'gold' when it is metallic, 'amber' when it is translicent, 'sodium' when
it is emissive - we may need to specify whether our colours are on a
display or a printed copy or a scene. Sometimes you cannot find a good
match, and sometimes the name can seem absurd in particular contexts.

> All of these color names are pointless without a standard for what the
> numbers mean.  If you can't obtain a reference print of the color, or
> reproduce the color via calibrated mean, then the color name is only
> useful as a guide for user interfaces. The MIT X11 color names were
> established in the late '80s when computer displays were not quite the
> same as now.  The W3 standardized colors for CSS/SVG
> (http://www.w3.org/TR/css3-color/#svg-color) are based on standard
> sRGB. It could be assumed that all numbers are specified in computer
> sRGB but of course hardly anyone does serious grading in sRGB. What
> are the equivalent numbers in CineonLog or Rec.709?

Don't see anything wrong in grading in sRGB. It has a sensible tone curve
with a linear region near the dark point. I prefer it to the DCI spaces.
However, that's another topic. I can certainly read in these values, and
convert them to equivalents in Video (not Rec.709, as that's a camera
standard) and some of the colors will have equivalents in Cineon Log, when
printed under typical conditions (say ArriLaser on 5242, printed on
Vision, with LAD 1.09, 1.06, 1.03). It would be a couple of minutes' work.
I am not sure I want my name associated with a list of named film colors,
at least not without a good understanding of what the numbers mean.

If you have a list of named colors, I can imagine a color picker that
would show you a patch of your picked color, and a few surrounding named
colors, so you could pick out a name that expressed what you understood.

But is it be useful? Anyone?

Cheers.
Richard Kirk
-- 
FilmLight Ltd.     Tel: +44-(0)20-7292-0400 or -0409-224 (direct)
Artists House,     Fax: +44-(0)20-7292-0401
14-15 Manette Street
London W1D 4AP, UK