More on Colour

[Chris Handley <cbmvax!uunet!pucc.princeton.edu!CHandley>]

Bruce writes

> 30-color theories? Or is the 0 a stray?

Oops, yes it was.

> The terms "red," "green,", "blue," "cyan," "magenta," and "yellow"
> have two meanings in color theory. Chris is talking about the
> points (Cyan = Hickethier 900, magenta = Hickethier 090, yellow =
> Hickethier 009) that define additive and subtractive primaries.
> But in some treatments, the color wheel is divided into six
> _regions_, based on which of the six are nearest; i. e., "magenta"
> means anything closer to subtractive-primary magenta than to
> additive-primary red or additive-primary blue, or in
> Hickethier notation anything of the form xyz with y>(x + z).

> (As you can see, I tend to like the Hickethier notation, which
> makes it very easy to describe colors in terms of primaries. The
> CIE notation is much harder in my thoughts to visualize.)

The problem with all the 3 (6) colour, additive/ subtractive systems
is that they are _wrong_! One cannot define the all spectral colours
in terms of any of them. CIE gets round it by specifying  a multi-
peak 'spectrum' for its primary colours, and allowing negative
weights on its additive colours, i.e. to get a match you may have to
add some colour to the sample. All the nice diagrams you see in the
colour textbooks omit to tell you that they are, pure and simple,
kludges. We can perceive the spectral colours, but there is no way
we can reproduce them (except spectrally) and we certainly cannot
photograph them or print them. You don't believe me? Fire up your
favourite painting program (even using 24 bit colour) and make me a
rainbow.

Returning to colour terminology. One thing I have seen on CIE
diagrams, although not on otther systems, are circles (ellipses) of
confusion. These are centred on a particular hue within the gamut of
three colour additive colours. They represent the 'average' area
within which colours are perceived toi be the same. As far as I am
aware, these are reasonably constant between people and cultures,
although most people would probably hard pressed to specify the
colours by names (BTW my wife and ai have the same trouble with blue
and green - jerseys, shirts etc). I suspect that the best would be
to split the visible spectrum up into some convenient number of
ranges (I could argue quite convincingly for 6, 7 or 8) and give the
middle of each range a colour name. It might be better to make this
as differenr as possible from any of the colour names in any of the
target languages (recognisability of -1 would be great (%^). Other
spectrals could then be defined (as precisely as you wish) between
these fixed points. For the less saturated colours, probably the
easiest is to take one of the standards (Hickethier seems pretty
good, although I was not aware of it) and go from there giving names
to all the hues and saturations it can produce. Whatever you do is
going to be wrong, because people just do not see that way, but at
least you can be consistent (and possibly consistently wrong)

Chris Handley                                    chandley@otago.ac.nz
Dept of Computer Science                       Ph     (+64) 3-479-8499
University of Otago                           Fax     (+64) 3-479-8577
Dunedin, NZ