Re: CNS colors

lilley (lilley@afs.mcc.ac.uk)
Wed, 7 Feb 1996 12:01:19 +0000 (GMT)


David Perrell writes:
> At 11:48 AM 2/2/96 +0100, Hakon Lie wrote:
> >Chris Lilley has brought to attention a color naming scheme that fits
> >neatly into CSS. CNS is described in [1].

> Hello. I'm a graphic designer of 23 years and a programmer of 12,

nice credentials. But you write:

> and I think CNS is a bad idea.

which would be fine if you had a technical reason, but you go right
on to say:

> Obviously CNS (which I'd never heard of before) is a subset of the HSB
> (hue-saturation-brightness) color model, widely used and intuitive.

So you have newver heard of it, have a false idea of what it is, and
confidently state it is a bad idea...

Perhaps you should have taken the trouble to find out more about it ;-)
In particular, it has nothing to do with HLS, HSB and suchlike polar
representations of RGB (which are, in usability studies, often shown to
be *not* very intuitive).

CNS colours are specified in terms of the Munsell system, which can be
related via CIE XyY to the RGB values required on a monitor.

However, to take your other points in order:

> Most HTML markup will soon be performed within an editing program, where
> instant visual feedback of color choices will be the norm. Being forced
> to choose from an arbitrary set of 627 colors is an unnecessary constraint.

A natural language syntax is used for hand typing. It is easier to say
'vivid blue-green' than to figure out the #RRGGBB values, which many
less experienced HTML authors have trouble with (as countless postings
to usenet newsgroups will attest).

A GUI system obviously hides the details of colour specification and so
can use RGB specification, thus limiting the user (with current syntax)
to 16.7 million colours. For the purposes of CSS, this is probably
enough.

The referenced paper, which it appears you have not read, does however
speak of using CNS to produce a rough colour which is then refined
in a GUI editor. The same paper also has some impressive usability test results.

> The Pantone matching system includes over 800 spot color inks

Indeed it does (considerably more than 800). Conversion of these
colours to other colour systems, for example for display on a computer
screen, requitres a Pantone license. The idea of using Pantone in CSS
was suggested (by me) but rejected because of this licensing
requirement. I am sure you are also aware that many of the Pantone spot
colours are not withing the gamut of most RGB monitors.

> the TOYO system (also based on HSB) over 1000

TOYO is not "based on HSB" to the best of my knowledge. It is another
proprietary ink specification system.

> yet I have occasionally had
> to specify custom ink formulations to produce a non-standard color.

The relevance of this to screen presentation escapes me.

> For CMYK printing, millions
> of color mixtures can be specified, and millions of distinct colors
> can be produced on a good monitor.

Monitors do not use the CMYK colour model. CMYK is intimately tied to a
particular ink set, screening algorithm, printing press and paper stock
and is of no use whatsoever for online presentation which is the main
aim of CSS. Systems such as Adobe Photoshop which allow on-screen
preview are converting from CMYK to CIE LAB (using these printing
details) and thence to RGB (using the monitor details) for display.
Even with all this effort, the colour match between the screen and print
representations is often not good, as I am sure you are well aware.
CMYK is of zero use for CSS.

> Not all cultures put the same stress on the same colors as we do, and
> "blue" may not mean the same thing to two different people.

Hence the need for an unambiguous definition which can be related directly
to colours which are objectively measurable in an international standard
system. This is provided by CNS.

> To base a
> color model on the basic set of crayons used by euro-cultured children
> seems short-sighted.

It would be, if anyone was proposing it ;-) I'm not, are you?

> I strongly believe we need a numerical color representation,

Indeed we do, and we have one (RGB). We also need a natural language
representation to aid less technical users, and CNS amply fits the
bill.

> and we should stick with color models currently in
> use, particulary RGB and HSB

HSB is a spectacularly bad idea as it is non intuitive. It claims for
example that yellow (RGB 00FFFF) and blue (RGB 00FFFF) have the same
"brightness" which is clearly false. It is extremely non linear, the
hue circle is not at all even .. generally, it is a mess.

The overall concept of a polar model (as used for example by Munsell,
OSA, Itten, NCS, CIE LCH and so on is fine. The particular
implementation in HSB is not.

> and possibly LAB (a CIE-standard color
> model based on human perception of color and widely used in the
> broadcast industry)

The broadcast industry uses LUV, not LAB, but I take your point.
Indeed, the reason I suggested CNS is that it may be unambiguously
related to CIE colour spaces.

and CMYK (standard for the print industry).

See above.

If the CNS scheme is specified, it should be as an optional subset to
a complete numerical representation of HSB.

Well since HSB is but a renotation of RGB, that is the current
situation exactly.

-- 
Chris Lilley, Technical Author and JISC representative to W3C 
+-------------------------------------------------------------------+
|  Manchester and North Training & Education Centre   ( MAN T&EC )  |
+-------------------------------------------------------------------+
| Computer Graphics Unit,             Email: Chris.Lilley@mcc.ac.uk |
| Manchester Computing Centre,        Voice: +44 161 275 6045       |
| Oxford Road, Manchester, UK.          Fax: +44 161 275 6040       |
| M13 9PL                            BioMOO: ChrisL                 |
| Timezone: UTC        URI: http://info.mcc.ac.uk/CGU/staff/lilley/ | 
+-------------------------------------------------------------------+