Image Segmentation, Representation and Description

Image segmentation, Repres
entation, and Description

:
OUTLINE
Image segmentation
ex: edge-based, region-based
Image representation
ex: Chain code , polygonal approximation signatures, sk
eletons
Image description
ex: boundary-based, regional-based
Conclusion
Image segmentation
edge-based: point, line, edge detection
edge-based segmentation(1)
There are three basic types of gray-level discontinuities

in a digital image: points, lines, and edges
The most common way to look for discontinuities is to r
un a mask through the image.
We say that a point, line, and edge has been detected at
the location on which the mask is centered if
,wher
e
R T
R w1 z1 w2 z2 ...... w9 z9
Point detection
a point detection mask
Line detection
a line detection mask

5
f
Edge detection:
Gradient op
G
x
f G f
eration
y
x
f mag (f ) Gx G y
2
( x, y ) tan (
1
Gy
Gx
2 f 2 f
f 2 2
x
y
2
r2
r 2 2 2 2
2
h(r )
e
4
Edge detection: Laplacian

operation
7
Image segmentation
Region-base: SRG, USRG, Fast scanning
region-based segmentation
SRG(1)
Region growing: Groups pixels or sub-region into

larger regions.
step1:
Start with a set of seed points and from these g
row regions by appending to each seed those neig
hboring pixels that have properties similar to the
seed.
step2:
Region splitting and merging
SRG(2)
Advantage:
With good connectivity
Disadvantage:
Initial seed-points:
different sets of initial seed-point cause different segmente
d result
Time-consuming problem
10
USRG(1)
Unseeded region growing:

no explicit seed selection is necessary, the seed
s can be generated by the segmentation proced
ure automatically.
It is similar to SRG except the choice of seed poi
nt
11
USRG(2)
Advantage:
easy to use
can readily incorporate high level knowledge of the ima
ge composition through region threshold
Disadvantage:
slow speed
12
fast scanning(1)
Fast scanning Algorithm:

The fast scanning algorithm
somewhat resembles unsee
ded region growing
the number of clusters of b
oth two algorithm would no
t be decided before image p
assing through them.
13
fast scanning(2)
14
fast scanning(3)
Last step:
merge small region to big re
gion
15
fast scanning(4)
Advantage:
The speed is very fast
The result of segmentation will be intact with good conn
ectivity
Disadvantage:
The matching of physical object is not good
It can be improved by morphology and geometric mathema
tic
16
fast scanning-improved by morphology
A B {c E N | c a b for some a A and b B}
dilation
A ! B {x E N x b A for every b B}
erosion
17
dilation
erosion
18
opening
Erosion=>Dilation
closing
Dilation=>Erosion
19
fast scanning-improved by Geometric Math
ematic
20
fast scanning-improved by Geometric Math
ematic
21
application
0.6
0.5
Muscle Injury Determination
0.3
The quadratic regression equation

Image of the unhealthy muscle fiber
Image of the healthy muscle fiber
Use fast scanning algorithm to

segment it.
0.4
How to judge for using image

segmentation?
0.2
0.1
0
-0.1
0.05
0.1
0.15
0.2
0.25
X
0.3
0.35
0.4
0.45
0.5
22
Representation
chain code, polynomial approximation,
signature, skeletons
23
Representation
chain code
4-direction
8-direction
24
Representation
polynomial approximations
Merging Techniques
Splitting Techniques
25
Representation
signature
Distance signature of circle sh

apes
Distance signature of rectang

ular shapes
26
Representation
skeletons
2 N ( p1 ) 6
T ( p1 ) 1
p2 gp4 gp6 0
p4 gp6 gp8 0
Step1:
(a)
(b)
(c)
(d)
p2 gp4 gp8 0
p2 gp6 gp8 0
Step2:
(c)
(d)
27
Descriptors
boundary descriptor: Fourier descriptor,
polynomial approximation
28
Boundary Descriptors
Fourier descriptors (1)
s( k ) x(k ) jy (k )
Step1:
Step2: (DFT)
a (u )
1
K
K 1
j 2 uk / K
s
(
k
)
e
k 0
Step3: (reconstruction)
if a(u)=0 for u>P-1
P 1
s$(k ) u 0 a(u )e j 2 uk / K
Disadvantage:
Just for closed boundaries
29
Whats the reason that p

revious Fourier descripto
rs cant be used for non-c
losed boundaries?
s1(k)
(xK1, yK1)
(x0, y0)
s2(k)
Step 2
Original segment
How can we use the met

hod to descript non-close
d boundaries?
(a)linear offset
(b)odd-symmetric
extension
(b) Linear
offset
s3(k)
Step 3
(c) Odd symmetric extension
30
The proposed method is

used not only for non-clo
sed boundaries but also f
or closed boundaries.
Why we used proposed

method to descript close
d boundaries rather than
previous method?
31
polynomial approximation(1)
n
P ( x) f ( x0 ) Ln ,0 ( x ) L f ( xn ) Ln, n ( x) f ( xk ) Ln ,k ( x)
k 0
Ln ,k ( x)
( x x0 )L ( x xk 1 )( x xk 1 )L ( x xn )
( xk x0 ) L ( xk xk 1 )( xk xk 1 )L ( xk xn )
f ( n 1) ( ( x))
f ( x) P ( x)
( x x0 )( x x1 ) L ( x xn )
( n 1)!
e Lagrange
f ( x ) P ( x)Polynomial
Cubic Spline Interpolation
32
Proposed method(1)
Step1: rotate the boundary an
d let two end point locate at
x-axis
4b second
a
Step2:
poly
y 2use
( x ' ) 2 order
b
a to approximate
2
nomial
the bo
undary
n 1
e y ' y
2
j 0
4b
a
y j ' 2 ( x j ' ) 2 b
a
2
33
Proposed method(2)
If the boundary is closed, h
ow can we do?
Step1: use split approach di
vide the boundary to two pa
rts.
Step2: use parabolic functio
n to fit the boundary.
34
Descriptors
Regional descriptors: Topological, Texture
35
Regional Descriptors
Topological
E=V-Q+F=CH
E: Euler number
V: the number of vertices
Q: the number of edges
F: the number of faces
C: the number of
connected component
H: the number of holes
36
Regional Descriptors
Texture
un ( z ) i 0 ( zi m) n p( zi )
nth moment:
L 1
L 1
i 0
zi p ( z i )
2th moment:
is a measure of gray level contra
st(relative smoothness)
3th moment:
is a measure of the skewness of
the histogram
Statistical approaches
smooth, coarse, regular
4th moment:
is a measure of its relative flatne
ss
5th and higher moments:

are not so easily related to histo
gram shape
37
Conclusion
Image segmentation
speed, connectivity, match physical objects or not
match physical objects:
morphological: how to choose foreground or background?
geometric mathematic: wrong connection
Representation & Description

Boundary descriptor:
rotation, translation, degree of match boundary, closed or n
on-closed boundary
38
Reference
[1] R.C. Gonzalez, R.E. Woods, Digital Image Processing s

econd edition, Prentice Hall, 2002
[2] J.J. Ding, W.W. Hong, Improvement Techniques for Fa
st Segmentation and Compression
[3] J.J. Ding, Y.H. Wang, L.L. Hu, W.L. Chao, Y.W. Shau, Mu
scle Injury Determination By Image Segmentation
[4] J.J. Ding, W.L. Chao, J.D. Huang, C.J. Kuo, Asymmetric
Fourier Descriptor Of Non-Closed segments
39
Thank you for listening
40

Image Segmentation, Representation and Description

Uploaded by

Copyright:

Available Formats

Image Segmentation, Representation and Description

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

Image Segmentation, Representation and Description

Uploaded by

Copyright:

Available Formats

Image segmentation, Repres

entation, and Description

There are three basic types of gray-level discontinuities

a point detection mask

a line detection mask

Edge detection: Laplacian

Region growing: Groups pixels or sub-region into

Unseeded region growing:

Fast scanning Algorithm:

A B {c E N | c a b for some a A and b B}

Muscle Injury Determination

The quadratic regression equation

Use fast scanning algorithm to

How to judge for using image

Distance signature of circle sh

Distance signature of rectang

Whats the reason that p

How can we use the met

The proposed method is

Why we used proposed

Cubic Spline Interpolation

5th and higher moments:

Representation & Description

[1] R.C. Gonzalez, R.E. Woods, Digital Image Processing s

Thank you for listening

You might also like