Data Visualization

Chapter 06

Color

Khoa Công Nghệ Thông Tin

2
 Color spaces
•The color space of what colors the human visual
system can detect is three-dimensional
•Some of these color spaces for computer
manipulation, and the others are a better match
with the characteristic of human vision
•The most common color space is the RGB system
▪ Each color is represented by the weights assigned to
the red, green and blue channels.

3
 Color spaces
• HSL system is
• HSL system
▪ Requires HUE, Saturation, and Lightness axes
▪ Each color is defined by the weights assigned to the
hue, saturation and lightness axes
o Hue axis captures what we normally think of as pure color
➢ Not mixed with pure white and pure black
o Saturation axis is the amount of white mixed with a pure
color
➢ Pink = red + pure white
o Lightness axis is the amount of black mixed with a color
• HSV space is similar, where V is for grayscale and
is linear to lightness L

4
 Color spaces
•HSL space does not truly reflect how we perceive
color
▪ Lightness L is wildly different from how we perceive
luminance

5
 Color spaces
•The amount of luminance that human perceives
depends on the wavelength.
•Human are much more sensitive to middle
wavelengths of green and yellow than to the outer
wavelengths of red and blue

6
Luminance, Saturation and Hue Channels

• Luminance, saturation and hue can be used as a

7
Luminance, Saturation and Hue Channels

▪ Magnitude channel
o Luminance
➢ Suitable for ordered data
➢ Low-accuracy channel in terms of perceiving whether
noncontiguous regions have the same luminance because of
contrast effects
➢ Limit grayscale bins to no more than 4
o Saturation
▪ Identify channel
o HUE

8
Luminance, Saturation and Hue Channels

▪ Magnitude channel
o Luminance
o Saturation
➢ Suitable for ordered data
➢ Low-accuracy for noncontiguous regions
➢ The number of discriminate steps for saturation is low:
around 3 bins
➢ Interact with the size channel
➢ Use bright, highly saturated colors so that small regions are
more distinguishable
▪ Identify channel
o HUE

9
Luminance, Saturation and Hue Channels

▪ Identify channel
o HUE
➢ Very effective for categorical data and showing groups (highest
ranked after spatial position)
➢ HUE and size can interact: hue is harder to distinguish in small
regions
➢ Make fine distinctions in hue for contiguous regions
➢ Very limited discriminability between separate regions
➢ Hue does not have an implicit perceptual ordering

10
Transparency

•Transparency is a channel related to the other

11
Colormaps

12
 Colormaps
• Categorical colormaps
• Ordered colormaps
• Bivariate colormaps
• Colorblind-safe colormaps design

13
 Colormaps
•A colormap defines a mapping between colors and
data values (visual encoding with colors)
•Colormaps can be categorical or ordered
▪ Ordered colormaps can be sequential or diverging
•Colormaps can also be a continuous range of values
or segmented into discrete bins of colors
▪ Continuous colormaps are used for showing quantitative
attributes
▪ Segmented colormaps are suitable for categorical data
▪ For ordinal data
o Segmented colormaps will emphasize their discrete nature
o While continuous colormaps will emphasize their ordered
nature

14
Colormaps

15
 Categorical Colormaps
•Categorical colormap uses color to encode
categories and groupings
•Categorical colormaps are typically designed by
using color as an integral identity channel to
encode a single attribute, rather than to encode
three completely separate attributes with the three
channels of HUE, saturation and luminance.
•The number of discriminable colors for coding
small separated regions is limited to between 6 and
12 bins

16
 Categorical Colormaps
•A good set of initial choices are the fully saturated
and easily nameable colors, which are also the
opponent color axes: red, blue, green, and yellow.

17
 Categorical Colormaps
•Be careful with luminance contrast: for some
uses, the colors should be close in luminance to
avoid major differences in salience and to ensure
that all can be seen against the background.

18
 Categorical Colormaps
• Example:
▪ Fully saturated green and yellow will have much less
luminance contrast against a white background than
red and blue

19
 Categorical Colormaps
•For other uses, colors should be sufficiently different
in luminance that they can be distinguished even in
black and white
•Colormaps for
▪ Small regions such as lines should be highly saturated
▪ Large regions such as areas should have low saturation
•An appropriate colormap may depend on the mark
type being used.

20
Categorical Colormaps - Example

21
Categorical Colormaps - Example

22
Categorical Colormaps - Example
•The 21 colors used as an
index for each mouse
chromosome can indeed be
distinguished in large regions
next to each other
•These colors act as a legend
and an index because:
▪ Regions are large
▪ The most subtle differences
are between regions that are
right next to each other.

23
Categorical Colormaps - Example
•The regions of the human
chromosomes that
correspond to those in the
mouse chromosomes have
been colored to illustrate
how genomic regions have
moved around as the
species evolved
independently of each
other after diverging from
a common ancestor.

24
Categorical Colormaps - Example

25
Categorical Colormaps - Example
• Solve this problem
▪ Transform the data that take into account the nature
of the data and task, so that each bin can be encoded
with a distinguishable color
o Find a possible hierarchical structure that can be used to
derive meaningful aggregated groups
o Filter the attributes to only encode a small set of the most
important ones with color, and aggregate all the reset into a
new category
▪ Use a different encoding idiom that uses other visual
channels instead of, or in addition to, the color
channel alone

26
 Ordered Colormaps
• An ordered colormap is appropriate for encoding
ordinal or quantitative attributes
• The two major variants of continuous colormaps
for ordered data have expressiveness characteristics
that should match up with the attribute type.
▪ Sequential
▪ Diverging

27
 Ordered Colormaps
• A sequential colormap
▪ Ranges from a minimum value to a maximum value
▪ If only the luminance channel is used, the result is a
grayscale ramp
o When hue is added, one end of the map is a specific hue at full
saturation and brightness
▪ If saturation is a variable, the other end is pale or white
o If luminance is the varying quantity, the other end is dark or
black
• A diverging colormap
▪ has two hues at the endpoints and a neutral color as a
midpoint
▪ This neutral color may be white, gray, or black, or a high-
luminance color such as yellow.

28
 Ordered Colormaps
•The number of unique HUEs to be used in
continuous colormaps depends on what level of
structure should be emphasized:
▪ The high-level structure
▪ The middle range of local neighborhoods
▪ Fine-grained detail

29
 Ordered Colormaps
• Rainbow colormap

30
 Ordered Colormaps
• Blue - yellow colormap

31
 Ordered colormaps
•With a rainbow colormap, people can easily
discuss specific subranges because the differences
are easily nameable
•With a colormap only using saturation or
luminance changes, people can only talk about the
blue side and yellow side.
Blue Purple Red Yellow

32
 Ordered colormaps
•Rainbow colormaps, which are usually a default
choice in many software packages, suffer from
three serious problems.
1. Hue is used to indicate order, despite being an
identity channel
without an implicit perceptual ordering
2. The scale is not perceptually linear: steps of the
same size at different points in the colormap range
are not perceived equally by our eyes
Expressively mismatch
3. Fine details cannot be perceived with the hue
channel; the luminance channel would be a better
choice
Accuracy mismatch
33
 Ordered colormaps
•A range of 1000 units has different characteristics
depending on where within the colormap it falls
▪ The range [-2000, -1000] has three distinct colors (i.e.,
cyan, green, and yellow)
▪ The range of the same size [-1000, 0] looks
all yellow!

34
 Ordered colormaps
•A colormap that combines monotonically
increasing luminance with multiple hues for
semantic categories with a clear segmentation at
the 0 point
▪In doing so, it shows high-level, mid-level, and
low-level structures clearly.

35
 Ordered colormaps
•One way to address all three problems is to
design monotonically increasing luminance
colormaps
▪ More precisely, the multiple hues being used are
ordered according to their luminance from lowest to
highest
▪ The varying hues allow easy segmentation into
categorical regions
▪ Luminance is a magnitude channel, that provides
perceptual ordering.

36
 Ordered colormaps
•HUE is used to creating a semantically meaningful
categorization:
▪ 0 point matches with sea level
▪ dark blue sea
▪ cyan continental shelf
▪ green lowlands
▪ white mountains

37
 Ordered colormaps
•It is possible to create a perceptually linear
rainbow colormap, but at the cost of losing part of
the dynamic range because the fully saturated
colors are not available for use.

38
Ordered colormaps

39
 Bivariate Colormaps
•The safest use of the color channel is to visually
encode a single attribute: these colormaps are
known as univariate
•Colormaps that encode two separate attributes
are bivariate

40
 Bivariate Colormaps
•Fixing a base set of hues and varying their
saturation
▪ When both attributes are categorical with multiple
levels, results will be poor
▪ Combinations of two sequential or diverging
attributes with multiple levels is a middle ground.

41
 Colorblind-Safe Colormaps
•Color blindness is a sex-linked inherited trait that
affects 8% of males and 0.5% of females.
▪ In the common forms of color blindness, the ability to
sense along the red-green color axis is limited or
absent.
▪ Many pairs that are discriminable to people with
normal color vision are confused
o Including red-black, blue-purple, light green-white, and
brown-green.

42
 Colorblind-Safe Colormaps
•Avoid the use of only the HUE channel to encode
information
▪ In addition to HUE design, categorical colormaps vary
in luminance or saturation
•If possible, avoid colormaps that emphasize red-
green
▪ Especially in divergent red-green ramps
•Check
▪ [Link]

43
Q&A