Causes of color

The sensation of color is caused Light could be produced in

by the brain. different amounts at different
wavelengths (compare the sun and
Some ways to get this sensation a fluorescent light bulb).
include:
Light could be differentially
Pressure on the eyelids reflected (e.g. some pigments).
Dreaming, hallucinations, etc. It could be differentially refracted -
Main way to get it is the (e.g. Newtons prism)
response of the visual system to Wavelength dependent specular
the presence/absence of light at reflection - e.g. shiny copper penny
various wavelengths. (actually most metals).
Flourescence - light at invisible
wavelengths is absorbed and
reemitted at visible wavelengths.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Radiometry for colour

All definitions are now per unit wavelength

All units are now per unit wavelength
All terms are now spectral
Radiance becomes spectral radiance
watts per square meter per steradian per unit wavelength
Radiosity --- spectral radiosity

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Black body radiators

Construct a hot body with near-zero albedo (black body)

Easiest way to do this is to build a hollow metal object with a tiny
hole in it, and look at the hole.
The spectral power distribution of light leaving this object
is a simple function of temperature
1 1
E 5
exp hc kT 1

This leads to the notion of color temperature --- the
temperature of a black body that would look the same

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Measurements of
relative spectral power
of sunlight, made by J.
Parkkinen and P.
Silfsten. Relative
spectral power is plotted
against wavelength in
nm. The visible range is
about 400nm to 700nm.
The color names on the
horizontal axis give the
color names used for
monochromatic light of
the corresponding
wavelength --- the
colors of the rainbow.
Violet Indigo Blue Green Yellow Orange Red
Mnemonic is Richard
of York got blisters in
Computer Vision - A Modern Approach
Venice.
Set: Color
Slides by D.A. Forsyth
 Relative spectral power
of two standard
illuminant models ---
D65 models sunlight,and
illuminant A models
incandescent lamps.
Relative spectral power
is plotted against
wavelength in nm. The
visible range is about
400nm to 700nm. The
color names on the
horizontal axis give the
color names used for
monochromatic light of
the corresponding
Violet Indigo Blue Green Yellow Orange Red
wavelength --- the
colors of the rainbow.
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Measurements of
relative spectral power
of four different artificial
illuminants, made by
H.Sugiura. Relative
spectral power is plotted
against wavelength in
nm. The visible range is
about 400nm to 700nm.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Spectral albedoes for
several different leaves,
with color names
attached. Notice that
different colours
typically have different
spectral albedo, but that
different spectral
albedoes may result in
the same perceived
color (compare the two
whites). Spectral
albedoes are typically
quite smooth functions.
Measurements by
E.Koivisto.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 The appearance of colors

Color appearance is strongly Film color mode:

affected by (at least): View a colored surface
other nearby colors, through a hole in a sheet, so that
adaptation to previous views the colour looks like a film in
state of mind space; controls for nearby colors,
We show several demonstrations in and state of mind.
what follows. Other modes:
Surface colour
Volume colour
Mirror colour
Illuminant colour

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 The appearance of colors

Hering, Helmholtz: Color appearance is

strongly affected by other nearby
By experience, it is possible to
colors, by adaptation to previous views,
and by state of mind
match almost all colors, viewed in
film mode using only three primary
Film color mode: View a
colored surface through a hole in a sources - the principle of
sheet, so that the colour looks like a trichromacy.
film in space; controls for nearby Other modes may have more
colors, and state of mind. dimensions
Other modes: Glossy-matte
Surface colour Rough-smooth
Volume colour
Most of what follows discusses
Mirror colour
Illuminant colour
film mode.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Why specify color numerically?

Accurate color reproduction is Color reproduction problems

commercially valuable increased by prevalence of digital
Many products are identified by imaging - eg. digital libraries of art.
color (golden arches; How do we ensure that everyone
Few color names are widely sees the same color?
recognized by English speakers -
About 10; other languages have
fewer/more, but not many more.
Its common to disagree on
appropriate color names.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Color matching experiments - I

Show a split field to subjects; one side shows the light

whose color one wants to measure, the other a weighted
mixture of primaries (fixed lights).
Each light is seen in film color mode.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Color matching experiments - II

Many colors can be represented as a mixture of A, B, C

write
M=a A + b B + c C
where the = sign should be read as matches
This is additive matching.
Gives a color description system - two people who agree
on A, B, C need only supply (a, b, c) to describe a color.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Subtractive matching

Some colors cant be matched like this:

instead, must write
M+a A = b B+c C
This is subtractive matching.
Interpret this as (-a, b, c)
Problem for building monitors: Choose R, G, B such
that positive linear combinations match a large set of
colors

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 The principle of trichromacy

Experimental facts:
Three primaries will work for most people if we allow subtractive
matching
Exceptional people can match with two or only one primary.
This could be caused by a variety of deficiencies.
Most people make the same matches.
There are some anomalous trichromats, who use three
primaries but make different combinations to match.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Grassmans Laws

For colour matches made in film colour mode:

symmetry: U=V <=>V=U
transitivity: U=V and V=W => U=W
proportionality: U=V <=> tU=tV
additivity: if any two (or more) of the statements
U=V,
W=X,
(U+W)=(V+X) are true, then so is the third
These statements are as true as any biological law. They
mean that color matching in film color mode is linear.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Linear color spaces

A choice of primaries yields a RGB: primaries are

linear color space --- the monochromatic energies are
coordinates of a color are given 645.2nm, 526.3nm, 444.4nm.
by the weights of the primaries CIE XYZ: Primaries are
used to match it. imaginary, but have other
Choice of primaries is convenient properties. Color
equivalent to choice of color coordinates are (X,Y,Z), where
space. X is the amount of the X
primary, etc.
Usually draw x, y, where
x=X/(X+Y+Z)
y=Y/(X+Y+Z)

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Color matching functions
Choose primaries, say A, B, C
Given energy function, E( ) Then our match is:
what amounts of primaries will

a( )E( )d A
match it?
For each wavelength, determine
how much of A, of B, and of C
is needed to match light of that
wavelength alone. a( )
b( )E( )dB
b( ) c( )E( )dC
c( )

These are colormatching

functions

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 RGB: primaries are
monochromatic, energies are
645.2nm, 526.3nm, 444.4nm.
Color matching functions have
negative parts -> some colors
can be matched only
subtractively.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 CIE XYZ: Color
matching functions are
positive everywhere, but
primaries are imaginary.
Usually draw x, y, where
x=X/(X+Y+Z)
y=Y/(X+Y+Z)

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 A qualitative rendering of
the CIE (x,y) space. The
blobby region represents
visible colors. There are
sets of (x, y) coordinates
that dont represent real
colors, because the
primaries are not real lights
(so that the color matching
functions could be positive
everywhere).

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 A plot of the CIE (x,y)
space. We show the
spectral locus (the colors
of monochromatic
lights) and the black-
body locus (the colors of
heated black-bodies). I
have also plotted the
range of typical
incandescent lighting.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Non-linear colour spaces

HSV: Hue, Saturation, Value are non-linear functions of

XYZ.
because hue relations are naturally expressed in a circle

Uniform: equal (small!) steps give the same perceived

color changes.

Munsell: describes surfaces, rather than lights - less

relevant for graphics. Surfaces must be viewed under
fixed comparison light
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
HSV hexcone

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Uniform color spaces

McAdam ellipses (next slide) demonstrate that differences

in x,y are a poor guide to differences in color
Construct color spaces so that differences in coordinates
are a good guide to differences in color.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
Variations in color matches on a CIE x, y space. At the center of the ellipse is the color of a
test light; the size of the ellipse represents the scatter of lights that the human observers tested
would match to the test color; the boundary shows where the just noticeable difference is.
The ellipses on the left have been magnified 10x for clarity; on the right they are plotted to
scale. The ellipses are known as MacAdam ellipses after their inventor. The ellipses at the top
are larger than those at the bottom of the figure, and that they rotate as they move up. This
means that the magnitude of the difference in x, y coordinates is a poor guide to the
difference in color. Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
CIE uv which is a
projective transform
of x, y. We transform
x,y so that ellipses are
most like one another.
Figure shows the
transformed ellipses.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Color receptors and color deficiency

Trichromacy is justified - in Some people have fewer than

color normal people, there are three types of receptor; most
three types of color receptor, common deficiency is red-green
called cones, which vary in color blindness in men.
their sensitivity to light at Color deficiency is less
different wavelengths (shown common in women; red and
by molecular biologists). green receptor genes are carried
Deficiency can be caused by on the X chromosome, and
CNS, by optical problems in the these are the ones that typically
eye, or by absent receptor types go wrong. Women need two
Usually a result of absent bad X chromosomes to have a
genes. deficiency, and this is less
likely.
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Color receptors

Principle of univariance: All experimental evidence

cones give the same kind of suggests that the response of the
response, in different amounts, kth type of cone can be written
to different wavelengths. The as
output of the cone is obtained
by summing over wavelengths. ( )E( )d
k
Responses are measured in a
variety of ways (comparing
behaviour of color normal and where k ( ) is the sensitivity
color deficient subjects). of the receptor and spectral
energy density of the incoming
light.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
Color receptors

Plot shows relative sensitivity

as a function of wavelength, for
the three cones. The S (for
short) cone responds most
strongly at short wavelengths;
the M (for medium) at medium
wavelengths and the L (for
long) at long wavelengths.
These are occasionally called B,
G and R cones respectively, but
thats misleading - you dont
see red because your R cone is
activated.
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Adaptation phenomena

The response of your color Common example: walk inside

system depends both on spatial from a bright day; everything
contrast and what it has seen looks dark for a bit, then takes
before (adaptation) its conventional brightness.
This seems to be a result of
coding constraints --- receptors
appear to have an operating
point that varies slowly over
time, and to signal some sort of
offset. One form of adaptation
involves changing this
operating point.

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Viewing coloured objects

Assume diffuse+specular model Diffuse

colour of reflected light
depends on both illuminant and
Specular
surface
specularities on dielectric
people are surprisingly good at
objects take the colour of the
disentangling these effects in
light
practice (colour constancy)
specularities on metals can be
this is probably where some of
coloured
the spatial phenomena in
colour perception come from

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 When one views a colored
surface, the spectral
radiance of the light
reaching the eye depends
on both the spectral
radiance of the illuminant,
and on the spectral albedo
of the surface. Were
assuming that camera
receptors are linear, like
the receptors in the eye.
Computer Vision - A Modern Approach This is usually the case.
Set: Color
Slides by D.A. Forsyth
 Subtractive mixing of inks

Inks subtract light from white, eg. C+M+Y=White-

whereas phosphors glow. White=Black
Linearity depends on pigment C+M=White-Yellow=Blue
properties Usually require CMY and
inks, paints, often hugely non- Black, because colored inks are
linear. more expensive, and
Inks: Cyan=White-Red, registration is hard
Magenta=White-Green, For good choice of inks, there is
Yellow=White-Blue. a linear transform between
For a good choice of inks, and XYZ and CMY
good registration, matching is
linear and easy

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Finding Specularities

Assume we are dealing with dielectrics

specularly reflected light is the same color as the source
Reflected light has two components
diffuse
specular
and we see a weighted sum of these two
Specularities produce a characteristic dogleg in the
histogram of receptor responses
in a patch of diffuse surface, we see a color multiplied by different
scaling constants (surface orientation)
in the specular patch, a new color is added; a dog-leg results

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 B
S

Illuminant color

Diffuse component

R
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Boundary of
Diffuse specularity
region
B B

G G

R R

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Color constancy
Assume weve identified and removed specularities
The spectral radiance at the camera depends on two things
surface albedo
illuminant spectral radiance
the effect is much more pronounced than most people think (see following
slides)
We would like an illuminant invariant description of the surface
e.g. some measurements of surface albedo
need a model of the interactions
Multiple types of report
The colour of paint I would use is
The colour of the surface is
The colour of the light is

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Notice how the
color of light at
the camera varies
with the illuminant
color; here we have
a uniform reflectance
illuminated by five
different lights, and
the result plotted on
CIE x,y

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Notice how the
color of light at
the camera varies
with the illuminant
color; here we have
the blue flower
illuminated by five
different lights, and
the result plotted on
CIE x,y. Notice how it
looks significantly more
saturated under some
lights.
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Notice how the
color of light at
the camera varies
with the illuminant
color; here we have
a green leaf
illuminated by five
different lights, and
the result plotted on
CIE x,y

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Lands Demonstration

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Lightness Constancy
Lightness constancy
how light is the surface, independent of the brightness of the illuminant
issues
spatial variation in illumination
absolute standard
Human lightness constancy is very good
Assume
frontal 1D Surface
slowly varying illumination
quickly varying surface reflectance

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 Lightness Constancy in 2D

Differentiation, thresholding are This yields a minimization

easy problem
integration isnt How do we choose the constant
problem - gradient field may of integration?
no longer be a gradient field average lightness is grey
One solution lightest object is white
Choose the function whose ?
gradient is most like
thresholded gradient

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Simplest colour constancy

Adjust three receptor channels independently

Von Kries
Where does the constant come from?
White patch
Averages
Some other known reference (faces, nose)

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Colour Constancy - I

We need a model of interaction Finite Dimensional Linear

between illumination and Model (or FDLM)
surface colour surface spectral albedo is a
finite dimensional linear model weighted sum of basis
seems OK functions
illuminant spectral exitance is
a weighted sum of basis
functions
This gives a quite simple form
to interaction between the two

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Finite Dimensional Linear Models

m
E ii
i1
m n
pk k ii
rj j d
i1 j1
m,n
r d
i j k i j
i1, j1
m,n

n
r g i j ijk

rj j i1, j1

j1
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth
 General strategies

Determine what image would We need some form of

look like under white light reference
Assume brightest patch is white
that we are dealing with flat spatial average is known
frontal surfaces gamut is known
Weve identified and removed specularities
specularities
no variation in illumination

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Obtaining the illuminant from
specularities
Assume that a specularity has Assuming
been identified, and material is we know the sensitivities and
dielectric. the illuminant basis functions
Then in the specularity, we have there are no more illuminant
basis functions than receptors
This linear system yields the
pk k E d illuminant coefficients.
m
i k i d
i1

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Obtaining the illuminant from average
color assumptions
Assume the spatial average Assuming
reflectance is known g_ijk are known
n average reflectance is known
r j j there are not more receptor
j1 types than illuminant basis
functions
We can recover the illuminant
We can measure the spatial coefficients from this linear
average of the receptor response system
to get
m,n
pk r g i j ijk
i1, j1

Computer Vision - A Modern Approach

Set: Color
Slides by D.A. Forsyth
 Computing surface properties

Two strategies to get appearance under white

compute reflectance light, plug in reflectance
coefficients coefficients and compute
compute appearance under
white light. m,n

These are essentially pk white

i
rj gijk
equivalent. i1, j1

Once illuminant coefficients are

known, to get reflectance
coefficients we solve the linear
system m,n
pk r g i j ijk
i1, j1
Computer Vision - A Modern Approach
Set: Color
Slides by D.A. Forsyth