Image Fundamentals

Reference:
[Gonzalez and Woods] Rafael C. Gonzalez and Richard E. Woods, Digital
Image Processing, Prentice-Hall, Inc.
http://www.imageprocessingplace.com/

1
 Digital Imaging
• Charged-Coupled Device (CCD)
– consists of photosites (regions for collecting light)
• Each photosite (region) is a silicon imaging element
(sensor) that gives a voltage output proportional to the
intensity of the incident light
– linear array (scanner)
– area array (Camera CCD)

• http://en.wikipedia.org/wiki/Charge-coupled_device

2
 Photosite

light energy

+
- output voltage
3
http://en.wikipedia.org/wiki/Charge-coupled_device
Line scan sensor = a line of photosites

1D signal
4
Area scan sensor = 2D array of photosites

2D signal
5
 Continuous image function
y

f(x,y)
x
The continuous image function (intensity/color) f can be
characterized by two components:
(1) illumination (light source) and
(2) reflectance (materials).

Image intensity is depending on these two factors.

6
Continuous image function

Illumination: the amount of

i source illumination (energy)
incident on the scene being
viewed.

Reflectance: the amount of

source illumination (energy)
r
reflected by the objects
(or elements) in the scene.

f(x,y) = i(x,y) r(x,y) 7

[Gonzalez and Woods]
 Continuous image function
• Basic nature of f(x,y) has two components:
f(x,y) = i(x,y) r(x,y)

• i(x,y) is illumination component

0 ≤ i(x,y) < +infinity

• r(x,y) is reflectance component

0 ≤ r(x,y) ≤ 1
0 = total absorption, 1 = total reflectance.

• f(x,y) is intensity

• 0 ≤ f(x,y) < +infinity

8
Digital image formation
Camera model

r
f(x,y) = i(x,y) r(x,y)

i and r are continuous functions

thus f(x,y) is continuous
9
Digital image formation

f(x,y) = i(x,y) r(x,y)

10
[Gonzalez and Woods]
 Image resolution
The sampling rate on the
image plane is the image
Camera model resolution.

For example, 128 samples

along one horizontal line =
128 columns in the 2D image.
i
100 samples alone one vertical
r line = 100 rows in the 2D
image.
f(x,y) = i(x,y) r(x,y)
The image resolution is 100 x
128 samples (pixels).

The image is represented by a

100 x 128 matrix.

11
Spatially discretized

12
[Gonzalez and Woods]
 Discretized image representation
origin (0,0)
y

f(x,y), a 2D function representing

image intensity; (x,y) represents a
x pair of spatial coordinates
13
 Discretized image representation
origin
y •Discretized in
spatial coordinates

f(x,y)
x

Like a matrix in Linear Algebra

Individual elements are called:

image elements, picture elements (pixels), or image points.

14
 Discretized image representation

f(0,0) f(0,1) . . . f(0, m-1)

f(1,0) f(1,1) . . . f(1, m-1)
f(x,y) ~ .
.
.
f(n-1,0) f(n-1,1) . . . f(n-1,m-1)

Like a matrix in Linear Algebra with n rows and m columns

15
 Discretized image representation
y
Dimension 2
f(x, y) f(0, 0) f(0, 1) f(0, 2) (Horizontal),
or the 2nd
dimension
For example, f(1,0)
represents one
individual element, f(1, 0) f(1, 1) f(1, 2)
which is called the
image element,
picture element
(pixels), or image
points f(2, 0) f(2, 1) f(2, 2)
x
Dimension 1 or the 1st dimension 16
(Vertical)
Sampling and quantization are affecting the image quality.

Camera model

r
f(x,y) = i(x,y) r(x,y)

i and r are continuous functions

thus f(x,y) is continuous
17
Discretized image representation

origin •Discretized in both

y •Spatial coordinates
(Sampling)

•Brightness/Intensity
f(x,y) (Quantization)
x

18
 Discretized image representation

Image can be
viewed as a 3D
surface with
height and
different
“landscapes”

Image Source: Chapter 2, Figure 2.18,

http://www.imageprocessingplace.com/DIP-3E/dip3e_classroom_presentations_downloads.htm 19
Effects of reducing spatial resolution

More pixelated!

20
[Gonzalez and Woods]
 Effects of reducing spatial resolution

a b

Image Source: Chapter 2, Figure 2.20, 21

http://www.imageprocessingplace.com/DIP-3E/dip3e_classroom_presentations_downloads.htm
 Effects of reducing spatial resolution

c d

Image Source: Chapter 2, Figure 2.20, 22

http://www.imageprocessingplace.com/DIP-3E/dip3e_classroom_presentations_downloads.htm
 Intensity quantization
We call intensity of a monochrome
image, its grey level (L). In this
example, there are 16 levels
f(x) represented by using 4 bits, e.g.,
1111=level 15, and 1001=level 9.

f(x=3) = 9 = L
9

0 x, position
23
 Effects of intensity quantization

Image resolution
is fixed but the
total number of
gray levels
decreases.
256 levels 128 levels

64 levels 32 levels 24
Image Source: Chapter 2, Figure 2.21,
http://www.imageprocessingplace.com/DIP-3E/dip3e_classroom_presentations_downloads.htm [Gonzalez and Woods]
 Effects of intensity quantization

16 levels 8 levels
Image resolution
is fixed but the
number of gray
levels decreases.

Image Source: Chapter 2, Figure 2.21,

4 levels 2 levels 25
http://www.imageprocessingplace.com/DIP-3E/dip3e_classroom_presentations_downloads.htm [Gonzalez and Woods]
 Storage requirements
M = number of columns,
“Width”
y

Amount of memory needed

M*N*(bits required to
represent the grey levels)

x
N = number of rows, “height”
26