Eg. Median Filter - Hanyang University

advertisement
Computer Vision –
Enhancement(Part II)
Hanyang University
Jong-Il Park
Local Enhancement
 Global enhancement
 The same operation for all pixels
 Local enhancement
 Different operation for each pixel
 According to the statistics of local support
Department of Computer Science and Engineering, Hanyang University
Local Histogram Equalization
 Using a fixed window at each point
 Computationally expensive
 Histogram equalization at each point
Department of Computer Science and Engineering, Hanyang University
Use of statistics of local support
 Eg.
E  f ( x, y)
g ( x, y)  
 f ( x, y)
m
Original
image
if m  k0 M G AND k1DG    k2 DG
otherwise

E
Enhanced
image
Department of Computer Science and Engineering, Hanyang University
Spatial Operations
 Spatial averaging and spatial LPF for noise
smoothing
Input
image
output
*
Spatial mask
( 33, 55, )
Department of Computer Science and Engineering, Hanyang University
Spatial Mask
Department of Computer Science and Engineering, Hanyang University
Spatial Averaging
 Mean-filtering
v(m, n)   a (k , l ) y (m  k , n  l )
( k .l ) W
ak , l  : filtercoefficients
equal weight filter: ak , l  

1
, N w  2M  1 2 N  1
Nw
Noise reduction
v(m, n)  u(m, n)  (m, n)
(m, n) ~ N (0, )
2
a(k , l ) 
1
Nw
1
v(m, n) 
 u (m  k , n  l )   (m, n)
N w ( k ,l ) W
 (m, n) ~ N (0,   ), where   
2
2
 2
Nw
Department of Computer Science and Engineering, Hanyang University
Spatial Averaging Mask

Spatial averaging masks a(k,l)

Disadvantage : blurring
Department of Computer Science and Engineering, Hanyang University
Effect of window size
Department of Computer Science and Engineering, Hanyang University
Eg. Spatial Averaging(1)
Department of Computer Science and Engineering, Hanyang University
Eg. Spatial Averaging(2)
Original image
Averaging 후의 image
Department of Computer Science and Engineering, Hanyang University
Cf. Multi-image averaging
Department of Computer Science and Engineering, Hanyang University
Spatial Operations - Filtering
 Parametric Low Pass Filter
1 b
 1  
2
H 
b
b

b  2  
1 b
2

but
 h
ij
i
1
1

b
1 
to preserve the mean
j
Department of Computer Science and Engineering, Hanyang University
Spatial LPF, BPF, HPF
hHP (m, n)   (m, n)  hLP (m, n)
hBP (m, n)  hL (m, n)  hL (m, n)
1
u (m, n)
Spatial
averaging
2
u (m, n)
VLP (m, n)
LPF
+
(a) Spatial low-pass filter
u (m, n)
VHP (m, n)
(b) Spatial high-pass filter
LPF
+
VBP (m, n)
hL (m, n)
1
LPF
hL (m, n)
2
(c) Spatial band-pass filter
Department of Computer Science and Engineering, Hanyang University
Eg. Spatial LPF
Original image
Lowpass Filter된 후의 image
Department of Computer Science and Engineering, Hanyang University
Spatial High-Pass Filtering
Department of Computer Science and Engineering, Hanyang University
Eg. Spatial HPF
Original image
Highpass filtered image
Department of Computer Science and Engineering, Hanyang University
Spatial Band-Pass Filtering
Original image
Lowpass Filter(Short Term) =A
Bandpass Filter된
후의 Image =B-A
Lowpass Filter(Long Term) =B
Department of Computer Science and Engineering, Hanyang University
Denoising by LPF
Noisy!
Blurred!
Trade-off?
Department of Computer Science and Engineering, Hanyang University
Directional Smoothing
 Directional Smoothing
 to protect the edges from blurring while smoothing
spatialaveragesvm, n,  are calculatedin severaldirection
1
v(m, n :  ) 
N
W
 y(m  k , n  l )
( k ,l ) W
 v(m, n)  v(m, n :   )
Find out  
such that min | y(m, n)  v(m, n :   ) |


























l
k
Department of Computer Science and Engineering, Hanyang University
Eg. Directional Smoothing
Original image
Lowpass Filter(Long Term)
Direc. Smoothing (대각선)
Direc. Smoothing (수 직)
Department of Computer Science and Engineering, Hanyang University
Median Filtering
 Median Filter
v(m, n)  median { y(m  k , n  l ), (k , l ) W }
filter length should be odd number

Properties
 nonlinear filter
median {x( m)  y ( m)}  median {x(m)}  median { y ( m)}

Example
 y  m   2,3,8, 4, 2 , 3 1 median filter
v  0   2  boundary value  , v 1  median 2,3,8  3
v  2   median 3,8, 4  4, v  3  median 8, 4, 2  4
v  5  2  boundary value 
v  m   2,3, 4, 4, 2
Numberof operations 3( NW2 1) / 8 comparisons
Department of Computer Science and Engineering, Hanyang University
Eg. 1D Median Filtering
L  5 1
Department of Computer Science and Engineering, Hanyang University
Discussion – Median filter
1) median filter preserve discontinuities in a step function
2) smooth a few pixels whose values differ significantly
from the surrounding, without affecting the other pixels.
3) pulse function, whose width is less than one half the
filter length, are suppressed
Department of Computer Science and Engineering, Hanyang University
2D Median Filtering
Filter
Filtered Image
Original Image
Filter
Filtered Image
Department of Computer Science and Engineering, Hanyang University
Eg. Median Filtering
Salt-and-pepper noise(=impulsive noise)
Original
7x7 Median filtered image
 Excellent performance!
Department of Computer Science and Engineering, Hanyang University
Eg. Median Filter – Impulsive Noise
Department of Computer Science and Engineering, Hanyang University
Eg. Median Filter – Impulsive Noise
Department of Computer Science and Engineering, Hanyang University
Eg. Median Filter – Gaussian Noise
Moderate
performance
Department of Computer Science and Engineering, Hanyang University
Various patterns for median filter
Neighborhood patterns used for median filtering
Department of Computer Science and Engineering, Hanyang University
Eg. Median filter – Square pattern
Original image
10% black, 10% white
Median filtering using
3 by 3 square region
Median filtering using
5 by 5 square region
Department of Computer Science and Engineering, Hanyang University
Eg. Median filter – Octagon pattern
Original image
5 by 5 octagonal median filter
Department of Computer Science and Engineering, Hanyang University
Eg. Median filter – Reconstruction
Original image
Median filtering and color compensation
Department of Computer Science and Engineering, Hanyang University
Sharpening Images
 Emphasis of high-frequency components
 Usually exploiting 1st order derivative and 2nd order
derivatives
 1D derivatives
f
st
 f ( x  1)  f ( x)
 1 order derivative:
x

2

f
nd
 f ( x  1)  f ( x  1)  2 f ( x)
2 order derivative:
2
x
Department of Computer Science and Engineering, Hanyang University
Eg. 1st & 2nd order derivatives
Department of Computer Science and Engineering, Hanyang University
Observation on derivatives
 2nd order derivative
 Thinner edges
 Stronger response to fine details
 Weaker response to a gray-level step
 Double response at step changes
 Intensity of response: point > line > step
 The 2nd order derivative is better suited than the 1st
order derivative for image enhancement.
Department of Computer Science and Engineering, Hanyang University
Laplacian Operator – Derivation
 The simplest isotropic derivative operator
2
2

f

f
2 f 

x 2
y 2
 x f (i, j )   x f (i  1, j )   x f (i, j )
2
 [ f (i  1, j )  f (i, j )]  [ f (i, j )  f (i  1, j )]
 f (i  1, j )  f (i  1, j )  2 f (i, j )
 y f (i, j )  f (i, j  1)  f (i, j  1)  2 f (i, j )
2
  2 f (i, j )   x f (i, j )   y f (i, j )
2
2
 [ f (i  1, j )  f (i  1, j )  f (i, j  1)  f (i, j  1)]  4 f (i, j )
Department of Computer Science and Engineering, Hanyang University
Laplacian Operator
 0 1 0 
H1   1 4  1
 0  1 0 
 1  1  1
H 2   1 8  1
 1  1  1
Eg)
0
0
0
1 2 3 4
x f
0
0
1
0 0 0 0 1 0 0 0
2
f  x f
2
0 0 1 1 2 3 4
5
 1 2 1 
H 3   2 4  2
 1  2 1 
5 5 5
6 5 5
5
Department of Computer Science and Engineering, Hanyang University
Sharpening by Laplacian operator
Department of Computer Science and Engineering, Hanyang University
Eg. Sharpening
Original SEM image
Laplacian operator
Original image
Laplacian operator
Subtraction of the
Laplacian from the original
Subtraction of the
Laplacian from the original
Department of Computer Science and Engineering, Hanyang University
Composite Laplacian mask
Department of Computer Science and Engineering, Hanyang University
Unsharp masking and Crispening
v(m, n)  u (m, n)  g (m, n),   0
where g (m, n) : Laplacianoutput (high pass filtered )
1
g (m, n)  u (m, n)  (u (m  1, n)  u (m, n  1)  u (m  1, n)  u (m, n  1))
4
(1)
Signal
(2)
Low-pass
(3)
High-pass
(1)+(3)
Department of Computer Science and Engineering, Hanyang University
Unsharp mask application
Original image
Processed image
Department of Computer Science and Engineering, Hanyang University
High-boost filtering
Let g(n1, n2) = u(n1, n2) - uL(n1, n2)
v(n1, n2) = u(n1, n2) + k g(n1, n2)
 k=1: Unsharp Masking
 Crispening an image
 k>1: High-boost filtering
 edge or line details to be emphasized
Department of Computer Science and Engineering, Hanyang University
Eg. High-boost filtering
Department of Computer Science and Engineering, Hanyang University
Zoom(1:2 magnification) revisited
 Nearest neighbor=Replication = zero - order hold
1
0
1
3
2



4 5 6  4



0
0
0
0
0
3
0
5
0
0
0
0
0
2
0
6
0
0
0

0

0
1
1
1
1
1
1
 
4

4
1
1
4
4
3
3
5
5
3
3
5
5
2
2
6
6
2
2
6

6
column, row zero-padding
Department of Computer Science and Engineering, Hanyang University
Zoom revisited(cont.)
 Linear Interpolation : first - order hold
1
0
1 7

3 1   3



0
0 7 0
1 4
0 0
0 0 0 
 
3 2
0 1 0


0 0 0
0 0



H  




1
4
1
2
1
4
1
2
1
1
2
1
4
1
2
1
4
7 3.5 
 1
 2
0 0 
 
 3
1 0.5 


0 0 
 1.5
3.5 
3 4
2 
2 1 0.5 

1 0.5 0.25 
4
7








Department of Computer Science and Engineering, Hanyang University
Download