0% found this document useful (0 votes)
90 views40 pages

Image Segmentation

The document provides an overview of three major image segmentation techniques: thresholding, edge-based segmentation, and region-based segmentation. It describes thresholding methods such as basic global thresholding, Otsu's method, multiple thresholding, and variable thresholding. For edge-based segmentation, it discusses basic edge detection using gradient operators, the Marr-Hildreth edge detector, short response Hilbert transform, and watersheds. Finally, it summarizes region-based segmentation including region growing and the use of markers in watershed segmentation.

Uploaded by

Arga Bz1
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
90 views40 pages

Image Segmentation

The document provides an overview of three major image segmentation techniques: thresholding, edge-based segmentation, and region-based segmentation. It describes thresholding methods such as basic global thresholding, Otsu's method, multiple thresholding, and variable thresholding. For edge-based segmentation, it discusses basic edge detection using gradient operators, the Marr-Hildreth edge detector, short response Hilbert transform, and watersheds. Finally, it summarizes region-based segmentation including region growing and the use of markers in watershed segmentation.

Uploaded by

Arga Bz1
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 40

An Introduction of Image

Segmentation
Outline & Content
1. Introduction
2. Thresholding
3. Edge-based segmentation
4. Region-based segmentation
5. conclusion

2
1. Introduction
What is segmentation?

 Three major ways to do.


 Thresholding
 Edge-based segmentation
 Region-based segmentation
3
Thresholding
Finding histogram of gray level intensity.

 Basic Global Thresholding


 Otsu’s Method
 Multiple Threshold
 Variable Thresholding

4
Edge-based segmentation
Using mask to detect edge in image by convolution.

 Basic Edge Detection


 The Marr-Hildreth edge detector(LoG)
 Short response Hilbert transform(SRHLT)
 Watersheds

5
Region-based segmentation
Finding region, but not finding edge.

 Region Growing
 Data Clustering (Hierarchical clustering)
 Partitional clustering
 Cheng-Jin Kuo`s method

6
2. Thresholding
Basic Global Thresholding
1) Select an initial To
2) Segment image use:

g ( x, y )   01 if f(x,y) T
if f(x,y)  T

3) Compute the average intensity m1


and mfor
2 the pixels in and .

1
4) Compute a new threshold: T (m 1  m 2 )
2
5) Until the difference between values of T is smaller than a predefined para
meter.

7
Otsu’s Method
{0,1,2,…,L-1} , L means gray level intensity
 MN  n 0  n 1  n 2 ...  n L 1
M*N is the total number of pixel.
n i denote the number of pixels with intensity i
we select a threshold T (k )  k , 0  k  L  1 , and use it to
classify C : intensity in the range [0, k ] and C : [k  1, L  1]
k
1
L 1
2

P (k )   p , P (k )   p  1  P (k )
1 i 2 i 1
i 0 i  k 1

P 1 m 1  P 2 m 2  m G , P 1 + P 2 =1
L1

 =  i  m  p , it is global variance.
2
G G
2
i
i 0

8
( m G P 1 (k )  m(k )) 2
  (k )  P 1 P 2 (m1  m 2 ) 
2
B
2
P 1 (k )(1  P 1 (k ))
it is 2between-class variance
 (k )
  = B 2 , 0   (k *)  1
G
it is a measure of separability between class.
  B2 (k  )  max  B2 (k )
o  k  L 1
 g ( x, y )   1 if f(x,y)k*
0 if f(x,y)  k*

For x = 0,1,2,…,M-1 and y = 0,1,2…,N-1.


Using image Smoothing/Edge to improve Global Thre
shold
Smoothing Edge
detection
What situation is Large object we Small object we
more suitable for the are interested. are interested
method
9
Multiple Threshold
 As Otsu’s method, it takes more area and k*
  B2  P 1 (m 1 m G )2  P 2 (m 2 m G )2  P 3 (m 3 m G )2
 P1m1P 2 m 2 P 3 m 3  m G
 P 1 + P 2 +P 3 =1
  B2 (k1* , k2* )  max  B2 (k 1 , k 2 )
o  k 1  k 2  L 1

* *  2
( k *
 (k , k )  B 12 2 ) , k *

G
1 2

Disadvantage: it becomes too complicate when numbe


r of area more than three.

10
Variable Thresholding
1) Image partitioning
.

It is work when the objects of interest and the backgro


und occupy regions of reasonably comparable size. If n
ot , it will fail.

11
2) Variable thresholding based on local image prop
erties
 Let  xy and m xy denote the standard deviation and m
ean value of the set of pixels contained in a neighbor
hood, . S xy
 g ( x, y )   1 if Q(local parameters) is true
0 if Q(local parameters) is true

 Q ( xy , m xy )   true if f ( x , y )  a
false otherwise
xy ANDf ( x , y ) bm xy

12
3) Using moving average
 It discussed is based on computing a moving average
along scan klines
1
of an image.
 m(k  1)  1  z i  m(k )  1 ( z k 1  z k n )
n i  k  2 n n
 z k 1 denote the intensity of the point at step k+1.
n denote the number of point used in the average.
 m(1)  z 1 / n is the initial value.
 T xy  bm xy ,where b is constant and is the moving a
verage at point (x,y)

13
3. Edge-based segmentation
Basic Edge Detection
Why we can find edge by difference?
 image
intensity
first-order deviation
second-order deviation
First-order produce
deviation thicker edges

Second-order stronger double-edge Determine edge


deviation response to fine response is from light to
detail dark or dark to
light
14
1) Gradient
The image gradient is to find edge strength and direction at
location (x,y) of image.
 f 
  g x   x 
f  grad( f )      
 g y   f 
 y 

The magnitude (length) of vector , denoted as M(x,y):


mag(f )  g x  g y

The direction of the gradient vector is given by the angle:


gy
1
 ( x, y )  tan  
 g x 

15
Roberts cross-gradient operators:

Prewitt operator:

Sobel operator:

G ( x, y )   1
0
if R ( x , y ) T
otherwise

16
The Marr-Hildreth edge detector(LoG)
This is second-order deviation, we call Laplacian.

 g ( x, y )  [ 2G ( x, y )]  f ( x, y ) g ( x, y )   2 [G ( x, y )  f ( x, y )]
 Filter the input image with an n*n Gaussian lowpass fi
lter. 99.7% of the volume under a 2-D Gaussian surface
lies between about the mean. So . n  6

17
Short response Hilbert Transform(SRHLT)
1) Hilbert transform
1
 g H ( )  h( x) * g ( x), where h( x) 
x
 GH ( f )  H ( f )G ( f ) , H ( f )   j sgn( f )

18
2) SRHLT
 g H     hb  x   g  x  , where hb  x   b csch( bx )
 GH ( f )  H b ( f )G ( f ) where GH ( f )  FT [ g H ( )],
G ( f )  FT [ g ( )], H b ( f )   j tanh( f / b).

Choose a suitable
HLT value differentiation
b0 b

Lower b Higher b

Impulse response Longer Shorter
Noise robustness Good Bad
Type of edge Ramp Step
Output Thick Sharp
19
Simulation result

20
Watersheds

Algorithm:
T [n]  {( s, t ) | g ( s, t )  n} , g(s,t) is intensity.
n= min+1 to n = max +1. And let T[n]=0, others 1.
R
C[n]  C ( M ) ,C ( M ) is minimum point beneath n.
 n i n i
i 1

21
Markers
External markers:
Points along the watershed
line along highest points.

Internal markers:
(1) That is surrounded higher points .
(2) Points in region form a connected component
(3) All points in connected component have the same
intensity.
22
4. Region-based segmentation
Region Growing
Algorithm:
a. Choose a random pixels
b. Use 8-connected and threshold to determine
c. Repeat a and b until almost points are classified.

23
Simulation of region growing (90% pixels )
Threshold/second: 20/4.7 seconds.

24
Data Clustering

 Using centroid to represent the huge numbers of clu


sters

 Hierarchical clustering, we can change the number o


f cluster anytime during process if we want.

 Partitional clustering , we have to decide the number


of clustering we need first before we begin the proces
s.

25
Hierarchical clustering
1) Algorithm of hierarchical agglomeration(built)
: Ci
A. See every single data as a cluster .
Ci Cj
B. Find out , for the distance is the shortest.

C. Repeat the steps until satisfies our demand.


d (a, b)
• as the distance between data a and b

26
2) Algorithm of hierarchical division (break up ) :
 Diameter of cluster
D(Ci )  max(d (a,b)) , for a  Ci , b  Ci

27
1) See the whole database as one cluster.
2) Find out the cluster having the biggest diameter
3) max (d ( x, c)) , for x  C .
4) Split out x as a new cluster C1 , and see the rest data
points as Ci .
5) If d ( y, Ci ) > d ( y, C1 ) , for y  Ci then split y out of Ci and c
lassify it to C1
6) Back to step2 and continue the algorithm until C1 an
d Cisi not change anymore.

28
Partitional clustering
Decide the numbers of the cluster(k-means)

Problem:
Initial problem

29
Number of regions are more than clusters you set.

Determine the number of clusters.


 intra
Vadility=
inter
k
1
intra-cluster distance=
N
  x 
i 1 xCi
zi 2

inter-cluster distance  min( zi  z j 2 )


, i  1, 2,3,......, K  1
; j  i  1,......K 30
Simulation of k-means
Clustering/time: 9 clustering/ 0.1

31
Advantage and disadvantage of data clustering
Hierarchical Partitional
algorithm algorithms
advantage 1.Concept is simple 1.Computing speed
is fast.
2. Result is reliable. 2.Numbers of
cluster is fixed, so
the concept is also
simple.

disadvantage It is consuming, so 1. Determine the


is not suitable for number of
a large database. clusters.
2. Initial problem
3. ….

32
Cheng-Jin Kuo`s method
Algorithm

33
1) Adaptive threshold decision with local variance

 Variance of Lena: 1943


716 447 1293 2470
899 1579 1960 2238
1497 1822 1974 1273
2314 1129 1545 1646

 Small variance cause small threshold.


34
2) Adaptive threshold decision with local variance
and frequency

 Variance of baboon: 1503


716 447 1293 2470
899 1579 1960 2238
1497 1822 1974 1273
2314 1129 1545 1646
6.9451 8.8807 7.2965 7.0914
8.0413 10.0076 8.4951 7.6421
9.6709 10.3219 7.9815 6.1310
9.0464 10.4821 7.1513 6.4118 35
High frequency, high variance. Set highest threshold.

(4,1)
High frequency, low variance. Set second high thresho
ld.

(4,2)
Low frequency, high variance. Set third high threshold
.
(1,4)
Low frequency, low variance. Set the lowest threshold.
 (1,1)
36
Comparison of all algorithm by data compr
ession
Region K-means Watershed Cheng-Jin
growing Kuo`s
method
Speed bad Good(worse bad Good
than C.J.K’s
method)
Shape intact fragmentary oversegmen Intact
connectivi tation
ty
Shape Good(better Good(equal bad Good
match than C.J.K’s C.J.K’s
method) method)

37
Conclusion
Speed

Connectivity

System reliability

38
Reference
1. R. C. Gonzalez, R. E. Woods, Digital Image Processin
g third edition, Prentice Hall, 2010.
2. C. J. Kuo, J. J. Ding, Anew Compression-Oriented Fas
t image Segmentation Technique, NTU,2009.

39
Q&A

40

You might also like