DIP3E Chapter10 Art

Digital Image Processing, 3rd ed.
Gonzalez & Woods

www.ImageProcessingPlace.com
Ch t 10
Chapter
Segmentation
Image
g Segmentation
g
Subdividing an image into
different regions based on some
criterion/criteria.
© 1992–2008 R. C. Gonzalez & R. E. Woods

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Image
g Derivatives
1. First order generally produces
thick edges
2. Second order has stronger
response to fine detail
3. Second order derivates may
produce double edges (in ramp
and step transition intensities).
4. The sign of the second
derivative indicates whether
the transition from dark to
light or light to dark
f
 f ( x  1)  f ( x)
x
2 f
 f ( x  2)  2 f ( x)  f ( x  1)
x 2

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Detection of Isolate Points

The advantages of the second
derivative implies using the
Laplacian
2 f 2 f
 f ( x, y )  2  2
2
x y
2 f
 f ( x  2, y )  2 f ( x, y )  f ( x  1, y )
x 2
2 f
 f ( x, y  2)  2 f ( x, y )  f ( x, y  1)
y 2
 2 f ( x, y )  f ( x  2, y )  f ( x, y  2)  4 f ( x, y )
 f ( x  1, y )  f ( x, y  1)
A point at location (x,y) can be
detected using its response g(x,y)
R(x,y)
R(x y) is the result of applying a filter
1 | R( x, y ) | T at the location x,y
g ( x, y )  
0 otherwise
Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Line Detection
Lines are detected by
1. Finding the points of the
lines
2. Collect these points into
line.
Applying Laplacian on an image.

1. Original
2. Laplacian Result
3. Absolute value
4. Positive only

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Line/Edge
g Detection
1. The points along the lines do not
usually fall on an ideal line.
2. The line may not be continuous.

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Edge
g Models
The transition from a bright intensity
to dark one, have different patterns.

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Edgeg Models
The existence of noise complicate line
detection therefore
1. Image smoothing often contribute
to noise reduction.
2. Detection of edge points- potential
candidates for edge points.
3 Edge
3. d localization-
l li i to select
l points
i
which are actually on the lines.

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Image
g Gradient
 f 
 g x   x 
f  grad ( f )      
 g y   f 
 x 
2 2
M ( x, y )  magg (f )  g x  g y
 gy 
 ( x, y )  tan 1  
 gx 
Gradient is often expressed as

M ( x, y )  g x  g y

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Robert operation
p
f
gx   ( z9  z5 )
x
f
gy   ( z8  z 6 )
y
The simplest approximation of partial
dervatives on 3x3 mask
f
gx   ( z7  z8  z9 )  ( z1  z 2  z3 )
x
f
gy   ( z3  z6  z9 )  ( z1  z 4  z7 )
y

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Marr-Hildreth Edge Detector

Marr and Hildreh argued
1. Intensity changes are not independent
of image scale
2 The studden change in intensity show
2.
a peak in the first derivative and zero
crossing in the second
x2  y2

G ( x, y )  e 2 2
 2 G ( x, y )  2 G ( x , y )
 G ( x, y ) 
2

x 2 y 2
  x      y  x  y2 
x2  y 2 2 2
  2e    2 e 2 
2
2
x    y   
x2  y2 x2  y 2
 x2 1    y2 1  
  4  2 e   4  2 e
2
2 2 2
     
This expression is often called Laplacian of Gaussian
x2  y2
 x  y  2  
2 2 2
(LoG)
 e
2 2
  4

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Canny Edge Detector

1. Low error rate
2. Edge point should be well localized
3. Single edge point response
Algorithm
1. Smooth image with a Gaussian
- optimizes the trade-off between noise
filtering and edge localization
2. Compute the Gradient magnitude using
approximations of partial derivatives
3 Thi
3. Thin edges
d by
b applying
l i non-maximai
suppression to the gradient magnitude
4. Detect edges by double thresholding

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Gradient
• At each point convolve with
  1 1 1 1
Gx   Gy   
 1 1   1  1
 
• magnitude and orientation of the Gradient are
computed as
M [i, j ]  P[i, j ]2  Q[i, j ]2
 [i, j ]  tan 1 (Q[i, j ], P[i, j ])
• Avoid
A id fl
floating
i point
i arithmetic
i h i forf fast
f computation
i
Canny Edge Detector 19
Non-Maxima Suppression Gonzalez & Woods

Ch t 10
Chapter
Segmentation
• Thin edges by keeping large values of Gradient

– not always at the location of an edge
– there are manyy thick edges
g
0 0 0 0 1 1 1 3
0 0 0 1 2 1 3 1
0 0 2 1 2 1 1 0
0 1 3 2 1 1 0 0
0 3 2 1 0 0 1 0
2 3 2 0 0 1 0 1
2 3 2 0 1 0 2 1

Non-Maxima Suppression (2) Gonzalez & Woods

Ch t 10
Chapter
Segmentation
• Thin the broad ridges in M[i,j] into ridges that are only one
pixel
i l wide
id
• Find local maxima in M[i,j] by suppressing all values along
the line of the Gradient that are not peak values of the ridge
0 0 0 0 1 1 1 3
3 0 0 1 2 1 3 1
0 0 2 1 2 1 1 0
false 0 1 3 2 1 1 0 0
edges 0 3 2 1 0 0 1 3
2 3 2 0 0 1 0 1 gaps
2 3 2 0 1 0 2 1

Gradient Orientation
Gonzalez & Woods
Ch t 10
Chapter
Segmentation
• Reduce angle of Gradient θ[i,j] to one of the 4 sectors

• Check the 3x3 region of each M[i,j]
• If the value at the center is not greater than the 2
values along the gradient, then M[i,j] is set to 0

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
0 0 0 0 1 1 1 3 local
0 0 0 1 2 1 3 1 maxima
0 0 2 1 2 1 1 0
removed
0 1 3 2 1 1 0 0
0 3 2 1 0 0 1 0 depends
on condition
2 3 2 0 0 1 0 1
2 3 2 0 1 0 2 1

Gonzalez & Woods
0 0 0 0Segmentation
0 0
Ch t 10
Chapter
0 3
0 0 0 0 2 1 3 0
0 0 2 1 2 0 0 0
0 0 3 0 0 0 0 0 false edges
g
0 3 2 0 0 0 0 0
0 3 0 0 0 1 0 1
0 3 0 0 1 0 2 0
• The suppressed magnitude image will contain many
false edges caused by noise or fine texture
Thresholding
g
Gonzalez & Woods
Ch t 10
Chapter
Segmentation
• Reduce number of false edges by applying a

threshold T
– all values below T are changed to 0
– selecting a good values for T is difficult
– some false edges will remain if T is too low
– some edges will disappear if T is too high
– some edges
d will
ill disappear
di due
d tot softening
ft i off the
th
edge contrast by shadows

Double Thresholding
g
Gonzalez & Woods
Ch t 10
Chapter
Segmentation
• Apply two thresholds in the suppressed image

– T2 = 2T2
– two images
g in the output
p
– the image from T2 contains fewer edges but has gaps in the
contours
– the
h image
i from
f T1 has
h many false
f l edges
d
– combine the results from T1 and T2
– link the edges of T2 into contours until we reach a gap
– link the edge from T2 with edge pixels from a T1 contour
until a T2 edge is found again

Gonzalez & Woods
T2=22 Ch t 10
Chapter T1=1
Segmentation
0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 3
0 0 0 0 0 2 0 3 0 0 0 0 2 1 3 0
0 0 2 0 0 2 0 0 0 gaps 0 2 1 2 0 0 0
0 0 3 0 0 0 0 0 0 filled 0 3 0 0 0 0 0
0 3 2 0 0 0 0 0 from 0 3 2 0 0 0 0 0
T1 0 3 0 0 0 1 0 1
0 3 0 0 0 0 0 0
0 3 0 0 0 0 2 0 0 3 0 0 1 0 2 0
• A T2 contour has pixels along the green arrows

• Linking: search in a 3x3 of each pixel and connect the
pixel at the center with the one having greater value
• Search
S h iin the
th direction
di ti off the
th edge
d (direction
(di ti off Gradient)
G di t)
Line Detection
Gonzalez & Woods
Ch t 10
Chapter
Segmentation
• Model of a line: two edges with opposite

polarity in distance less than the size of the
smoothing g filter
– edge detection filters respond to step edges
– they do not provide meaningful response to lines
• Apply nonmaxima suppression on the
smoothed output
– a line is the derivative of a step  the derivative
step of the Canny algorithm is not necessary

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Thresholding
1. Select initial estimate
2. Segment the images using T into two
groups G1 and G2
3. Compute the averages of the two
groups m1 and m2.
4. Computer a new threshold as
T = (m1+m2)/2
5. Repeat 2..4 until the difference
between two successive thresholds is
less than t.

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Optimal Thresholding
L 1
p
i 0
i 1
Suppose we select a thershold T(k)

k
P1 (k )   pi
i 0
L 1
P2 (k )  p
i  k 1
i  1  P1 (k )
The mean intensity in each group is

k k
1 k
m1 (k )   iP (i C1 )   iP (C1 i ) / P(i C1 )   iP (i )
i 0 i 0 P1 (k ) i 0
P ( A B )  P( B A) P( A) / P ( B)

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
The mean intensity of the second group is
L 1 L 1
1 L 1
m2 (k )   iP (i C2 )   iP(C2 i ) / P(i C2 )   iP (i )
i  k 1 i  k 1 P2 (k ) i  k 1
The accumulative mean up to k The average intensity of the entire
level is k
g is:
image L 1
m(kk))   ip
i i mG (k )   ipi
i 0 i 0
L 1
 mG  P1m1  P2 m2     (i  m) 2 pi
2
G
i 0
 2

mG P1 (k )  m(k )
2
P1 (k )[1  P1 (k )]
B

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
We can measure goodness of the threshold at level k is
 B2
 2
G
L 1
Where    (i  mG ) 2 pi
2
G
i 0
and (mG P1  m) 2
  P1 (m1  mG )  P2 (m2  mG )  P1 P2 (m1  m2 ) 
2 2 2 2
P1 (1  P2 )
B
Taking k into account
 (k ) 
 B2 ((kk )
Where  2
(k ) 
mG P1 (k )  m(k )
2
 G2 B
P1 (k )[1  P1 (k )]
The optimum threshold k* is
 B2 (k * )  max ( B2 (k ))
0 k  L 1

Gonzalez & Woods
Ch t 10
Chapter
Segmentation
Otsu’ss Algorithm for Optimal Threshold

Otsu
1. Compute the normalized histogram (pi) of the input image
2. Compute the accumulative sum Pk for k=0, .., L-1
3
3. C
Compute the
h accumulative
l i mean mk forf kk=0,
0 .., L-1
L1
4. Compute the global intensity mean mG 2
5. Compute the between class variance  B (k )
6
6.  2
Obtain the Otus threshold k* as the value of k that maximize B (k )
7. Obtain the separablility measure  at k = k*

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

Gonzalez & Woods
Ch t 10
Chapter
Segmentation

DIP3E Chapter10 Art

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

DIP3E Chapter10 Art

Caricato da

Copyright:

Formati disponibili

Digital Image Processing, 3rd ed.

Gonzalez & Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

Detection of Isolate Points

Applying Laplacian on an image.

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

Gradient is often expressed as

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

Marr-Hildreth Edge Detector

© 1992–2008 R. C. Gonzalez & R. E. Woods

Canny Edge Detector

© 1992–2008 R. C. Gonzalez & R. E. Woods

Non-Maxima Suppression Gonzalez & Woods

• Thin edges by keeping large values of Gradient

Canny Edge Detector 20

Non-Maxima Suppression (2) Gonzalez & Woods

Canny Edge Detector 21

• Reduce angle of Gradient θ[i,j] to one of the 4 sectors

Canny Edge Detector 22

Canny Edge Detector 23

• Reduce number of false edges by applying a

Canny Edge Detector 25

• Apply two thresholds in the suppressed image

Canny Edge Detector 26

• A T2 contour has pixels along the green arrows

• Model of a line: two edges with opposite

Canny Edge Detector 28

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

Suppose we select a thershold T(k)

The mean intensity in each group is

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

Taking k into account

© 1992–2008 R. C. Gonzalez & R. E. Woods

Otsu’ss Algorithm for Optimal Threshold

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods

© 1992–2008 R. C. Gonzalez & R. E. Woods