Graphics Reference
In-Depth Information
Lighter colors tend to make areas look larger than darker colors, so using colors
withunequal luminancevaluesmakesitdi
culttocompareareasizes(Cleveland
and McGill,
).
Color palettes derived from nonuniform color spaces may contain unbalanced
colors with respect to their colorfulness or brightness. When tiles are shaded us-
ing such a palette, some of them might appear to be more important than others
in an uncontrolled way.
Due to the three-dimensional nature of human color perception (Mollon,
), it
has been common to specify colors using the three primaries red, green, and blue
(RGBcolors),especially forcomputerdevices.heappearance oftheon-screencolor
is affected by the characteristics of the device used. For example, the intensity I of
a primary on a particular device follows the rule I
L
γ
,whereL is the value for
the primary color and γ is device-dependent (but typically close to
.
). herefore,
a first caveat is that if colors are to appear identical on different devices, their gamma
characteristics must be taken into account.
Choosing colors and color palettes in RGB space can be a bit inconvenient and
hencesotwareimplementationsareotenbasedonhue-saturation-value(HSV)col-
ors (or the comparable hue-luminance-saturation color scheme). Both spaces are
rather similar transformations of the RGB space (Brewer,
; Poynton,
) and
are very common implementations of colors in many computer packages (Moretti
and Lyons,
). Each color in HSV space is represented by three dimensions (H,
S, V): the hue H (the dominant wavelength in the spectrum, in
=
,
), the satura-
[
]
tion S (the“colorfulness”or“pureness,”in
,
),andthevalue V (the“brightness,”
[
]
).
hese intuitive dimensions make HSV colors easier
to specify than RGB colors. However, HSV colors have several disadvantages. Most
importantly, HSVcolorsarenotperceptuallyuniformbecausethethreeHSVdimen-
sions map only poorly to the three perceptual dimensions of the human visual sys-
tem (Brewer,
; Ihaka,
). One important issue here is that HSV dimensions
are confounded, e.g., saturation is not uniform across different hues. As an exam-
ple, see Fig.
.
(let), which shows a qualitative color palette (colors
the amount of gray,in
[
,
]
)
for varying hues H) in the HSV space: although saturation and value are fixed, the
fully saturated blue is perceived to be much darker than the fully saturated red or
green, making it di
cult to judge the size of shaded areas. Furthermore, the flashy
fully saturated HSV colors are hard to look at for a long time. For similar reasons,
it is equally di
cult to derive acceptable diverging palettes from the HSV space, i.e.,
bipolar scales containing colors ranging between two very distinct colors. he up-
per part of Fig.
.
shows a diverging palette in the HSV space with colors ranging
from a saturated red
(
H,
,
(
,
,
)
to a neutral white
(
H,
,
)
to a saturated blue
(
.Although the palette should be balanced with respectto colorfulness
and brightness, the red colors are perceived to be more intense and flashy than the
corresponding blue colors.
,
,
)
In many implementations, all three dimensions are scaled tothe unit intervalinstead of the
coordinatesused here.
