Wavelet Transform
A very brief look
2
Wavelets vs. Fourier Transform
 In Fourier transform (FT) we represent a
signal in terms of sinusoids
 FT provides a signal which is localized
only in the frequency domain
 It does not give any information of the
signal in the time domain
3
Wavelets vs. Fourier Transform
 Basis functions of the wavelet transform
(WT) are small waves located in different
times
 They are obtained using scaling and
translation of a scaling function and
wavelet function
 Therefore, the WT is localized in both time
and frequency
4
Wavelets vs. Fourier Transform
 In addition, the WT provides a
multiresolution system
 Multiresolution is useful in several
applications
 For instance, image communications and
image data base are such applications
5
Wavelets vs. Fourier Transform
 If a signal has a discontinuity, FT produces
many coefficients with large magnitude
(significant coefficients)
 But WT generates a few significant
coefficients around the discontinuity
 Nonlinear approximation is a method to
benchmark the approximation power of a
transform
6
Wavelets vs. Fourier Transform
 In nonlinear approximation we keep only a few
significant coefficients of a signal and set the
rest to zero
 Then we reconstruct the signal using the
significant coefficients
 WT produces a few significant coefficients for
the signals with discontinuities
 Thus, we obtain better results for WT nonlinear
approximation when compared with the FT
7
Wavelets vs. Fourier Transform
 Most natural signals are smooth with a few
discontinuities (are piece-wise smooth)
 Speech and natural images are such signals
 Hence, WT has better capability for representing
these signal when compared with the FT
 Good nonlinear approximation results in
efficiency in several applications such as
compression and denoising
8
Series Expansion of Discrete-Time Signals
 Suppose that is a square-summable sequence, that
is
 Orthonormal expansion of is of the form
 Where
is the transform of
 The basis functions satisfy the orthonormality
constraint
[ ]
x n
2
[ ] ( )
x n  Z
[ ]
x n
[ ] [ ], [ ] [ ] [ ] [ ]
k k k
k k
x n l x l n X k n
  
 
 
 
Z Z
*
[ ] [ ], [ ] [ ] [ ]
k k
l
X k l x l n x l
 
  
[ ]
x n
k

[ ], [ ] [ ]
k l
n n k l
  
 
2 2
x X

9
 Haar expansion is a two-point avarage
and difference operation
 The basis functions are given as
 It follows that
Haar Basis
2
1 2 , 2 , 2 1
[ ]
0, otherwise
k
n k k
n

  
 

2 1
1 2 , 2
[ ] 1 2 , 2 1
0, otherwise
k
n k
n n k
 
 

   



2 0
[ ] [ 2 ],
k n n k
 
  2 1 1
[ ] [ 2 ]
k n n k
 
  
10
 The transform is
 The reconstruction is obtained from
Haar Basis
 
2
1
[2 ] , [2 ] [2 1] ,
2
k
X k x x k x k

   
[ ] [ ] [ ]
k
k
x n X k n


 
Z
 
2 1
1
[2 1] , [2 ] [2 1]
2
k
X k x x k x k
 
    
11
Two-Channel Filter Banks
 Filter bank is the building block of discrete-
time wavelet transform
 For 1-D signals, two-channel filter bank is
depicted below
12
Two-Channel Filter Banks
 For perfect reconstruction filter banks we have
 In order to achieve perfect reconstruction the
filters should satisfy
 Thus if one filter is lowpass, the other one will be
highpass
x̂ x

0 0
1 1
[ ] [ ]
[ ] [ ]
g n h n
g n h n
  

  

13
Two-Channel Filter Banks
14
Two-Channel Filter Banks
 To have orthogonal wavelets, the filter bank
should be orthogonal
 The orthogonal condition for 1-D two-channel
filter banks is
 Given one of the filters of the orthogonal filter
bank, we can obtain the rest of the filters
1 0
[ ] ( 1) [ 1]
n
g n g n
   
15
Haar Filter Bank
 The simplest orthogonal filter bank is Haar
 The lowpass filter is
 And the highpass filter
0
1
, 0, 1
[ ] 2
0, otherwise
n
h n

 

 


1
1
, 0
2
1
[ ] , 1
2
0, otherwise
n
h n n





   





16
Haar Filter Bank
 The lowpass output is
 And the highpass output is
0 0 0
2
1 1
[ ] [ ]* [ ] [ ] [2 ] [2 ] [2 1]
2 2
n k
l
y k h n x n h l x k l x k x k


     

1 1 1
2
1 1
[ ] [ ]* [ ] [ ] [2 ] [2 ] [2 1]
2 2
n k
l
y k h n x n h l x k l x k x k


     

17
Haar Filter Bank
 Since and , the filter
bank implements Haar expansion
 Note that the analysis filters are time-reversed
versions of the basis functions
since convolution is an inner product followed by
time-reversal
0[ ] [2 ]
y k X k
 1[ ] [2 1]
y k X k
 
0 0
[ ] [ ]
h n n

  1 1
[ ] [ ]
h n n

 
18
Discrete Wavelet Transform
 We can construct discrete WT via iterated (octave-band) filter banks
 The analysis section is illustrated below
Level 1
Level 2
Level J
19
Discrete Wavelet Transform
 And the synthesis section is illustrated here
 If is an orthogonal filter and , then we have an
orthogonal wavelet transform
0
V
1
V
2
V
J
V
1
W
2
W
J
W
[ ]
i
h n [ ] [ ]
i i
g n h n
 
20
Multiresolution
 We say that is the space of all square-
summable sequences if
 Then a multiresolution analysis consists of
a sequence of embedded closed spaces
 It is obvious that
0
V
0 2( )
V 
2 1 0 2( )
J
V V V V
    
0 2
0
( )
J
j
j
V V

 
21
Multiresolution
 The orthogonal component of in will
be denoted by :
 If we split and repeat on , , …., ,
we have
1
j
V 
1 1
j j j
V V W
 
 
j
V
1
j
W 
1 1
j j
V W
 

0
V 1
V 2
V 1
J
V 
0 1 1 J J
V W W W V
    
22
2-D Separable WT
 For images we use separable WT
 First we apply a 1-D filter bank to the rows of the
image
 Then we apply same transform to the columns of
each channel of the result
 Therefore, we obtain 3 highpass channels
corresponding to vertical, horizontal, and
diagonal, and one approximation image
 We can iterate the above procedure on the
lowpass channel
23
2-D Analysis Filter Bank
1
h
0
h
1
h
1
h
0
h
0
h
x diagonal
vertical
horizontal
approximation
24
2-D Synthesis Filter Bank
x̂
diagonal
vertical
horizontal
approximation
1
g
1
g
1
g
0
g
0
g
0
g
25
2-D WT Example
Boats image WT in 3 levels
26
WT-Application in Denoising
Boats image Noisy image (additive Gaussian noise)
27
WT-Application in Denoising
Boats image Denoised image using hard thresholding
28
Reference
 Martin Vetterli and Jelena Kovacevic, Wavelets and
Subband Coding. Prentice Hall, 1995.

Wavelet Transform.ppt

  • 1.
  • 2.
    2 Wavelets vs. FourierTransform  In Fourier transform (FT) we represent a signal in terms of sinusoids  FT provides a signal which is localized only in the frequency domain  It does not give any information of the signal in the time domain
  • 3.
    3 Wavelets vs. FourierTransform  Basis functions of the wavelet transform (WT) are small waves located in different times  They are obtained using scaling and translation of a scaling function and wavelet function  Therefore, the WT is localized in both time and frequency
  • 4.
    4 Wavelets vs. FourierTransform  In addition, the WT provides a multiresolution system  Multiresolution is useful in several applications  For instance, image communications and image data base are such applications
  • 5.
    5 Wavelets vs. FourierTransform  If a signal has a discontinuity, FT produces many coefficients with large magnitude (significant coefficients)  But WT generates a few significant coefficients around the discontinuity  Nonlinear approximation is a method to benchmark the approximation power of a transform
  • 6.
    6 Wavelets vs. FourierTransform  In nonlinear approximation we keep only a few significant coefficients of a signal and set the rest to zero  Then we reconstruct the signal using the significant coefficients  WT produces a few significant coefficients for the signals with discontinuities  Thus, we obtain better results for WT nonlinear approximation when compared with the FT
  • 7.
    7 Wavelets vs. FourierTransform  Most natural signals are smooth with a few discontinuities (are piece-wise smooth)  Speech and natural images are such signals  Hence, WT has better capability for representing these signal when compared with the FT  Good nonlinear approximation results in efficiency in several applications such as compression and denoising
  • 8.
    8 Series Expansion ofDiscrete-Time Signals  Suppose that is a square-summable sequence, that is  Orthonormal expansion of is of the form  Where is the transform of  The basis functions satisfy the orthonormality constraint [ ] x n 2 [ ] ( ) x n  Z [ ] x n [ ] [ ], [ ] [ ] [ ] [ ] k k k k k x n l x l n X k n          Z Z * [ ] [ ], [ ] [ ] [ ] k k l X k l x l n x l      [ ] x n k  [ ], [ ] [ ] k l n n k l      2 2 x X 
  • 9.
    9  Haar expansionis a two-point avarage and difference operation  The basis functions are given as  It follows that Haar Basis 2 1 2 , 2 , 2 1 [ ] 0, otherwise k n k k n        2 1 1 2 , 2 [ ] 1 2 , 2 1 0, otherwise k n k n n k             2 0 [ ] [ 2 ], k n n k     2 1 1 [ ] [ 2 ] k n n k     
  • 10.
    10  The transformis  The reconstruction is obtained from Haar Basis   2 1 [2 ] , [2 ] [2 1] , 2 k X k x x k x k      [ ] [ ] [ ] k k x n X k n     Z   2 1 1 [2 1] , [2 ] [2 1] 2 k X k x x k x k       
  • 11.
    11 Two-Channel Filter Banks Filter bank is the building block of discrete- time wavelet transform  For 1-D signals, two-channel filter bank is depicted below
  • 12.
    12 Two-Channel Filter Banks For perfect reconstruction filter banks we have  In order to achieve perfect reconstruction the filters should satisfy  Thus if one filter is lowpass, the other one will be highpass x̂ x  0 0 1 1 [ ] [ ] [ ] [ ] g n h n g n h n        
  • 13.
  • 14.
    14 Two-Channel Filter Banks To have orthogonal wavelets, the filter bank should be orthogonal  The orthogonal condition for 1-D two-channel filter banks is  Given one of the filters of the orthogonal filter bank, we can obtain the rest of the filters 1 0 [ ] ( 1) [ 1] n g n g n    
  • 15.
    15 Haar Filter Bank The simplest orthogonal filter bank is Haar  The lowpass filter is  And the highpass filter 0 1 , 0, 1 [ ] 2 0, otherwise n h n         1 1 , 0 2 1 [ ] , 1 2 0, otherwise n h n n              
  • 16.
    16 Haar Filter Bank The lowpass output is  And the highpass output is 0 0 0 2 1 1 [ ] [ ]* [ ] [ ] [2 ] [2 ] [2 1] 2 2 n k l y k h n x n h l x k l x k x k          1 1 1 2 1 1 [ ] [ ]* [ ] [ ] [2 ] [2 ] [2 1] 2 2 n k l y k h n x n h l x k l x k x k         
  • 17.
    17 Haar Filter Bank Since and , the filter bank implements Haar expansion  Note that the analysis filters are time-reversed versions of the basis functions since convolution is an inner product followed by time-reversal 0[ ] [2 ] y k X k  1[ ] [2 1] y k X k   0 0 [ ] [ ] h n n    1 1 [ ] [ ] h n n   
  • 18.
    18 Discrete Wavelet Transform We can construct discrete WT via iterated (octave-band) filter banks  The analysis section is illustrated below Level 1 Level 2 Level J
  • 19.
    19 Discrete Wavelet Transform And the synthesis section is illustrated here  If is an orthogonal filter and , then we have an orthogonal wavelet transform 0 V 1 V 2 V J V 1 W 2 W J W [ ] i h n [ ] [ ] i i g n h n  
  • 20.
    20 Multiresolution  We saythat is the space of all square- summable sequences if  Then a multiresolution analysis consists of a sequence of embedded closed spaces  It is obvious that 0 V 0 2( ) V  2 1 0 2( ) J V V V V      0 2 0 ( ) J j j V V   
  • 21.
    21 Multiresolution  The orthogonalcomponent of in will be denoted by :  If we split and repeat on , , …., , we have 1 j V  1 1 j j j V V W     j V 1 j W  1 1 j j V W    0 V 1 V 2 V 1 J V  0 1 1 J J V W W W V     
  • 22.
    22 2-D Separable WT For images we use separable WT  First we apply a 1-D filter bank to the rows of the image  Then we apply same transform to the columns of each channel of the result  Therefore, we obtain 3 highpass channels corresponding to vertical, horizontal, and diagonal, and one approximation image  We can iterate the above procedure on the lowpass channel
  • 23.
    23 2-D Analysis FilterBank 1 h 0 h 1 h 1 h 0 h 0 h x diagonal vertical horizontal approximation
  • 24.
    24 2-D Synthesis FilterBank x̂ diagonal vertical horizontal approximation 1 g 1 g 1 g 0 g 0 g 0 g
  • 25.
    25 2-D WT Example Boatsimage WT in 3 levels
  • 26.
    26 WT-Application in Denoising Boatsimage Noisy image (additive Gaussian noise)
  • 27.
    27 WT-Application in Denoising Boatsimage Denoised image using hard thresholding
  • 28.
    28 Reference  Martin Vetterliand Jelena Kovacevic, Wavelets and Subband Coding. Prentice Hall, 1995.