Color Basics
Lecture 2
Gamma X rays Infrared Radar FM TV AMUltra-violet
10-12
10-8
10-4
104
1 108
electricityACShort-
wave
400nm 500nm 600nm 700nm
Wavelength
Wavelength in meters (m)
Visible light
Wavelength EncodingTrichromatic Color Theory
Color Matching Experiments
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Newton’s Experiment
1665, Cambridge University
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Electromagnetic Radiation -Spectrum
Gamma X rays Infrared Radar FM TV AMUltra-violet
10-12
10-8
10-4
104
1 108
electricityACShort-
wave
400 nm 500 nm 600 nm 700 nmWavelength in nanometers (1nm=10-9 m)
Wavelength in meters (m)
Visible light
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
The Spectral Power Distribution (SPD) of a light is a function f(l) which defines the energy at each wavelength.
Wavelength (λ)
400 500 600 7000
0.5
1
Re
lative
Po
wer
Spectral Power Distribution
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Monochromators
Monochromators measure the power or energy at different wavelengths
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Examples of Spectral power Distributions
Blue Skylight Tungsten bulb
Red monitor phosphor Monochromatic light
400 500 600 7000
0.5
1
400 500 600 7000
0.5
1
400 500 600 7000
0.5
1
400 500 600 7000
0.5
1
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Superposition of Light SPDs
wavelength
pow
er
pow
er
pow
er
wavelength
wavelength
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
• Low illumination levels (Scotopic vision).
• Highly sensitive (respond to a single photon).
• 100 million rods in each eye.
• No rods in fovea.
Retinal Photoreceptors
Rods -
Wavelength (nm)
Re
lative
sensiti
vity
400 500 600 7000
0.25
0.5
0.75
1
Rod Spectral Sensitivity
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Cones - • High illumination levels (Photopic vision)
• Less sensitive than rods.
• 5 million cones in each eye.
• Only cones in fovea (aprox. 50,000).
• Density decreases with distance from fovea.
• 3 cone types differing in their spectral
sensitivity: L , M, and S cones.
Retinal Photoreceptors
Wavelength (nm)
Re
lative
sensiti
vity
Cone Spectral Sensitivity
400 500 600 7000
0.25
0.5
0.75
1
ML
SM
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
L and M Cones -Density decreases with distance from fovea.None past 40 deg.
S Cones -None in the fovea (central 25’).Very sparse elsewhere.
Sensitivity to color decreases with distance from fovea in the order: green, red, yellow, blue.
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Color Sensitivity test – distance from fovea
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Color Deficiency
Trichromats - use 3 sensors
Dichromats - use 2 sensors (8% males, 0.05% females)protanopia - missing red conedeuteranopia - missing green conetritanopia - missing blue cone
Monochromats - use 1 sensor.
protanopia deuteranopia tritanopia
Dichromatic confusions:
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Color Deficiency
ProtanopeNormal
Deuteranope Tritanopia
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Ishihara Plates (1917).
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Reverse Ishihara Plates
Normal Should see Nothing.CVD should see 5
Normal Should see both 2 and 6Deutanopes should see 2 more easilyProtanopes should see 6 more easily
Normal Should see 3.CVD should see 5
Normal Should see 73.CVD should nothing.
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Wavelength (λ)
400 500 600 7000
0.5
1
Re
lative
Po
wer
Spectral Power Distribution
High dimensional data
RGB – 3 dimensional
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Trichromatic Color Theory
Thomas Young (1773-1829) -A few different retinal receptors operating with different wavelength sensitivities will allow humans to perceivethe number of colors that they do.Suggested 3 receptors.
Helmholtz & Maxwell (1850) -Color matching with 3 primaries.
Trichromatic: “tri”=three “chroma”=colorcolor vision is based on three primaries(i.e., it is 3 dimensional).
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Color Matching Experiment
Thomas Young 1802Helmholtz & Maxwell 1850Wright 1929Stiles & Burch 1959Judd & Wyszeki 1975
Metamer - two lights that appear the same visually. They might have different SPDs(spectral power distributions).
400 500 600 7000
400
800
400 500 600 7000
100
200
Wavelength (nm)
Pow
er
The phosphors of the monitor were set to match the tungsten light.
Tungsten light Monitor emission
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
+ -
+ -
+ -
test match
Three primary lights are set to match a test light.
=~
400 500 600 7000
0.25
0.5
0.75
1
400 500 600 7000
0.25
0.5
0.75
1
Test light Match light
Color Matching Experiment
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
If
and matches
matches
then matches
Color Matching Experimentis Linear
Homogeneity + additivity
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Color Matching Experimentis a Linear System
There exists a system matrix thatmaps test SPD to Match intensities.
400 500 600 700
Test Match
LinearSystem
C T = M
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Given a set of primaries, one can determine for every spectral wavelength, the intensity of the guns required to match a monochromatic light of that spectral wavelength.
Color Matching Functions
Monochromatic lights
400 500 600 7000
0.5
1
400 500 600 7000
0.5
1
400 500 600 7000
0.5
1
Primary Intensities
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
These values form the Color Matching Functions associated with the primaries.
400 500 600 7000
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Color Matching Functions
The intensity of the primaries required to match any spectra can then be determined by inner product of the spectra with the 3 color matching functions.
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Wavelength (λ)400 500 600 700
0
0.5
1
Wavelength (λ)400 500 600 700
0
0.5
1
Test Light = sum of monochromatic lights
=
Color Matching Functions
400 500 600 7000
0.2
0.4
0.6
0.8
1
Wavelength (nm)
ColorMatchingFunctions
TristimulusValues
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
)(3 λM
Wavelength (λ)
400 500 600 7000
0.5
1
Tristimulus Values = Inner product of SPD and CMF
Tristimulus Calculation
400 500 600 7000
0.2
0.4
0.6
0.8
1
dotproduct
λλλλ
dMte ∫= )()( 11
)(λt)(1 λM
)(2 λM
λλλλ
dMte ∫= )()( 22
λλλλ
dMte ∫= )()( 33
Wavelength (λ)
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Wavelength (λ)400 500 600 700
0
0.5
1
Rela
tive
Po
we
r
Light
Matrix Representation of the color matching system
Wavelength (λ)400 500 600 700
0
0.5
1
Rela
tive
Po
we
r
Wavelength (λ)
0
0.5
1
Se
nsi
tivity
400 500 600 700
Sensor
s
0.5
Wavelength (λ)
0
1
Se
nsi
tivity
400 500 600 700
I
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Wavelength (λ)400 500 600 700
0
0.5
Rela
tive
Po
we
r
s
Wavelength (λ)
0
0.5
Se
nsi
tivity
400 500 600 700
t
Sensor Response Calculation
e = <s,t> = sT t =
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
M1
t
Tristimulus Calculation
)(3 λM
Wavelength (λ)
400 500 600 7000
0.5
1
400 500 600 7000
0.2
0.4
0.6
0.8
1
dotproduct
)(λt)(1 λM
)(2 λM
Wavelength (λ)
M2
M3
e1
e2
e3
=
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
+ -
+ -
+ -
test match
Color Matching Experiment
Primaries
P1 P2 P3
Primaries
=
SpectraTristimulus
values
e1
e2
e3
t~
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Thus there must exist a 3 x nλ system matrix that mapsinput to output:
nλ
Color Matching is Linear
Color Matching is a linear system.i.e.
Color Matching defines a linear mapping from the test spd (nλ x 1 vector) to 3 primary intensities (3 x 1 vector).
M3
M2
M1
=
e1
e2
e3
t
SpectraTristimulus
values
e1
e2
e3
t
e = C t
nλ
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
[0 0 . . . 1 ]t
[1 0 . . . 0 ]t
Measure intensities of primaries for monochromatic test lights:
e1
en
. . .
.
λ
. . .
.
t2
t1 =
=
e are the columns of the color matching system matrix C.
i
C = [ e1
. . . .
enλ
]
Now, given the color matching functions, we
can calculate the response to any test light t :
e = C t
Calculating the system matrix
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
The rows of the system matrix are the color matching functions
with respect to the given primaries.
Color Matching Functions
M3
M2
M1
=C
400 500 600 7000
0.2
0.4
0.6
0.8
1
Wavelength (nm)
ColorMatchingFunctions
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
M3
M2
M1
=1 0 00 1 00 0 1
MatchingFunctions Primaries
Note an important relationship between Primaries and
their Matching Functions:
P1 P2 P3
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
400 500 600 7000
20
40
60
80
Wavelength (nm)
Every color can be represented by 3 values.
Space of visible colors is 3 Dimensional.
“tri”=three “chroma”=color
Trichromatic Color Theory
e1
e2
e3
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Stiles & Burch (1959) Color matching functions.Primaries are: 444.4 525.3 645.210 deg field..
Given the color matching functions, we can describeany light with 3 values (CIE-RGB):
r(λ)
g(λ)b(λ)
400 500 600 700
0
1
2
3
Wavelength (nm)
Prim
ary
Inte
nsity
(85, 38, 10) (21, 45, 72) (65, 54, 73)
Color Representation
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Caveat: For some matches e may be negative.i
t = e p + e p + e p1 1 22 3 3
~ e < 01
This does not make sense physically, however mathematically:
t = - |e1| p1 + e2 p2 + e3 p3~
t + |e1| p1 = e2 p2 + e3 p3~then
with all positive coefficients.
Physically this can be interpreted as adding
primary light p to the test : 1
+ -
+ -
+ -
test match
e.g.:
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
The Color Matching Experiment Predicts Metameric Matches
S and T match perceptually; they are metamers
400 500 600 7000
400
800
400 500 600 7000
100
200
Wavelength (nm)
Pow
er
Tungsten light Monitor emission
M3
M2
M1
t =M3
M2
M1
s
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Using different primary lights
1) Primary lights must be visually independent.
2) Uniqueness of the color matching functions:
p p p1 2 3
cmf C
or
This is true for all t.
primaries p p p` ` `1 2 3
C̀
given a test light t: e = C t e = C t``
denote P = [p1 p2 p3 ] P = [p p p ]1 2 3
` ` ` `
Since t is metameric to Pe and to Pe :` `
C t = C P e = C P e` `
e = C P e` `
t = P e t = P e` `we have ~ ~
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
test match
Primaries
test match
Primaries
test match
PrimariesPrimaries
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
C = (C P ) C` `
CP’ is a 3 x 3 matrix relating the two sets of color matching functions.
The color matching functions are unique up to a free 3 x 3 linear transformation.
i.e.
For monochromatic t:e are the columns of C,e are the columns of C.
e = C P e` ` for all t.
``
3 x 3 matrix
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Linear Transformation of The CMFsPredict the Same Metamers:
S and T are metamers
M3
M2
M1
t =M3
M2
M1
s
M3
M2
M1
t =M3
M2
M1
s LL
The color matching functions are uniqueup to a free linear transformation.
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
CMF vs Human Photoreceptors
Single Unit Cone Photocurrent Measurements(Schnapf, Baylor et al – 1987).
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Current Recordings From a Cone
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Cone Spectral Sensitivities
Wavelength (nm)
Re
lative
sensitiv
ity
Cone Spectral Sensitivity
400 500 600 7000
0.25
0.5
0.75
1
ML
SM
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
There are Three Types of Cone Wavelength Responsivity
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Stimuli Causing Equal Cone Signals Match Perceptually
L M S
L M S
Ab
sorp
tion
rat
e
Cone type
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Behavioral CMFs are accurately predicted by cone responsivities
B
R
G
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
The cone responsivities are a linear transformation from the CMFs
SML
=M3
M2
M1L
3x3 matrix
SensorResponsivities
Color MatchingFunctions
= L
3x3 matrix
SensorRespones
TristimulusValues
e1
e2
e3
L
M
S
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
400 500 600 7000
20
40
60
80
Wavelength (nm)
Every color can be represented by 3 values.
Space of visible colors is 3 Dimensional.
“tri”=three “chroma”=color
Trichromatic Color Theory
Color Representation ?
e1
e2
e3
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.
Color Matching Predicts Matches, Not Appearance
Albers (1975)
Generated by Foxit PDF Creator © Foxit Softwarehttp://www.foxitsoftware.com For evaluation only.