Digital Image Processing

Chapter 11: Image Description and

Representation

24 May 2014

Dr.Mogeeb A.A. Mosleh

Image Representation and Description?
Objective:
To represent and describe information embedded in
an image in other forms that are more suitable than the
image itself.
Benefits:
- Easier to understand
- Require fewer memory, faster to be processed
- More “ready to be used”
What kind of information we can use?
- Boundary, shape
- Region
- Texture
- Relation between regions
Shape Representation by Using Chain Codes
Why we focus on a boundary?
The boundary is a good representation of an object shape
and also requires a few memory.
Chain codes: represent an object boundary by a connected
sequence of straight line segments of specified length
and direction.

4-directional 8-directional (Images from Rafael C.

chain code
Gonzalez and Richard E.

chain code Wood, Digital Image

Processing, 2nd Edition.
Examples of Chain Codes

Object Boundary
boundary vertices
(resampling)

4-directional 8-directional
chain code chain code

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
The First Difference of a Chain Codes
Problem of a chain code:
a chain code sequence depends on a starting point.
Solution: treat a chain code as a circular sequence and redefine the
starting point so that the resulting sequence of numbers forms an
integer of minimum magnitude.
The first difference of a chain code: counting the number of direction
change (in counterclockwise) between 2 adjacent elements of the code.
Example: Chain code : The first Example:
1 difference - a chain code: 10103322
01 1 - The first difference = 3133030
2 0 02 2 - Treating a chain code as a
03 3 circular sequence, we get
3 23 1 the first difference = 33133030
20 2
21 3 The first difference is rotational
invariant.
Polygon Approximation

Represent an object boundary by a polygon

Object boundary Minimum perimeter

polygon

Minimum perimeter polygon consists of line segments that

minimize distances between boundary pixels.
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Polygon Approximation:Splitting Techniques

1. Find the line joining

two extreme points
0. Object boundary 2. Find the
farthest points
from the line

3. Draw a polygon (Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Distance-Versus-Angle Signatures
Represent an 2-D object boundary in term of a 1-D function
of radial distance with respect to .

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Boundary Segments
Concept: Partitioning an object boundary by using vertices
of a convex hull.

Partitioned boundary

Convex hull (gray color)

Object boundary

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
 Convex Hull Algorithm
points on a cornea boundary
points on a boundary of a convex hull of
s by x-coordinate to get a sequence p1, p
de of a convex hull
p1 and p2 in a list Lupper with p1 as the firs

pi to Lupper
ntains more than 2 points and the last 3
in Lupper do not make a right turn
elete the middle point
Turn
of the last 3 poin
Turn Left
Turn
Right NOK!
Right
OK!
OK!
 Convex Hull Algorithm (cont.)
side of a convex hull
nts pn and pn-1 in a list Llower with pn as the
down to 1
d pi to Llower
contains more than 2 points and the las
Llower do not make a right turn
delete the middle point of the last 3 poin
e first and the last points from Llower
wer
to Lupper resulting in the list L
Turn
Turn Right
Right OK!
Turn
OK!
Left
NOK!
Skeletons
Obtained from thinning or skeletonizing processes

Medial axes (dash lines)

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
 Thinning Algorithm
ncept: 1. Do not remove end points
2. Do not break connectivity
3. Do not cause excessive erosion
Apply only to contour pixels: pixels “1” having at least one of its 8
neighbor pixels valued “0”
Notation:
p9 p2 p3 Neighborhood
arrangement
Let p 8 p 1 p 4 =
for the thinning
p7 p6 p5 algorithm
Example
Let N ( p1 )  p2  p3    p8  p9 0 0 1
T(p1) = the number of transition 0-1 in 1 p1 0
the ordered sequence p2, p3, … 1 0 1
, p8, p9, p2. N(p1) = 4
T(p ) = 3
Thinning Algorithm (cont.)

Step 1. Mark pixels for deletion if the following conditions are true.
a) 2  N ( p1 ) 6
(Apply to all border pixels) p9 p2 p3
b) T(p1) =1
p  p p  0 p8 p1 p4
c) 2 4 6
p7 p6 p5
d) p4 p6 p8 0
Step 2. Delete marked pixels and go to Step 3.
Step 3. Mark pixels for deletion if the following conditions are true.
a) 2  N ( p1 ) 6
(Apply to all border pixels)
b) T(p1) =1
c) p2 p4 p8 0
d) p2 p6 p8 0
Step 4. Delete marked pixels and repeat Step 1 until no change
occurs.
Example: Skeletons Obtained from the Thinning Alg.

Skeleton

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Boundary Descriptors
1. Simple boundary descriptors:
we can use
- Length of the boundary
- The size of smallest circle or box that can totally
enclosing the object

2. Shape number

3. Fourier descriptor

4. Statistical moments
Shape Number
Shape number of the boundary definition: 1
the first difference of smallest magnitude
The order n of the shape number: 2 0
the number of digits in the sequence
3

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Shape Number (cont.)

Shape numbers of order

4, 6 and 8

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Example: Shape Number

2. Find the smallest rectangle

that fits the shape
1. Original boundary

Chain code:
000030032232221211

First difference:
300031033013003130

4. Find the nearest Shape No.

3. Create grid Grid. 000310330130031303
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Fourier Descriptor
Fourier descriptor: view a coordinate (x,y) as a complex number
(x = real part and y = imaginary part) then apply the Fourier
transform to a sequence of boundary points.
Let s(k) be a coordinate s( k )  x ( k )  jy ( k )
of a boundary point k :

K1
1
Fourier descriptor : a (u ) 
K
 s
k 0
( k ) e  2uk / K

Reconstruction formula
K1
1
s(k ) 
K
 a
k 0
( u ) e 2uk / K

Boundary
points (Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Example: Fourier Descriptor
Examples of reconstruction from Fourier descriptors
P 1
1
sˆ( k ) 
K
 a (
k 0
u ) e 2uk / K

P is the number of
Fourier coefficients
used to reconstruct
the boundary

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Fourier Descriptor Properties

Some properties of Fourier descriptors

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Statistical Moments
Definition: the nth moment Example of moment:
K1 The first moment = mean
n ( r )  ( ri  m)n g ( ri ) The second moment = variance
i 0

K1
where
m  ri g ( ri )
i 0

Boundary
segment 1D graph

1. Convert a boundary segment into 1D graph

2. View a 1D graph as a PDF function (Images from Rafael C.
3. Compute the nth order moment of the graph Gonzalez and Richard E.
Wood, Digital Image
Processing, 2nd Edition.
Regional Descriptors
Purpose: to describe regions or “areas”
1. Some simple regional descriptors
- area of the region
- length of the boundary (perimeter) of the region
- Compactness
A( R )
C 2
P ( R)
where A(R) and P(R) = area and perimeter of region R
Example: a circle is the most compact shape with C = 1/4

2. Topological Descriptors
3. Texture
4. Moments of 2D Functions
Example: Regional Descriptors

White pixels represent

“light of the cities”

% of white pixels
Region no. compared to the
total white pixels

1 20.4%
2 64.0%
3 4.9%
4 10.7%

Infrared image of America at night (Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Topological Descriptors
Use to describe holes and connected components of the region

Euler number (E):

E C  H

C = the number of connected

components
H = the number of holes

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Topological Descriptors (cont.)

E = -1

E=0 Euler Formula

V  Q  F C  H  E

V = the number of vertices

Q = the number of edges
F = the number of faces

E = -2 (Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
 Example: Topological Descriptors
ginal image:
ared image After intensity
Washington Thresholding
. area (1591 connected
components
with 39 holes)
Euler no. = 1552

The largest
After thinning
connected
area
(8479 Pixels)
(Hudson river)

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Texture Descriptors
Purpose: to describe “texture” of the region.
Examples: optical microscope images:

B C
A

Superconductor Cholesterol Microprocessor

(smooth texture) (coarse texture) (regular texture)
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Statistical Approaches for Texture Descriptors
We can use statistical moments computed from an image histogram:
K1
z = intensity
n ( z )  ( zi  m) p( zi )
n
p(z) = PDF or histogram of z
i 0

where K1
m  zi p( zi )
i 0

Example: The 2nd moment = variance  measure “smoothness”

The 3rd moment  measure “skewness”
The 4th moment  measure “uniformity” (flatness)

A
B
C
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Fourier Approach for Texture Descriptor
Concept: convert 2D spectrum into 1D graphs

Fourier
Original FFT2D Divide into areas
coefficient
image +FFTSHIFT by angles
image

Divide into areas Sum all pixels

by radius in each area

R0
S ( )  Sr ( )
r 1


Sum all pixels
S ( r )  S ( r )
in each area  0
Fourier Approach for Texture Descriptor

Original 2D Spectrum
image (Fourier Tr.)

S(r) S()

Another Another S()

image
(Images from Rafael C.
Gonzalez and Richard E.
Wood, Digital Image
Processing, 2nd Edition.
Moments of Two-D Functions
The moment of order p + q
m pq  x p y q f ( x, y ) m10 m01
x y
x y m00 m00
The central moments of order p + q
 pq  ( x  x ) p ( y  y ) q f ( x, y )
x y

00 m00 01 10 0

11 m11  x m01 m11  ym10
20 m20  x m10 02 m02  ym01
21 m21  2 x m11  ym20  2 x 2m01 30 m30  3x m20  2 x 2m10
12 m12  2 ym11  x m02  2 y 2m10 03 m03  3 ym02  2 y 2m01
Invariant Moments of Two-D Functions
The normalized central moments of order p + q
 pq pq
 pq   where   1
00 2

Invariant moments: independent of rotation, translation, scaling,

and reflection
1 20  02 2 20  02 2  4112
3 30  312 2  321  03 2 4 30  12 2  21  03 2

5  30  312 30  12 30  12  3 21  03  

  
 2
 2

321  03 21  03 330  12 2  21  03 2 

6  20  02 30  12  21  03  
 
 2
 2

411 30  12 21  03 

Example: Invariant Moments of Two-D Functions
1. Original image 2. Half size 3. Mirrored

4. Rotated 2 degree 5. Rotated 45 degree

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Example: Invariant Moments of Two-D Functions

Invariant moments of images in the previous slide

Invariant moments are independent of rotation, translation,

scaling, and reflection

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Principal Components for Description
Purpose: to reduce dimensionality of a vector image while maintaining
information as much as possible.

Let x [ x1 x2 ... xn ]T

K
1
Mean: m x  E{x} 
K
x
k 1
k

K
1
Covariance matrix C x  E{( x  m x )( x  m x )T } 
K
 k k x x
x x
k 1
T
 m mT
Hotelling transformation
Let y A ( x  m x )

Where A is created from eigenvectors of Cx as follows

Row 1 contain the 1st eigenvector with the largest eigenvalue.
Row 2 contain the 2nd eigenvector with the 2nd largest eigenvalue.
….
Then we get
 1 0 ... 0 
m y  E{y} 0  0  ... 0 
and C y   1 
 ... ... ... ... 
C y AC x A T  
 0 ... ...  1

Then elements of y A ( x  m x ) are uncorrelated. The

component of y with the largest  is called the principal
component.
Eigenvector and Eigenvalue

Eigenvector and eigenvalue of Matrix C are defined as

Let C be a matrix of size NxN and e be a vector of size Nx1.

If
Ce e

Where  is a constant

We call e as an eigenvector and  as eigenvalue of C

Example: Principal Components

6 spectral images
from an airborne
Scanner.

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Example: Principal Components (cont.)

Component 
1 3210
2 931.4
3 118.5
4 83.88
5 64.00
6 13.40
(Images from Rafael C. Gonzalez and Richard E.
Wood, Digital Image Processing, 2nd Edition.
Example: Principal Components (cont.)

Original image After Hotelling transform

Principal Components for Description

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Relational Descriptors

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Relational Descriptors

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Relational Descriptors

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Relational Descriptors

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Relational Descriptors

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.
Structural Approach for Texture Descriptor

(Images from Rafael C. Gonzalez and Richard E.

Wood, Digital Image Processing, 2nd Edition.