Lecture 13: Color

Brian J. Smith

2026-03-24

Color

The beginning of this lecture is based on Chapter 10 of Visualization Analysis & Design.

“Map Color and Other Channels”

For this lecture, I’m going to start with some color theory from the VAD Text, incorporate a few other sources of information, then transition over to working with color in R.

Color Theory

Color Vision

• Our eyes have two kinds of receptors:
  • Rods
    • Actively contribute to vision only in low-light settings.
    • Provide low-resolution black & white information.
  • Cones
    • Main sensors in normal lighting.
    • Three types of cones for different wavelengths.

Color Vision

• Our visual system processes these signals into three opponent color channels:
  • One from red to green,
  • one from blue to yellow,
  • and one from black and white (luminance).
• The luminance channel provides high-resolution edge detection.
  • Red-green and blue-yellow channels are lower resolution.
• There is an important distinction between:
  • Luminance – how bright or dark a color is.
  • Chromaticity – what most people informally call “color”.

Color Vision

• Color blindness, more accurately termed color deficiency, affects a person’s ability to distinguish color along one of the chormaticity channels.
• There are three types of color deficiency:
  • Deuteranopia and protanopia
    • Affect red-green channel.
    • Are sex-linked (~ 8% of men).
  • Tritanopia
    • Affects blue-yellow channel.
    • Extremely rare, not sex-linked.

Color Spaces

• The color space of what we can detect is three-dimensional.
  • It can be adequately described by three orthogonal axes.
  • There are many ways to define these axes and convert between them.
  • Some color spaces are:
    • convenient/commonly used (e.g., RGB)
    • designed to match our vision (e.g., HSL)

RGB

• Red, green, blue.
• Computationally convenient.
• Not useful as separable channels; create an integral perception.
  • Additive color model.
  • Screens emit light.
• Not covered in the VAD textbook.

RGB

• Device-dependent color model.
  • Colors can vary slightly between devices,
  • or on a device over time.
• TVs, cameras, scanners, video cameras, smartphones, etc.
• See the Wikipedia RGB color model page.

RGB

• Given as a triplet of red, green, blue values.
• May range from 0 – 1.
• More commonly range from 0 – 255.
• Often coded using two hexadecimal digits (0-9, A-F).
  • Google color picker

CMYK

• Subtractive color model.
  • Inks absorb light.
• Cyan, magenta, yellow, black (= “key”).
• Used in color printing.
• Not even mentioned in the VAD textbook.
• See the Wikipedia CMYK color model page.

HSL/HSV

• HSL = Hue, saturation, lightness.
• HSV = Hue, saturation, grayscale value.
  • V is linearly related to L.
• More intuitive and heavily used by artists and designers.

HSL

• Hue captures what we normally think of as “pure” colors.
  • E.g., red, blue, green, yellow, purple, etc.
  • Not white or black.
• Saturation captures the amount of white mixed with the color.
  • E.g., pink is desaturated red (more white mixed in).
• Lightness captures the amount of black mixed with the color.
  • E.g., controlling the darkness of a gray.

HSL

• HSL is pseudoperceptual.
  • It doesn’t truly reflect how we perceive color.

“In particular, lightness L is wildly different from how we perceive luminance.”

HSL

Fig. 10.3

• Lightness is all the same here.
• True luminance is as measured with an instrument.
• L* is perceptually linear, the closest to what we see.

HSL

• Our perception of luminance depends on the wavelength of the light.
• Our spectral sensitivity curve for daylight vision is almost bell-shaped.

Fig. 10.4

HCL

• Hue, chroma, luminance.
• Used to address perceptual issues with saturation.
• Chroma = “colorfulness”.
• Not mentioned in the VAD textbook, but seems increasingly popular in the visualization world.

Choosing Color Channels

Choosing Color Channels

• Magnitude channels (ordered):
  • Luminance
  • Saturation
• Identity channels (categorical):
  • Hue
• Other channels (complementing the others):
  • Transparency

Choosing Color Channels

• We automatically perceive luminance and saturation as ordered.
• We do not automatically perceive hue as ordered.

Fig. 10.5

Magnitude Channels

• Luminance is suitable for ordered data.
• We have low accuracy in perceiving luminance of noncontiguous regions.
  • Due to contrast effects.
• Number of discriminable bins is probably < 5 if the background is not uniform.
• Luminance contrast is very important for resolving fine details.
  • E.g., we can’t read text without luminance contrast.
  • Hue and saturation are not important for this.
  • If we use the luminance channel for our data, we may not have it to use for other reasons.

Magnitude Channels

• Saturation is also suitable for ordered data.
• Also has the low accuracy problem for noncontiguous regions.
• Number of discernable steps is around 3.
• Interacts strongly with size:
  • Small marks (e.g., points/lines) are harder to discern.
  • Probably only want to use two saturation bins for small marks.

Identity Channels

• Hue is appropriate for categorical data.
• Hue is the highest ranked channel for categorical data after spatial position.
• Also hard to make small comparisons in noncontiguous regions.
• Also harder to distinguish in smaller marks.

Other Channels

• Transparency is a fourth available channel.
• It interacts strongly with luminance and saturation and should not be used in conjunction with those two.
• It can be used with hue for a small number of bins, often two.
• Most often used with superimposed layers.
  • Create a foreground layer distinguishable from the background.
• Frequently used redundantly.
  • Same information also in another channel.

Colormaps

Colormaps

• Colormaps specify the mapping between colors and data values.
  • AKA, pseudocoloring or color ramps.
• Mostly use a combination of hue, saturation, and luminance to encode data.
  • Rather than using the three independently as different channels.

Colormaps

• Categorical
• Ordered
  • Sequential
  • Diverging

Fig. 10.6

Colormaps

• Colormaps can be continuous or segmented.

Colormaps

• ColorBrewer is a good source for colormaps:

colorbrewer2.org.

• Easily accessible in R with RColorBrewer.

Colormaps

• viridis was created to have perceptual advantages.
• It was created for R and has since been ported to other places.
• Check out the viridis website.
• The viridis vignette is a good source of information about the various colormaps.

Other Useful R Packages

• colorspace
• nationalparkcolors
• paletteer (a comprehensive collection of other palettes)

Questions?

BCB5200 Home