Graphics Reference
In-Depth Information
ever be far apart, or will they generally be close together? Indeed, an interesting
experiment to perform when you want to write a color-interpolating program is to
create some sample inputs for your program, and try to approximate the outputs
that you'd like to get. If
you
can't approximate the output, you're unlikely to be
able to write a program that does so.
We use color for aesthetics, to establish tone or mood; for realism, to identify asso-
ciated groups of entities; and for coding of types of interaction. With care, color
use is effective in many of these roles. Careless use of color, however, can be disas-
trous; in one experiment, the introduction of meaningless color to a monochrome
interface reduced user performance by about two-thirds [KW79]. Color should be
employed conservatively; decorative uses should be subordinate to functional uses
so that color cannot be misinterpreted as having functional meaning. The use of
color, like any other aspect of a user interface, must be tested with real users to
identify and resolve problems. One conservative approach is to design first for a
monochrome display, which ensures that the color use is purely redundant (and
guarantees usability by color-deficient users).
There are many topics about the use of color for aesthetics, including [Bir61];
we state here just a few of the simpler rules to produce color harmony. The
most fundamental rule is to select colors according to some method, typically
by traversing a smooth path in some color model, or by restricting colors to a
plane in some color space. This might mean using colors of constant saturation
or value, for instance. Furthermore, it's wise to choose colors at equal percep-
tual spacing, which is not the same as equal coordinate distance in whatever color
model we're using: Conversion to CIE
L
∗
u
∗
v
∗
coordinates, or some other system
in which perceptual distances are accurate, is essential.
A random selection of hues and saturations is usually quite garish; grouping
colors so that those of similar hue or similar saturation are nearby is more attrac-
tive (but the distinction between colors may be less obvious).
If a chart or table contains just a few colors, the complement of one of them
makes a good choice for the background; a neutral gray is a good background
for photographic or similar imagery. If adjoining colors are not harmonious, a
thin black border between them will often help resolve the contrast. In general, a
parsimonious approach to choosing a color palette is wise (except in the case of
realistic images, of course).
Color can be used to code data (indeed, this is a standard tool in scientific
visualization applications), but several cautions are in order. First, color codes can
carry unintended meanings. If we display the earnings of company A in red and
those of company B in green, we may suggest to the viewer that company A is
in financial trouble because of our learned association of “red” with “in debt” in
financial situations. Bright, saturated colors stand out more than dim, pastel colors;
this may give unintended emphasis. And two unrelated elements of an interface
that have similar colors may be perceived as related, even if the color was intended
as purely ornamental.
A number of color-usage rules are based on physiological rather than aesthetic
considerations. For example, because the eye is more sensitive to spatial variation
in intensity than it is to variation in chromaticity, lines, text, and other fine detail
