Chap 4 Lecture 1, 2

Fundamentals of Multimedia, Chapter
Chapter 4
Color in Im a g e and
Video
1. Color Science
2. Color Models in Images
3. Color Models in Video
4. Further Exploration
1 Li & Drew §c Prentice Hall

2003
4
4 . 1 C olor Science
L i g h t and S pe ctra
• L i g h t is an electromagnetic wave. Its color is characterized

by the wavelength content of the light.
(a) Laser light consists of a single wavelength: e.g., a ruby laser produces
a bright, scarlet-red beam.
(b) Mos t light sources produce contributions over many wavelengths.
(c) However, humans cannot detect all light, just contributions t h a t fall
in the “visible wavelengths”.
(d) Short wavelengths produce a blue sensation, long wavelengths pro-
duce a red one.
• S p e c t r o p h o t o m e t e r : device used t o measure visible light,

by reflecting light f r o m a diffraction grating ( a ruled surface)
t h a t spreads o u t the different wavelengths.
2003
4
• Figure 4.1 shows the phenomenon t h a t white light contains

all the colors of a rainbow.
Fig. 4.1: Sir Isaac Newton’s experiments.
• Visible light is an electromagnetic wave in the range 400 n m

t o 700 n m (where n m stands for nanometer, 10 − 9 meters).

2003
Fundamentals of Multimedia, Chapter 4
• Fig. 4.2 shows the relative power in each wavelength interval

for typical outdoor light on a sunny day. T h is type of curve
is called a Spectral Power Distribution ( S P D ) or a
spectrum .
• T h e symbol for wavelength is λ. T h i s curve is called

E(λ).
Spectral power
distribution
400 450 500 550 600 650 700

Wavelength (nm)
Fig. 4.2: Spectral power distribution of daylight.

2003
4
H u m a n V ision
• T h e eye works like a camera, w i t h the lens focusing an

image o n t o the retina (upside-down and left-right
reversed).
• T h e retina consists of an array of rods and three kinds of

cones.
• T h e rods come in t o play when light levels are low and

produce a image in shades of gray ( “ a l l cats are gray a t
night!”).
• For higher light levels, the cones each produce a signal. Be-
cause of their differing pigments, the three kinds of cones
are m o s t sensitive t o red ( R ) , green ( G ) , and blue ( B ) light.
• I t seems likely t h a t the brain makes use of differences R-G,

G-B, and B - R , as well as combining all of R, G, and B in to
a high-light-level achromatic channel. 2003
4
Sp ectral Sensitivit y of t he E ye
• T h e eye is m os t sensitive t o light in the middle of the visible

spectrum.
• T h e sensitivity of our receptors is also a function of wave-
length ( F ig . 4.3 below).
• T h e Blue receptor sensitivity is n o t shown t o scale because
i t is much smaller than the curves for Red or Green — Blue
is a late addition, in evolution.
– Statistically, Blue is the favorite color of humans, regardless of na-
tionality — perhaps for this reason: Blue is a latecomer and thus is a
bit surprising!
• Fig. 4.3 shows the overall sensitivity as a dashed line — this

im p o r t a n t curve is called the luminous-efficiency function.
– I t is usually denoted V ( λ ) and is formed as the sum of the response
curves for Red, Green, and Blue.

2003
• The ro d sensitivity curve lo oks like t he luminous-

efficiency function V ( λ ) b u t is shifted t o the red end of the
spectrum.
• The achromatic channel pro duced by the cones is

approxi- mately proportional t o 2R + G + B/20.
1.0
V
0.8
Relative response
0.6
G
0.4
R
0.2
B
0.0
400 450 500 550 600 650 700
Wavelength (nm)
Fig. 4.3: R,G, and B cones, and Luminous

Efficiency curve V ( λ ) .
2003
4
• These spectral sensitivity functions are usually denoted by

letters other than “ R , tt, B ” ; here let’s use a vector function
q ( λ ) , w i t h components
q (λ) = (q R ( λ ) , qG(λ), q B ( λ ) )T (4.1)
• T h e response in each color channel in the eye is proportional

t o the number of neurons firing.
• A laser light a t wavelength λ would result in a certain n u m-

ber of neurons firing. A n S P D is a combination of single-
frequency lights (like “ las e rs ” ), so we add up the cone re-
sponses for all wavelengths, weighted by the eye’s relative
response a t t h a t wavelength.

2003
4
• We can succinctly write dow n t his idea in t he for m of

an
integral:
∫
R = E(λ)
q R (λ) dλ
∫
tt = E ( λ ) q G (λ) ( 4 .2 )
∫
dλ B = E(λ)
q B (λ) dλ

2003
4
I m a g e F o rm a ti o n
• Surfaces reflect different amounts of light a t different wave-

lengths, and dark surfaces reflect less energy than light sur-
faces.
• Fig. 4.4 shows the surface spectral reflectance f r o m ( 1 )

orange sneakers and ( 2 ) faded bluejeans. T h e reflectance
function is denoted S ( λ ) .

2003
4
Surface spectral reflectance

0.1 0.2 0.3 0.4 0.5 0.6 Sneakers
Jeans
400 450 500 550 600 650 700

Wavelength (nm)
Fig. 4.4: Surface spectral reflectance functions S ( λ ) for objects.

2003
4
• Image formation is thus:
– L i g h t f r o m the illuminant w i t h S P D E ( λ ) impinges on a

surface, w i t h surface spectral reflectance function S ( λ ) ,
is reflected, and then is filtered by the eye’s cone
functions q ( λ ) .
– Reflection is shown in Fig. 4.5 below.
– T h e function C ( λ ) is called the color signal and consists

of the product of E ( λ ) , the illuminant, times S ( λ ) , the
reflectance:
C(λ) = E ( λ ) S(λ).

2003
4
Fig. 4.5: Image formation model.

2003
4
• T h e equations t h a t take in to account the image formation

model are:
∫
R = E ( λ ) S ( λ ) q R (λ)
dλ
∫
tt = E ( λ ) S ( λ ) q G (λ) ( 4 .3 )
∫
dλ B = E(λ) S(λ)
q B (λ) dλ

2003
4
C a m e r a Systems
• Camera systems are made in a similar fashion; a studio-

quality camera has three signals produced a t each pixel lo-
cation (corresponding t o a retinal position).
• Analog signals are converted t o digital, truncated t o

integers, and stored. If the precision used is 8-bit, then the
m a x imu m value for any of R, G, B is 255, and the m in i m u m
is 0.
• However, the light entering the eye of the computer user is

t h a t which is emitted by the screen — the screen is essentially
a self-luminous source. Therefore we need t o know the light
E ( λ ) entering the eye.

2003
4
G a m m a Correction
• T h e light emitted is in f act roughly proportional t o the volt-

age raised to a power; this power is called g a m m a , w i t h
symbol γ.
(a) T h u s , if the file value in the red channel is R, the screen
emits light proportional t o R γ , w i t h S P D equal t o t h a t of

the red phosphor paint on the screen t h a t is the target
of the red channel electron gun. T h e value of gamma is
around 2.2.
(b) I t is customary t o append a prime t o signals t h a t are
g a m m a - c o rr e c t e d by raising t o the power (1/γ) before
transmission. T h u s we arrive a t linear signals:
R → R j = R 1/γ ⇒ (R j ) γ → R (4.4)

2003
4
• Fig. 4. 6 (a ) shows light o u t p u t w i t h no gamma-correction

applied. We see t h a t darker values are displayed t o o dark.
T h is is also shown in Fig. 4.7(a), which displays a linear
ramp f r o m left t o right.
• Fig. 4 . 6 ( b ) shows the effect of pre-correcting signals by ap-

plying the power law R 1/γ ; i t is customary t o normalize volt-
age t o the range [0,1].

2003
4
No gamma correction Gamma correction
1.
1.
0
0
0.
0.
8
8
0. 0.
0. 0.
Light
Light
6
6
4
4
0.
0.
2
2
0.
0.
0
0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Voltage Voltage
Fig. 4.6: ( a ) : Effect of C R T on light emitted f r o m screen

(voltage is normalized t o range 0..1). ( b ) : Gamma
correction of signal.

2003
• T h e combined effect is shown in Fig. 4 . 7( b ). Here, a

ramp is shown in 16 steps f r o m gray-level 0 t o gray-level
255.
Fig. 4.7: ( a ) : Display of ramp f r o m 0 t o 255, w i t h no

gamma correction. ( b ) : Image w i t h gamma correction
applied
2003
4
C ol or-Matc hi ng Functions
• Even w i t h o u t knowing the eye-sensitivity curves of Fig.4.3, a

technique evolved in psychology for matching a
combination of basic R, G, and B lights t o a given shade.
• T h e particular set of three basic lights used in an experiment

are called the set of color primaries.
• To m a t c h a given color, a subject is asked t o separately adjust

the brightness of the three primaries using a set of controls
until the resulting spot of light m o s t closely matches the
desired color.
• T h e basic situation is shown in Fig.4.8. A device for carrying

o u t such an experiment is called a colorimeter.

2003
4
Red Room illumination

Green
Blue
Black
partition
White
Eye
screen
Light to be matched
Fig. 4.8: Colorimeter experiment.

2003
• T h e amounts of R, G, and B the subject selects t o m a t c h

each single-wavelength light forms the color-matching
curves. These are denoted r¯(λ), ¯g(λ), ¯b(λ) and are shown in
Fig. 4.9.
r¯(
b̄( )
)
Relative sensitivity
g¯ (
)
400 450 500 550 600 650 700

Wavelength (nm)
Fig. 4.9: CIE R G B color-matching functions ¯r(λ), ¯g(λ),

¯b(λ).

2003
4
C I E Chromaticity D i a g r a m
• Since the ¯r(λ) color-matching curve has a negative lobe, a

set of fictitious primaries were devised t h a t lead t o color-
matching functions w i t h only positives values.
(a) T he resulting curves are show n in Fig. 4.10;

these are usually referred t o as the color-matching
( b ) functions.
T h e y are a 3 × 3 m a trix away f r o m r¯, ¯g , b curves, and
are denoted x¯(λ), y¯(λ), z¯(λ).
( c ) T h e m atrix is chosen such t h a t the middle standard color-

matching function y¯(λ) exactly equals the luminous-
efficiency curve V ( λ ) shown in Fig. 4.3.

2003
4
Fig. 4.10: CIE standard X Y Z color-matching functions x¯(λ), y¯(λ), z¯(λ).

2003
4
• F or a general SP D E ( λ ) , the essential “ color i m e t r i c ”

information required t o characterize a color is
the set of tristimulus values X , Y , Z defined in analogy t o
( E q . 4.2)
∫
X = E ( λ ) x¯(λ)
dλ
∫
Y = E ( λ ) y¯(λ)
( 4 .6 )
∫
dλ Z = E(λ)
• 3 D data is difficult t o visualize, so the CIE devised a 2 D
diagram based on z¯(λ) dλ
the values of ( X , Y, Z ) triples implied by
the curves in Fig. 4.10.

2003
CIE Chromaticity Diagram. This is an international standard for primary colors
established in 1931. It allows all other colors to be defined as weighted sum of the
three "primary" colors. ... The CIE diagram is a plot of X vs. Y for all visible colors.
4
• We go t o 2 D by factoring .o u t the m agnitude of

vectors ( X , Y, Z ) ; we could divide2 by X 2+ Y 2 + Z , but
divide by the sum X + Y + Z t o make the chromaticity:
weinstead
x = X/(X + Y +
Z ) y = Y/(X + Y
+ Z ) z = Z/(X + ( 4.7 )
Y + Z)
• T h i s effectively means t h a t one value o u t of the set (x, y, z)
is redundant since we have
X + Y +
x+ y+ z = ≡ 1 ( 4.8 )
Z X + Y
+ Z
so t h a t
z = 1 − x − ( 4 .9 )
y
2003
4
• Effectively, we are projecting each tristimulus vector ( X , Y, Z )

o n t o the plane connecting points (1, 0, 0 ) , (0, 1, 0 ) , and (0, 0,
1).
• Fig. 4.11 shows the locus of points for monochromatic light
1.0
0.8
0.6
y
0.4
700
0.2
0.0
400
0.0 0.2 0.4 0.6 0.8 1.0
x
Fig. 4.11: CIE chromaticity diagram.

2003
4
(a) T h e color matching curves each add up t o the same

value
— t he area under each curve is the same f or
each of
x¯(λ), y¯(λ), z¯(λ).
(b) For an E ( λ ) = 1 for all λ, — an “equi-energy white l i g h t ”

— chromaticity values are (1/3, 1/3). Fig. 4.11
displays a typical actual white po in t in the middle of the
diagram.
(c) Since x, y ≤ 1 and x + y ≤ 1, all p ossible

ch r o m a t ic i ty values lie below the dashed diagonal line in
Fig. 4.11.

2003
4
• T h e CIE defines several “ w h i t e ” spectra: illuminant A ,

illu- minan t C, and standard daylights D 6 5 and D 100 .
(F ig . 4.12)
Spectral power distribution
D100 A
C
D65
F2
400 450 500 550 600 650 700

Wavelength (nm)
Fig. 4.12: Standard illuminant SPDs.

2003
4
• Chromaticities on the spectrum locus ( t h e “horseshoe” in

Fig. 4.11) represent “ p u r e ” colors. These are the m o st “ s a t -
u r a t e d ” . Colors close t o the white p o int are more unsatu-
rated.
• T h e chromaticity diagram: for a mixture of t w o lights, the

resulting chromaticity lies on the straight line joining the
chromaticities of the t w o lights.
• T h e “ d o m i n a n t wavelength” is the position on the spectrum

locus intersected by a line joining the white p oin t t o the
given color, and extended through i t .

2003
4
Color M onitor Sp ecifications
• Color monitors are specified in part by the white po in t chro-

maticity t h a t is desired if the RGB electron guns are all ac-
tivated a t their highest value (1.0, if we normalize t o [0,1]).
• T h ere are several monitor specifications in current use ( Ta -

ble 4.1).

2003
4
Table 4.1: Chromaticities and White Points of Monitor Specifications
Red Green Blue W h i t e P oint

System xr yr xg yg xb yb xW yW
NTSC 0.67 0.33 0.21 0.71 0.14 0.08 0.3101 0.3162
SMPTE 0.630 0.340 0.310 0.595 0.155 0.070 0.3127 0.3291
EBU 0.64 0.33 0.29 0.60 0.15 0.06 0.3127 0.3291
NTSC stands for National Television Standards Committee.
SMPTE stands for the Society of Motion Picture and Television Engineers
EBU stands for European Broadcasting Union

2003
4
Out-of-Gamut
C olors
• For any (x, y) pair we wish t o find t h a t RGB triple giving the
specified (x, y, z): We fo rm the z values for the phosphors,
via z = 1 − x − y and solve for RGB f r o m the phosphor
chromaticities.
• We combine nonzero values of R, G, and B via
x r xg xb R x
yr y b G = ( 4 .1 0 )
yg y B z
z r zg
zb

2003
4
• If (x, y) [color w i t h o u t magnitude] is specified, instead of

derived as above, we have t o invert the matrix of phosphor
(x, y, z) values t o obtain RGB .
• W h a t do we do if any of the RGB numbers is negative? —

t h a t color, visible t o humans, is o u t - o f - g a m u t for our display.
1. On e meth od: simply use the closest in-gamut color

available, as in Fig. 4.13.
2. Another approach: select the closest complementary

color.

2003
4
1.0
Green
0.8
0.6
y
(a)
0.4
(b)
Red
0.2
Blue
0.0
0.0 0.2 0.4 0.6 0.8 1.0

x
Fig. 4.13: Approximating an out-of-gamut
color by an in-gamut one.
T h e out-of-gamut color shown by a triangle is approximated by the in-
tersection of ( a ) the line from t h a t color t o the white point with ( b ) the
boundary of the device color gamut.
2003
4
• Additive color results f r o m self-luminous sources, such as

lights projected on a white screen, or the phosphors glowing
on the mon itor glass. (Subtractive color applies for printers,
and is very different).
• Fig. 4.13 above shows the triangular g a m u t for the N T S C

system, drawn on the CIE diagram — a monito r can display
only the colors inside a triangular g a m ut .

2003

Chap 4 Lecture 1, 2

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chap 4 Lecture 1, 2

Uploaded by

Copyright:

Available Formats

Fundamentals of Multimedia, Chapter

1 Li & Drew §c Prentice Hall

• L i g h t is an electromagnetic wave. Its color is characterized

• S p e c t r o p h o t o m e t e r : device used t o measure visible light,

• Figure 4.1 shows the phenomenon t h a t white light contains

Fig. 4.1: Sir Isaac Newton’s experiments.

• Visible light is an electromagnetic wave in the range 400 n m

3 Li & Drew §c Prentice Hall

• Fig. 4.2 shows the relative power in each wavelength interval

• T h e symbol for wavelength is λ. T h i s curve is called

400 450 500 550 600 650 700

4 Li & Drew §c Prentice Hall

• T h e eye works like a camera, w i t h the lens focusing an

• T h e retina consists of an array of rods and three kinds of

• T h e rods come in t o play when light levels are low and

• I t seems likely t h a t the brain makes use of differences R-G,

• T h e eye is m os t sensitive t o light in the middle of the visible

• Fig. 4.3 shows the overall sensitivity as a dashed line — this

6 Li & Drew §c Prentice Hall

• The ro d sensitivity curve lo oks like t he luminous-

• The achromatic channel pro duced by the cones is

400 450 500 550 600 650 700

Fig. 4.3: R,G, and B cones, and Luminous

• These spectral sensitivity functions are usually denoted by

q (λ) = (q R ( λ ) , qG(λ), q B ( λ ) )T (4.1)

• T h e response in each color channel in the eye is proportional

• A laser light a t wavelength λ would result in a certain n u m-

8 Li & Drew §c Prentice Hall

• We can succinctly write dow n t his idea in t he for m of

9 Li & Drew §c Prentice Hall

• Surfaces reflect different amounts of light a t different wave-

• Fig. 4.4 shows the surface spectral reflectance f r o m ( 1 )

10 Li & Drew §c Prentice Hall

Surface spectral reflectance

400 450 500 550 600 650 700

Fig. 4.4: Surface spectral reflectance functions S ( λ ) for objects.

11 Li & Drew §c Prentice Hall

• Image formation is thus:

– L i g h t f r o m the illuminant w i t h S P D E ( λ ) impinges on a

– Reflection is shown in Fig. 4.5 below.

– T h e function C ( λ ) is called the color signal and consists

12 Li & Drew §c Prentice Hall

Fig. 4.5: Image formation model.

13 Li & Drew §c Prentice Hall

• T h e equations t h a t take in to account the image formation

14 Li & Drew §c Prentice Hall

• Camera systems are made in a similar fashion; a studio-

• Analog signals are converted t o digital, truncated t o

• However, the light entering the eye of the computer user is

15 Li & Drew §c Prentice Hall

• T h e light emitted is in f act roughly proportional t o the volt-

(a) T h u s , if the file value in the red channel is R, the screen

emits light proportional t o R γ , w i t h S P D equal t o t h a t of

16 Li & Drew §c Prentice Hall

• Fig. 4. 6 (a ) shows light o u t p u t w i t h no gamma-correction

• Fig. 4 . 6 ( b ) shows the effect of pre-correcting signals by ap-

17 Li & Drew §c Prentice Hall

No gamma correction Gamma correction

Fig. 4.6: ( a ) : Effect of C R T on light emitted f r o m screen

18 Li & Drew §c Prentice Hall

• T h e combined effect is shown in Fig. 4 . 7( b ). Here, a

Fig. 4.7: ( a ) : Display of ramp f r o m 0 t o 255, w i t h no