2XWOLQH

&RORXU • Spectrum and colour

• Measuring colour
• The CIE XYZ colour space
• Colour spaces for computer graphics

1 2

:KDWLV/LJKW" /LJKWDV:DYHV
• When light flies through space and • Maxwell described the electromagnetic
interacts with itself, it behaves as a wave. spectrum and showed that visible light was
• When light interacts with matter, it just part of the spectrum.
behaves as a particle.
• The truth is more complicated than both of 102
these! 100 Radio waves
2
10 700 nm
Red
4
10 Orange
Infrared
6 Yellow
10
8
Visible Green
10
Ultraviolet Blue
10
10 Indigo
X-rays
10 12 Violet
400 nm

Wavelength (meters)

3 4
 /LJKWDV3DUWLFOHV :KDW:H'R:LWK/LJKW
• At any given moment, a light source emits The eyes and brain turn an incoming emission
some relative amount of photons at each spectrum into a discrete set of values.
frequency.
• We can plot the emission spectrum of a Mathematically, this is accomplished by
light source as power vs. wavelength. integrating the product of emission spectrum
with each of the three cone response curves.

The signal sent to our brain is somehow

interpreted as colour.

Emission spectra for daylight and a tungsten bulb (Wandell 4.4)

• A light source is characterized by its

emission spectrum. 5 6

0HDVXULQJWKH(PLVVLRQ
1HZWRQ·V([SHULPHQWV
6SHFWUXP
Newton showed that a prism spreads apart a light
source’s emission spectrum in space (why?).
– Light emerging from the prism cannot be
further decomposed.
– Newton called the colours of these “atomic”
lights primaries.
– We call one-colour light monochromatic.

A spectroradiometer
Newton’s experimental setup (Wandell 4.1)
7 8
 (PLVVLRQ6SHFWUXPLV
&RORXU0DWFKLQJ
QRW&RORXU
• Recall how much averaging the eye does. Conjecture: every colour can be uniquely
Light is infinite dimensional! expressed as a mixing of a small number of
primaries. (Why is this plausible?)
• Different light sources can evoke exactly
the same colours. Such lights are called If true, this gives us a meaningful definition of
“metamers”. colour as a set of primaries and the range of
possible combinations between them.
Given a choice of primaries, how can we
verify the conjecture?

A dim tungsten bulb and an RGB monitor set up to emit a

metameric spectrum (Wandell 4.11)
9 10

7KH&RORXU0DWFKLQJ &RORXU0DWFKLQJ
([SHULPHQW )XQFWLRQV
Subject attempts to match a combination of A colour matching experiment yields a colour
primaries with a test light. matching function for each primary.

Experimental setup for colour matching (Wandell 4.10)

It turns out that three primaries suffice to Colour matching functions for primaries at 460, 530
and 650 nm (Wasserman 3.3)
produce all perceivable colours.
How can we use the matching functions to
The experiment tells us a lot more about
match an arbitrary monochromatic light?
colour perception…
11 What about an arbitrary light? 12
 &KRRVLQJ3ULPDULHV &RORXUHG6XUIDFHV
So far, we’ve discussed the colours of lights.
How do surfaces acquire colour?

Emission spectra for RGB monitor phosphors (Wandell B.3)

Primaries don’t have to be monochromatic.

You can still derive colour matching functions. Subtractive colour mixing (Wasserman 2.2)

It turns out that colour matching functions are A surface’s reflectance is its tendency to reflect
always linearly related! Why? incoming light across the spectrum.

Reflectance is combined subtractively with

13 incoming light. 14

,OOXVWUDWLRQRI&RORXU
6XEWUDFWLYH0HWDPHUV
$SSHDUDQFH

Surfaces that are metamers under only some lighting conditions

(Wasserman 3.9)

Reflectance adds a whole new dimension of

complexity to colour perception.
How light and reflectance become cone responses (Wandell, 9.2)

The solid curve appears green indoors and out.

The dashed curve looks green outdoors, but
brown under incandescent light.
15 16
 7KH&,(;<=6\VWHP &,(&RRUGLQDWHV
A standard created in 1931 by CIE, defined in Given an emission spectrum, we can use the
terms of three colour matching functions. CIE matching functions to obtain the X, Y and Z
coordinates.

Then we can compute chromaticity coordinates.

This gives a brightness independent notion of
colour.

The XYZ colour matching functions (Wasserman 3.8)

What are the primaries?

Why were these colour matching functions

chosen? 17 18

7KH&,(&KURPDWLFLW\
7KH&,(&RORXU%ORE
'LDJUDP

The chromaticity diagram (Wasserman 3.7)

Different views of the CIE colour space (Foley II.1)
• A projection of the plane X+Y+Z=1.
• Each point is a chromaticity value, which
depends on dominant wavelength, or hue,
and excitation purity, or saturation.
19 20
 0RUH$ERXW&KURPDWLFLW\ *DPXWV
Not every output device can reproduce every
• Dominant wavelengths go around the
colour. A device’s range of reproducible colours
perimeter of the chromaticity blob.
is called its gamut.
– A colour’s dominant wavelength is where a line
from white through that colour intersects the
perimeter.
– Some colours, called nonspectral colours, don’t
have a dominant wavelength.
• Excitation purity is measured in terms of a
colour’s position on the line to its
dominant wavelength.
• Complementary colours lie on opposite
sides of white, and can be mixed to get
white.

Gamuts of a few common output devices in CIE space (Foley, II.2)

21 22

3HUFHSWXDO
QRQ 8QLIRUPLW\
The XYZ colour space is not perceptually
uniform!

Enlarged ellipses of constant colour in XYZ space

(Wasserman 3.10)

Some modified spaces attempt to fix this:

•L*u*v*
•L*a*b*
23 24
 &RORXU6SDFHVIRU
5*%
&RPSXWHU*UDSKLFV
In practice, there’s a set of more commonly-used Perhaps the most familiar colour space, and the
colour spaces in computer graphics: most convenient for display on a CRT (why?)
• RGB for display
• HSV for user selection What does the RGB colour space look like?
• CMY (or CMYK) for hardcopy
• YIQ for television broadcast

25 26

+69 &0<

More natural for user interaction, corresponds to A subtractive colour space used for printing.
the artistic concepts of tint, shade and tone.
Involves three subtractive primaries:
The HSV space looks like a “hexcone”: • Cyan - subtracts red
• Magenta - subtracts green
• Yellow - subtracts blue
Mixing two pigments subtracts their opposites
from white.

CMYK adds blacK ink rather than using equal

amounts of all three.

27 28
 5*%YV&0< <,4

Used in TV broadcasting, YIQ exploits useful

properties of the visual system.
• Y - luminance (taken from CIE)
• I - major axis of remaining colour space
• Q - remaining axis
YIQ is broadcast with relative bandwidth ratios
8:3:1
• We’re best as distinguishing changes in
luminance.
• Small objects can be compressed into a
single colour dimension.
Why do we devote a channel to luminance?

29 30

6XPPDU\ 5HDGLQJ

Lights are characterized by emission spectra; surfaces have Suggested:

reflectance spectra.
• Angel, section 1.3
Emission spectrum is not colour. Our visual systems integrate and • Foley et al., sections 13.2 - 13.6
average emission spectra into cone responses, which are • Glassner, sections 1.7, 2.1, 2.2, 3.6
interpreted as colour.

Many different spectra can have the same colour. Such spectra are Further reading:
called metamers. • Gerald S. Wasserman. Color Vision: An Historical
Introduction. John Wiley & Sons, New York, 1978
The CIE XYZ system is a standardized colour space defined in
terms of three matching functions. The chromaticity diagram, • Brian Wandell. Foundations of Vision. Sinauer
derived from the XYZ space, gives a useful interpretation of colour. Associates, Sunderland, MA, 1995.

A gamut is the range of colours a device can reproduce.

There are several important colour spaces commonly used in

computer graphics and broadcast.

31 32