an empirical study of colour use - aucklandpaul/talks/colour.pdfan empirical study of colour use...
TRANSCRIPT
JJIIJI
Back
Close
An Empirical Study of Colour Use
Paul Murrell and Ross Ihaka
The University of AucklandNew Zealand
JJIIJI
Back
Close
Introduction
• A motivating example
• Why is it so hard to choose colours?
• Colour spaces
• Learning from the experts
JJIIJI
Back
Close
Introduction
• A motivating example
• Why is it so hard to choose colours?
• Colour spaces
• Learning from the experts
JJIIJI
Back
Close
Introduction
• A motivating example
• Why is it so hard to choose colours?
• Colour spaces
• Learning from the experts
JJIIJI
Back
Close
Introduction
• A motivating example
• Why is it so hard to choose colours?
• Colour spaces
• Learning from the experts
JJIIJI
Back
Close
Introduction
• A motivating example
• Why is it so hard to choose colours?
• Colour spaces
• Learning from the experts
JJIIJI
Back
Close
A motivating example
• Filling regions in barplots (or piecharts, or ...)
02
46
8
JJIIJI
Back
Close
A motivating example
• Filling regions in barplots (or piecharts, or ...)
02
46
8
JJIIJI
Back
Close
Why is it so hard to choose colours?
• Lack of natural talent
• Lack of knowledge about how colour works
• Lack of tools to work with colour
• Lack of knowledge about how to select colours
JJIIJI
Back
Close
Why is it so hard to choose colours?
• Lack of natural talent
• Lack of knowledge about how colour works
• Lack of tools to work with colour
• Lack of knowledge about how to select colours
JJIIJI
Back
Close
Why is it so hard to choose colours?
• Lack of natural talent
• Lack of knowledge about how colour works
• Lack of tools to work with colour
• Lack of knowledge about how to select colours
JJIIJI
Back
Close
Why is it so hard to choose colours?
• Lack of natural talent
• Lack of knowledge about how colour works
• Lack of tools to work with colour
• Lack of knowledge about how to select colours
JJIIJI
Back
Close
Why is it so hard to choose colours?
• Lack of natural talent
• Lack of knowledge about how colour works
• Lack of tools to work with colour
• Lack of knowledge about how to select colours
JJIIJI
Back
Close
Colour spacesThere are three main perceptual components to colour:
• hue (colour)
• lightness (light or dark)
• saturation (brightness, colourfulness)
JJIIJI
Back
Close
Colour spacesThere are three main perceptual components to colour:
• hue (colour)
• lightness (light or dark)
• saturation (brightness, colourfulness)
JJIIJI
Back
Close
Colour spacesThere are three main perceptual components to colour:
• hue (colour)
• lightness (light or dark)
• saturation (brightness, colourfulness)
JJIIJI
Back
Close
Colour spacesThere are three main perceptual components to colour:
• hue (colour)
• lightness (light or dark)
• saturation (brightness, colourfulness)
JJIIJI
Back
Close
Colour spaces• RGB colour space confounds hue, lightness, and sat-
uration.
● ●
●
●●
●●
●
JJIIJI
Back
Close
Colour spaces• HSV colour space directly addresses hue, lightness,
and saturation.
●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●
●
● ●●●
●● ●
JJIIJI
Back
Close
Colour spaces• CIE L*u*v* colour space directly addresses hue, light-
ness, and saturation AND attempts to make unit stepsperceptually uniform.
●
JJIIJI
Back
Close
Learning from the experts
• Interior designers select colours for large areas
• Interior design palettes are available on the internet
• Are there any obvious patterns to these palettes?
• View the palettes in CIE L*u*v* space
JJIIJI
Back
Close
Learning from the experts
• Interior designers select colours for large areas
• Interior design palettes are available on the internet
• Are there any obvious patterns to these palettes?
• View the palettes in CIE L*u*v* space
JJIIJI
Back
Close
Learning from the experts
• Interior designers select colours for large areas
• Interior design palettes are available on the internet
• Are there any obvious patterns to these palettes?
• View the palettes in CIE L*u*v* space
JJIIJI
Back
Close
Learning from the experts
• Interior designers select colours for large areas
• Interior design palettes are available on the internet
• Are there any obvious patterns to these palettes?
• View the palettes in CIE L*u*v* space
JJIIJI
Back
Close
Learning from the experts
• Interior designers select colours for large areas
• Interior design palettes are available on the internet
• Are there any obvious patterns to these palettes?
• View the palettes in CIE L*u*v* space
JJIIJI
Back
Close
Victorian Eclectic (“Home Decore” site)
Victorian Eclectic
JJIIJI
Back
Close
Victorian Eclectic
L
−20 0 20 40 60
●
●
●
●
●
●● ●
●●
●●
●
●
●
5060
7080
●
●
●
●
●
●● ●
●●
●●
●
●
●
−20
020
4060
●
●
● ●●
●
●
●
●
●
● ●● ●●
U●
●
● ●●
●
●
●
●
●
● ●●● ●
50 60 70 80
●● ●
●
●
●
●
●●
●
●
●
●●
●
● ●●
●
●
●
●
●●
●
●
●
●●●
−20 0 20 40 60
−20
020
4060
V
Victorian Eclectic
JJIIJI
Back
Close
Victorian Eclectic
1020
3040
5060
70
Victorian Eclectic Palette
Average Linkage Clustering
Hei
ght
●●
● ●● ● ● ●
● ●
●
● ● ● ●
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• L between 50 and 80
• U between -20 and 60
• V between -20 and 60
Modifications for barplots:
• Evenly spaced for “equal” difference
• Equal lightness for “equal impact”
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• L between 50 and 80
• U between -20 and 60
• V between -20 and 60
Modifications for barplots:
• Evenly spaced for “equal” difference
• Equal lightness for “equal impact”
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• L between 50 and 80
• U between -20 and 60
• V between -20 and 60
Modifications for barplots:
• Evenly spaced for “equal” difference
• Equal lightness for “equal impact”
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• L between 50 and 80
• U between -20 and 60
• V between -20 and 60
Modifications for barplots:
• Evenly spaced for “equal” difference
• Equal lightness for “equal impact”
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• L between 50 and 80
• U between -20 and 60
• V between -20 and 60
Modifications for barplots:
• Evenly spaced for “equal” difference
• Equal lightness for “equal impact”
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• L between 50 and 80
• U between -20 and 60
• V between -20 and 60
Modifications for barplots:
• Evenly spaced for “equal” difference
• Equal lightness for “equal impact”
JJIIJI
Back
Close
Applying the Expert Example
• Filling regions in barplots (or piecharts, or ...)
02
46
8
JJIIJI
Back
Close
Learning from the experts
• The EasyRGB web site.
JJIIJI
Back
Close
EasyRGB
L
−20 20 60 100
●
● ●● ●
● ●●●
●●● ●
4050
6070
8090
●
● ● ●●
● ●●●
●● ●●
−20
2060
100
●●
●
●
●
●
●●
●
●
●
●
● U
●●
●
●
●
●
●●
●
●
●
●
●
40 50 60 70 80 90
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
−80 −40 0 20 40
−80
−40
020
40
V
Dark Sea Green
JJIIJI
Back
Close
EasyRGB
−20 0 20 40 60 80 100
−80
−60
−40
−20
020
40
Dark Sea Green
U
V
●
●
●●
●
●
●
●
●
●
●
●
●
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• Halve or double the saturation
• Increase or decrease the luminance
• Generate complementary colours or triads
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• Halve or double the saturation
• Increase or decrease the luminance
• Generate complementary colours or triads
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• Halve or double the saturation
• Increase or decrease the luminance
• Generate complementary colours or triads
JJIIJI
Back
Close
Applying the Expert Example
Basic observations from the experts:
• Halve or double the saturation
• Increase or decrease the luminance
• Generate complementary colours or triads
JJIIJI
Back
Close
Applying the Expert Example
• Filling regions in barplots (or piecharts, or ...)
01
23
4
JJIIJI
Back
Close
Caveats
• This is just for barplots
• This is not for colour-blind
• This is not for grayscale printing
JJIIJI
Back
Close
Caveats
• This is just for barplots
• This is not for colour-blind
• This is not for grayscale printing
JJIIJI
Back
Close
Caveats
• This is just for barplots
• This is not for colour-blind
• This is not for grayscale printing
JJIIJI
Back
Close
Caveats
• This is just for barplots
• This is not for colour-blind
• This is not for grayscale printing
JJIIJI
Back
Close
Conclusions
If you collect colour palettes from the web ...
... and you work in the right colour space ...
... and you treat the palette as a data set ...
... and you observe simple patterns in the data ...
... you can generate simple colour palettes of your own(which don’t make you physically ill).
JJIIJI
Back
Close
Conclusions
If you collect colour palettes from the web ...
... and you work in the right colour space ...
... and you treat the palette as a data set ...
... and you observe simple patterns in the data ...
... you can generate simple colour palettes of your own(which don’t make you physically ill).
JJIIJI
Back
Close
Conclusions
If you collect colour palettes from the web ...
... and you work in the right colour space ...
... and you treat the palette as a data set ...
... and you observe simple patterns in the data ...
... you can generate simple colour palettes of your own(which don’t make you physically ill).
JJIIJI
Back
Close
Conclusions
If you collect colour palettes from the web ...
... and you work in the right colour space ...
... and you treat the palette as a data set ...
... and you observe simple patterns in the data ...
... you can generate simple colour palettes of your own(which don’t make you physically ill).
JJIIJI
Back
Close
Conclusions
If you collect colour palettes from the web ...
... and you work in the right colour space ...
... and you treat the palette as a data set ...
... and you observe simple patterns in the data ...
... you can generate simple colour palettes of your own(which don’t make you physically ill).
JJIIJI
Back
Close
Conclusions
If you collect colour palettes from the web ...
... and you work in the right colour space ...
... and you treat the palette as a data set ...
... and you observe simple patterns in the data ...
... you can generate simple colour palettes of your own(which don’t make you physically ill).
http://www.stat.auckland.ac.nz/ paul/