Raster Graphics and Color
Aaron Bloomfield
CS 445: Introduction to Graphics
Fall 2006
Overview

Display hardware


Raster graphics systems


How are imaging systems organized?
Color models


How are images displayed?
How can we describe and represent colors?
All non-credited images in this slide
set are from Wikipedia
2
Overview

Display hardware


Raster graphics systems


How are images displayed?
How are imaging systems organized?
Color models

How can we describe and represent colors?
3
Display Hardware

Video display devices






Cathode Ray Tube (CRT)
Liquid Crystal Display (LCD)
Plasma panels
Thin-film electroluminescent displays
Light-emitting diodes (LED)
Hard-copy devices





Ink-jet printer
Laser printer
Film recorder
Electrostatic printer
Pen plotter
4
Cathode Ray Tube (CRT)
1.
2.
3.
4.
Electron guns
Electron beams
Focusing coils
Deflection coils
5.
6.
7.
8.

Image via Wikipedia: http://en.wikipedia.org/wiki/Cathode_ray_tube
Anode
connection
Mask for
separating
beams for
RGB part of
displayed
image
Phosphor
layer with
RGB zones
Close-up of
the phosphor-coated
inner side
of the
screen
5
Liquid Crystal Display (LCD)
Figure 2.16 from H&B6
Display Hardware

Video display devices
»
»




Cathode Ray Tube (CRT)
Liquid Crystal Display (LCD)
Plasma panels
Thin-film electroluminescent displays
Light-emitting diodes (LED)
Hard-copy devices





Ink-jet printer
Laser printer
Film recorder
Electrostatic printer
Pen plotter
7
Overview

Display hardware


Raster graphics systems


How are images displayed?
How are imaging systems organized?
Color models

How can we describe and represent colors?
8
Raster Graphics Systems
I/O Devices
System Bus
Display
Processor
CPU
System
Memory
Frame
Buffer
Video
Controller
Monitor
Figure 2.29 from H&B9
Frame Buffer
Frame Buffer
Figure 1.2 from FvDFH10
Frame Buffer Refresh
Refresh rate is usually 60-120 Hz for CRTs
Figure 1.3 from FvDFH11
Direct Color Framebuffer


Store the actual intensities of R, G, and B
individually in the framebuffer
24 bits per pixel = 8 bits red, 8 bits green, 8 bits
blue
DAC
12
Red component vs. monochromatic

The red component only
has the red components
of each pixel (duh!)

Monochromatic is a grayscale image that uses
another color instead of white
13
Color Lookup Framebuffer


Store indices (usually 8 bits) in framebuffer
Display controller looks up the R,G,B values
before triggering the electron guns
DAC
Color indices
14
Color CRT
Figure 2.8 from H&B15
Overview

Display hardware


Raster graphics systems

»
How are images displayed?
How are imaging systems organized?
Color models

How can we describe and represent colors?
16
Specifying Color

Color perception usually involves three quantities:





Hue: Distinguishes between colors like red, green, blue,
etc
Saturation: How far the color is from a gray of equal
intensity
Lightness: The perceived intensity of a reflecting object
Sometimes lightness is called brightness if the
object is emitting light instead of reflecting it.
In order to use color precisely in computer
graphics, we need to be able to specify and
measure colors.
17
How Do Artists Do It?




Artists often specify color as tints, shades, and tones of
saturated (pure) pigments
Tint: Adding white to a pure pigment, decreasing saturation
Shade: Adding black to a pure pigment, decreasing
lightness
Tone: Adding white and black to a pure pigment
White
Tints
Pure Color
Tones
Grays
Black
Shades
18
Additive color vs. Subtractive color

Additive colors models are used in light


Subtractive color models are used with paint



Start with black, and add colored light to make your desired shade
Start with white, and add colors
A given color – red – subtracts away (from the reflected light) any
wavelength that is not red
Additive color mixing:

Subtractive color mixing:
19
HSV Color Model
H
0
120
240
*
*
*
60
270
270
S
1.0
1.0
1.0
0.0
0.0
*
1.0
0.5
0.0
V
1.0
1.0
1.0
1.0
0.5
0.0
1.0
1.0
0.7
Color
Red
Green
Blue
White
Gray
Black
?
?
?
Figure 15.16&15.17 from H&B
20
Intuitive Color Spaces





HSV is an
intuitive color
space
Corresponds
to our perceptual
notions of tint, shade,
and tone
Hue (H) is the angle
around the vertical axis
Saturation (S) is a value from
0 to 1 indicating how far from
the vertical axis the color lies
Value (V) is the height of the “hexcone”
21
Precise Color Specifications






Pigment-mixing is subjective --- depends on human
observer, surrounding colors, lighting of the environment,
etc
We need an objective color specification
Light is electromagnetic energy in the 400 to 700 nm
wavelength range
Dominant wavelength is the wavelength of the color we
“see”
Excitation purity is the proportion of pure colored light to
white light
Luminance is the amount (or intensity) of the light
22
Electromagnetic Spectrum

Visible light frequencies range between ...


Red = 4.3 x 1014 hertz (700nm)
Violet = 7.5 x 1014 hertz (400nm)
Figures 15.1 from H&B 23
Visible Light



Hue = dominant frequency (highest peak)
Saturation = excitation purity (ratio of highest to rest)
Lightness = luminance (area under curve)
White Light
Orange Light
Figures 15.3-4 from H&B
24
Color Matching

In order to match a color, we can adjust the
brightness of 3 overlapping primaries until the two
colors look the same.


C = color to be matched
RGB = laser sources (R=700nm, G=546nm, B=435nm)
C
R B
G
C=R+G+B

C R
G B
C+R=G+B
Humans have trichromatic color vision
25
Linear Color Matching

Grassman’s Laws:
1.
Scaling the color and the primaries by the same factor
preserves the match:
2C = 2R + 2G + 2B
2.
To match a color formed by adding two colors, add the
primaries for each color
C1 + C2 = (R1 + R2) + (G1 + G2) + (B1 + B2)
26
RGB Spectral Colors


Match each pure color in the visible spectrum
(rainbow)
Record the color coordinates as a function of
wavelength
?
27
Perception of color intensities



Which shade of gray is half-way between white and black?
It’s the second one
Humans perceive color intensity (and sound, etc.) on a
logarithmic scale

The first one is (about) 3/4 lit


The second one is 1/2 lit


We perceive it as 1/2 lit
We perceive it as 1/4 lit
That exponent is called gamma ()

2.0 is a sample value for a CRT or LCD monitor
28
Human Color Vision


Humans have 3 light sensitive pigments in their
cones, called L, M, and S
The cones
respond to
different lights:




L to red
M to green
S to blue
This leads to
metamerism

“Tristimulus”
color theory
29
Just Noticeable Differences


The human eye can distinguish hundreds of
thousands of different colors
When two colors differ only in hue, the wavelength
between just noticeably different colors varies with
the wavelength!





More than 10 nm at the extremes of the spectrum
Less than 2 nm around blue and yellow
Most JND hues are within 4 nm.
Altogether, the eye can distinguish about 128 fully
saturated hues
Human eyes are less sensitive to hue changes in
less saturated light (not a surprise)
30
Luminance


Compare color
source to a gray
source
Luminance


Y = .30R + .59G
+ .11B
Color signal on a
black and white
TV
31
Chromaticity and the CIE





Negative spectral
matching functions?
Some colors
cannot be
represented by
RGB
Enter the CIE
Three new standard
primaries called X, Y, and Z
Y has a spectral matching function exactly equal to
the human response to luminance
32
XYZ Matching Functions




Match all visible
colors with only
positive weights
Y matches
luminance
These functions
are defined
tabularly at 1nm intervals
Linear
combinations of
the R,G,B
matching
functions
33
CIE Color Space
34
Spectral Locus


Human perceptual gamut
The cone keeps going
towards the right

Brightness (not whiteness!)
keeps increasing
From http://pages.infinit.net/graxx/Theorie4.html
35
Chromaticity Diagram
Converting from RGB
to XYZ is a snap:
X  2.77 1.75 1.13 R 
Y  1.00 4.59 0.06 G 
  
  


Z 
 
0.00 0.57 5.59 

B 
X
X Y  Z
Y
y
X Y  Z
x
Given x, y, and Y, we can
recover the X,Y,Z coordinates
36
Measuring Color





Colorimeters measure the X, Y, and Z values for any color
A line between the “white point” of the chromaticity diagram
and the measured color intersects the horseshoe curve at
exactly the dominant wavelength of the measured color
A ratio of lengths will give the excitation purity of the color
Complementary colors are two colors that mix to produce
pure white
Some colors are non-spectral --- their dominant wavelength
is defined as the same as their complimentary color, with a
“c” on the end
37
Gamuts
38
Gamut problems



Monitor gamuts are
RGB
Printer gamuts are
CMYK
Each can display
colors the other
cannot
39
A Problem With XYZ Colors




If we have two colors C1 and C2, and we add C
to both of them, the differences between the
original and new colors will not be perceived to be
equal

C1:
add green

C2:
add green
This is due to the variation of the just noticeable
differences in saturated hues
XYZ space is not perceptually uniform
LUV space was created to address this problem
40
The RGB Color Model



This is the model used in color CRT
monitors
RGB are additive primaries
We can represent this space as a unit cube:
From http://ian-albert.com/graphics/rgb.php
41
More on RGB




The color gamut covered by the RGB model is
determined by the chromaticites of the three
phosphors
To convert a color from the gamut of one monitor
to the gamut of another, we first measure the
chromaticities of the phosphors
Then, convert the color to XYZ space, and finally
to the gamut of the second monitor
We can do this all with a single matrix multiply
42
The CMY Color Model

Cyan, magenta, and yellow are
complements of red, green, and blue



the
We can use them as filters to subtract from
white
The space is the same as RGB except the origin
is white instead of black
This is useful for hardcopy devices like
laser printers

If you put cyan ink on the page, no red light is
reflected
 C  1  R 
 M   1  G 
    
 Y  1  B 
43
CMYK


Most printers actually add a fourth
color, black
Use black in place of equal amounts
of C, M, and Y
K  min C , M , Y 
C CK
M M K
Y Y K

Why?


Black ink is darker than mixing C, M,
and Y
Black ink is cheaper than colored ink
44
CMY vs CMYK





You can create (more or
less) any color with each
gamut
Colored printer ink is more
expensive
Notice how much less
CMY is needed in the
CMYK version
One of the reasons
printers use CMYK
And color mixing…
45
The YIQ Color Model




YIQ is used to encode television signals
Y is the CIE Y primary, not yellow
Y is luminance, so I and Q encode the
chromaticity of the color
If we just throw I and Q away, we have
black and white TV
0.114   R 
Y  0.299 0.587
 I   0.596  0.275  0.321 G 
  
 
Q  0.212  0.528 0.311   B 



This assumes known chromaticities for your
monitor
Backwards compatibility with black and
white TV
More bandwidth can be assigned to Y
46
HSV color space aside

Consider a HSV picture space:



Let’s call the blue axis Cb


Blue and red are at right angles to each other
Thus, with 2 coordinates,
you can define any
saturation/hue combination
It defines the blue/yellow
combination
And the red axis Cr

It defines the red/cyan
combination
47
The YCbCr Color Model

Y is luma (similar to luminance)


Cb and Cr define the chrominance




The brightness of a pixel
Meaning they each define saturation
and hue
Cb is the blue chroma, Cr is the red
From the last slide
Notice the murkiness of the Cr and
Cb components

The human eye does not notice
differences in them nearly as much
48
JPEG Image Compression

Take an image in the (r,g,b) color space


Convert it to YCbCr




Let’s say 4 bits (24 = 16)
So it can have values 0, 15, 31, 47, … 255
Each pixel now takes up 16 bits


Also 8 bits per image
Downsample Cb and Cr to
fewer bits


Assume it’s 8 bits per image (24 bits total)
8 for Y, 4 for Cb and 4 for Cr
Then do some other magic (including zip-like compression)
And you have a (lossy) compressed image
49
Future of color displays

Future color displays may
have more pixels


Will allow much more vivid
color


RGB plus yellow, cyan, etc.
A greater gamut of color
possibilities
Note that both the pictures
on the right are being
displayed by an RGB
output device…
50
Photo printers

Photo printers use many ink colors for rich, vivid
color

Also a scam to sell you more ink (the razor business
model)
51