wavelet
-
Upload
chandan-bangera -
Category
Documents
-
view
213 -
download
0
Transcript of wavelet
![Page 1: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/1.jpg)
Wavelets (Chapter 7)
CS474/674 – Prof. Bebis
![Page 2: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/2.jpg)
STFT (revisited)
• Time/Frequency localization depends on window size.• Once you choose a particular window size, it will be the same
for all frequencies.• Many signals require a more flexible approach - vary the
window size to determine more accurately either time or frequency.
![Page 3: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/3.jpg)
The Wavelet Transform
• Overcomes the preset resolution problem of the STFT by using a variable length window:
– Use narrower windows at high frequencies for better time resolution.
– Use wider windows at low frequencies for better frequency resolution.
![Page 4: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/4.jpg)
The Wavelet Transform (cont’d)
Wide windows do not provide good localization at high frequencies.
![Page 5: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/5.jpg)
The Wavelet Transform (cont’d)
Use narrower windows at high frequencies.
![Page 6: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/6.jpg)
The Wavelet Transform (cont’d)
Narrow windows do not provide good localization at low frequencies.
![Page 7: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/7.jpg)
The Wavelet Transform (cont’d)
Use wider windows at low frequencies.
![Page 8: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/8.jpg)
What are Wavelets?• Wavelets are functions that “wave” above and below the
x-axis, have (1) varying frequency, (2) limited duration, and (3) an average value of zero.
• This is in contrast to sinusoids, used by FT, which have infinite energy.
Sinusoid Wavelet
![Page 9: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/9.jpg)
What are Wavelets? (cont’d)• Like sines and cosines in FT, wavelets are used as basis
functions ψk(t) in representing other functions f(t):
• Span of ψk(t): vector space S containing all functions f(t) that can be represented by ψk(t).
( ) ( )k kk
f t a t
![Page 10: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/10.jpg)
What are Wavelets? (cont’d)
• There are many different wavelets:
MorletHaar Daubechies
![Page 11: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/11.jpg)
(dyadic/octave grid)
What are Wavelets? (cont’d)
( ) jk t
![Page 12: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/12.jpg)
What are Wavelets? (cont’d)
time localization
scale/frequency localization
/2( ) 2 2 j jjk t t k
j
![Page 13: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/13.jpg)
Continuous Wavelet Transform (CWT)
1( , )t
tC s f t dtss
Continuous Wavelet Transform of signal f(t)
translation parameter, measure of time
scale parameter (measure of frequency)
Mother wavelet (window)
normalization constant
ForwardCWT:
Scale = 1/j = 1/Frequency
![Page 14: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/14.jpg)
CWT: Main Steps
1. Take a wavelet and compare it to a section at the start of the original signal.
2. Calculate a number, C, that represents how closely correlated the wavelet is with this section of the signal. The higher C is, the more the similarity.
![Page 15: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/15.jpg)
CWT: Main Steps (cont’d)
3. Shift the wavelet to the right and repeat steps 1 and 2 until you've covered the whole signal.
![Page 16: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/16.jpg)
CWT: Main Steps (cont’d)
4. Scale the wavelet and repeat steps 1 through 3.
5. Repeat steps 1 through 4 for all scales.
![Page 17: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/17.jpg)
Coefficients of CTW Transform
1( , )t
tC s f t dtss
• Wavelet analysis produces a time-scale view of the input signal or image.
![Page 18: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/18.jpg)
Continuous Wavelet Transform (cont’d)
• Inverse CWT:
1( ) ( , ) ( )s
tf t C s d dsss
double integral!
![Page 19: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/19.jpg)
FT vs WT
weighted by F(u)
weighted by C(τ,s)
![Page 20: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/20.jpg)
Properties of Wavelets
• Simultaneous localization in time and scale- The location of the wavelet allows to explicitly represent
the location of events in time.- The shape of the wavelet allows to represent different
detail or resolution.
![Page 21: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/21.jpg)
Properties of Wavelets (cont’d)
• Sparsity: for functions typically found in practice, many of the coefficients in a wavelet representation are either zero or very small.
• Linear-time complexity: many wavelet transformations can be accomplished in O(N) time.
1( ) ( , ) ( )s
tf t C s d dsss
![Page 22: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/22.jpg)
Properties of Wavelets (cont’d)
• Adaptability: wavelets can be adapted to represent a wide variety of functions (e.g., functions with discontinuities, functions defined on bounded domains etc.).– Well suited to problems involving images, open or closed
curves, and surfaces of just about any variety.– Can represent functions with discontinuities or corners more
efficiently (i.e., some have sharp corners themselves).
![Page 23: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/23.jpg)
Discrete Wavelet Transform (DWT)
( ) ( )jk jkk j
f t a t
/2( ) 2 2 j jjk t t k
(inverse DWT)
(forward DWT)
where
*( ) ( )jkjkt
a f t t
![Page 24: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/24.jpg)
DFT vs DWT
• FT expansion:
• WT expansion
or
one parameter basis
( ) ( )l ll
f t a t
( ) ( )jk jkk j
f t a t
two parameter basis
![Page 25: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/25.jpg)
Multiresolution Representation using
( ) ( )jk jkk j
f t a t
( )f t
( )jk t
j
finedetails
coarsedetails
wider, large translations
![Page 26: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/26.jpg)
Multiresolution Representation using
( ) ( )jk jkk j
f t a t
( )f t
( )jk t
j
finedetails
coarsedetails
![Page 27: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/27.jpg)
Multiresolution Representation using
( ) ( )jk jkk j
f t a t
( )f t
( )jk t
j
finedetails
coarsedetails
narrower, small translations
![Page 28: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/28.jpg)
Multiresolution Representation using
high resolution (more details)
low resolution (less details)
…
( ) ( )jk jkk j
f t a t
( )f t
1̂( )f t
2̂ ( )f t
ˆ ( )sf t
( )jk t
j
![Page 29: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/29.jpg)
Prediction Residual Pyramid (revisited)
• In the absence of quantization errors, the approximation pyramid can be reconstructed from the prediction residual pyramid.
• Prediction residual pyramid can be represented more efficiently.
(with sub-sampling)
![Page 30: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/30.jpg)
Efficient Representation Using “Details”
details D2
L0
details D3
details D1
(no sub-sampling)
![Page 31: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/31.jpg)
Efficient Representation Using Details (cont’d)
representation: L0 D1 D2 D3
A wavelet representation of a function consists of (1)a coarse overall approximation (2)detail coefficients that influence the function at various scales.
(decompositionor analysis)in general: L0 D1 D2 D3…DJ
![Page 32: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/32.jpg)
Reconstruction (synthesis)H3=L2+D3
details D2
L0
details D3
H2=L1+D2
H1=L0+D1
details D1
(no sub-sampling)
![Page 33: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/33.jpg)
Example - Haar Wavelets
• Suppose we are given a 1D "image" with a resolution of 4 pixels:
[9 7 3 5]
• The Haar wavelet transform is the following:
L0 D1 D2 D3(with sub-sampling)
![Page 34: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/34.jpg)
Example - Haar Wavelets (cont’d)
• Start by averaging the pixels together (pairwise) to get a new lower resolution image:
• To recover the original four pixels from the two averaged pixels, store some detail coefficients.
1
![Page 35: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/35.jpg)
Example - Haar Wavelets (cont’d)
• Repeating this process on the averages gives the full decomposition:
1
![Page 36: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/36.jpg)
Example - Haar Wavelets (cont’d)
• The Harr decomposition of the original four-pixel image is:
• We can reconstruct the original image to a resolution by adding or subtracting the detail coefficients from the lower-resolution versions.
2 1 -1
![Page 37: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/37.jpg)
Example - Haar Wavelets (cont’d)
Note small magnitudedetail coefficients!
Dj
Dj-1
D1L0
How tocompute Di ?
![Page 38: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/38.jpg)
Multiresolution Conditions• If a set of functions can be represented by a weighted
sum of ψ(2jt - k), then a larger set, including the original, can be represented by a weighted sum of ψ(2j+1t - k):
time localization
scale/frequency localization
low resolution
high resolution
j
![Page 39: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/39.jpg)
Multiresolution Conditions (cont’d)• If a set of functions can be represented by a weighted
sum of ψ(2jt - k), then a larger set, including the original, can be represented by a weighted sum of ψ(2j+1t - k):
Vj: span of ψ(2jt - k): ( ) ( )j k jkk
f t a t
Vj+1: span of ψ(2j+1t - k): 1 ( 1)( ) ( )j k j kk
f t b t
1j jV V
![Page 40: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/40.jpg)
Nested Spaces Vj
ψ(t - k)
ψ(2t - k)
ψ(2jt - k)
…
V0
V1
Vj
Vj : space spanned by ψ(2jt - k)
Multiresolution conditions nested spanned spaces:
( ) ( )jk jkk j
f t a t f(t) ϵ Vj
Basis functions:
i.e., if f(t) ϵ V j then f(t) ϵ V j+1
1j jV V
![Page 41: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/41.jpg)
How to compute Di ? (cont’d)
( ) ( )jk jkk j
f t a t f(t) ϵ Vj
IDEA: define a set of basisfunctions that span thedifferences between Vj
![Page 42: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/42.jpg)
Orthogonal Complement Wj
• Let Wj be the orthogonal complement of Vj in Vj+1
Vj+1 = Vj + Wj
![Page 43: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/43.jpg)
How to compute Di ? (cont’d)• If f(t) ϵ Vj+1, then f(t) can be represented using basis functions φ(t) fromVj+1:
1( ) (2 )jk
k
f t c t k
( ) (2 ) (2 )j jk jk
k k
f t c t k d t k Vj+1 = Vj + Wj
Alternatively, f(t) can be represented using two basis functions, φ(t) from Vj and ψ(t) from Wj:
Vj+1
![Page 44: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/44.jpg)
Think of Wj as a means to represent the parts of a function in Vj+1 that cannot be represented in Vj
1( ) (2 )jk
k
f t c t k
( ) (2 ) (2 )j jk jk
k k
f t c t k d t k Vj, Wj
How to compute Di ? (cont’d)
differencesbetweenVj and Vj+1
![Page 45: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/45.jpg)
How to compute Di ? (cont’d)• using recursion on Vj:
( ) ( ) (2 )jk jk
k k j
f t c t k d t k
V0 W0, W1, W2, …basis functions basis functions
Vj+1 = Vj-1+Wj-1+Wj = …= V0 + W0 + W1 + W2 + … + Wj
if f(t) ϵ Vj+1 , then:
Vj+1 = Vj + Wj
![Page 46: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/46.jpg)
Summary: wavelet expansion (Section 7.2)
• Wavelet decompositions involve a pair of waveforms (mother wavelets):
• The two shapes are translated and scaled to produce wavelets (wavelet basis) at different locations and on different scales.
φ(t) ψ(t)
φ(t-k) ψ(2jt-k)
encode lowresolution info
encode details orhigh resolution info
![Page 47: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/47.jpg)
Summary: wavelet expansion (cont’d)
• f(t) is written as a linear combination of φ(t-k) and ψ(2jt-k) :
( ) ( ) (2 )jk jk
k k j
f t c t k d t k
scaling function wavelet function
![Page 48: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/48.jpg)
1D Haar Wavelets
• Haar scaling and wavelet functions:
computes average computes details
φ(t) ψ(t)
![Page 49: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/49.jpg)
1D Haar Wavelets (cont’d)
• Think of a one-pixel image as a function that is constant over [0,1)
• We will denote by V0 the space of all such functions.
Example:0 1
![Page 50: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/50.jpg)
1D Haar Wavelets (cont’d)
• Think of a two-pixel image as a function having two constant pieces over the intervals [0, 1/2) and [1/2,1)
• We will denote by V1 the space of all such functions.
• Note that
Examples: 0 ½ 1
0 1V V
= +
![Page 51: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/51.jpg)
1D Haar Wavelets (cont’d)• V j represents all the 2j-pixel images• Functions having constant pieces over 2j equal-sized
intervals on [0,1).
• Note that
Examples:
width: 1/2j
ϵ Vj ϵ Vj
1j jV V
![Page 52: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/52.jpg)
1D Haar Wavelets (cont’d)
V0, V1, ..., V j are nested
i.e.,
VJ
…
V2
V1coarse details
fine details
1j jV V
![Page 53: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/53.jpg)
1D Haar Wavelets (cont’d)
• Mother scaling function:
• Let’s define a basis for V j :
0 1
note alternative notation: ( ) ( )ji jix x
1
![Page 54: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/54.jpg)
1D Haar Wavelets (cont’d)
![Page 55: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/55.jpg)
1D Haar Wavelets (cont’d)
• Suppose Wj is the orthogonal complement of Vj in Vj+1
![Page 56: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/56.jpg)
1D Haar Wavelets (cont’d)
• Mother wavelet function:
• Note that φ(x) . ψ(x) = 0 (i.e., orthogonal)
1-1
0 1/2 1
0 1 . = 0 1
-1 0 1/2 1
1
![Page 57: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/57.jpg)
1D Haar Wavelets (cont’d)
• Mother wavelet function:
• Let’s define a basis ψ ji for Wj :
1-1
0 1
( ) ( )ji jix x note alternative notation:
![Page 58: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/58.jpg)
1D Haar Wavelets (cont’d)
j=1
basis W 1 :
basis for V 1 :
Notethat innerproductis zero!
![Page 59: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/59.jpg)
1D Haar Wavelets (cont’d)
Basis functions ψ ji of W j
Basis functions φ ji of V j
form a basis in V j+1
![Page 60: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/60.jpg)
1D Haar Wavelets (cont’d)
V3 = V2 + W2
![Page 61: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/61.jpg)
1D Haar Wavelets (cont’d)
V2 = V1 + W1
![Page 62: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/62.jpg)
1D Haar Wavelets (cont’d)
V1 = V0 + W0
![Page 63: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/63.jpg)
1D Haar Wavelets- Summary
ψ(t)
φ(t)
![Page 64: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/64.jpg)
Example - Haar basis (revisited)
1
![Page 65: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/65.jpg)
Decomposition of f(x)
V2
φ0,2(x)
φ1,2(x)
φ2,2(x)
φ3,2(x)
f(x)=
![Page 66: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/66.jpg)
Decomposition of f(x) (cont’d)
V1and W1
V2=V1+W1
φ0,1(x)
φ1,1(x)
ψ0,1(x)
ψ1,1(x)
![Page 67: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/67.jpg)
Example - Haar basis (revisited)
1
![Page 68: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/68.jpg)
Decomposition of f(x) (cont’d)
![Page 69: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/69.jpg)
Decomposition of f(x) (cont’d)
V2=V1+W1=V0+W0+W1
V0 ,W0 and W1
φ0,0(x)
ψ0,0(x)
ψ0,1(x)
ψ1,1(x)
![Page 70: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/70.jpg)
Example - Haar basis (revisited)
1
![Page 71: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/71.jpg)
Example
![Page 72: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/72.jpg)
Example (cont’d)
![Page 73: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/73.jpg)
Filter banks (analysis)
• The lower resolution coefficients can be calculated from the higher resolution coefficients by a tree-structured algorithm (filter bank).
a0k (j=0)
a1k (j=1)
ψ(t - k)
0 0 0ˆ ( ) ( )k k
k j
f t a t
1 1 1ˆ ( ) ( )k k
k j
f t a t
ψ(2t - k) Example:
![Page 74: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/74.jpg)
Filter banks (analysis) (cont’d)• The lower resolution coefficients can be calculated
from the higher resolution coefficients by a tree-structured algorithm (filter bank).
h0(-n) is a lowpass filter and h1(-n) is a highpass filter
Subbandencoding!
![Page 75: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/75.jpg)
Example - Haar basis (revisited)
[9 7 3 5]low-pass,down-sampling
high-pass, down-sampling
(9+7)/2 (3+5)/2 (9-7)/2 (3-5)/2 V1 basis functions
![Page 76: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/76.jpg)
Filter banks (analysis) (cont’d)
![Page 77: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/77.jpg)
Example - Haar basis (revisited)
[9 7 3 5]
high-pass, down-sampling
low-pass,down-sampling
(8+4)/2 (8-4)/2
V1 basis functions
![Page 78: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/78.jpg)
Convention for illustrating 1D Haar wavelet decomposition
x x x x x x … x x
detailaverage
…
re-arrange:
re-arrange:
V1 basis functions
![Page 79: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/79.jpg)
Convention for illustrating1D Haar wavelet decomposition (cont’d)
x x x x x x … x x detailaverage
…
![Page 80: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/80.jpg)
Orthogonality and normalization
• The Haar basis forms an orthogonal basis• It can become orthonormal through the following
normalization:
since ( ) 2 , ( ) 2j jji jix x
/2
/2
( ) 1/ 2 (2 ) 2 (2 )
( ) 1/ 2 (2 ) 2 (2 )
j j j j ji
j j j j ji
x x i x i
x x i x i
![Page 81: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/81.jpg)
Examples of lowpass/highpass analysis filters
Daubechies
Haarh0
h1
h0
h1
![Page 82: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/82.jpg)
Filter banks (synthesis)
• The higher resolution coefficients can be calculated from the lower resolution coefficients using a similar structure.
![Page 83: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/83.jpg)
Filter banks (synthesis) (cont’d)
![Page 84: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/84.jpg)
Examples of lowpass/highpass synthesis filters
Daubechies
Haar (same as for analysis):
+
g0
g1
g0
g1
![Page 85: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/85.jpg)
2D Haar Wavelet Transform
• The 2D Haar wavelet decomposition can be computed using 1D Haar wavelet decompositions (i.e., 2D Haar wavelet basis is separable).
• Two decompositions– Standard decomposition– Non-standard decomposition
• The two decompositions correspond to a different set of 2D basis functions.
![Page 86: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/86.jpg)
Standard Haar wavelet decomposition
• Steps
(1) Compute 1D Haar wavelet decomposition of each row of the original pixel values.
(2) Compute 1D Haar wavelet decomposition of each column of the row-transformed pixels.
![Page 87: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/87.jpg)
Standard Haar wavelet decomposition (cont’d)
x x x … xx x x … x
… … .x x x ... x
(1) row-wise Haar decomposition:
…
detailaverage
…
… … .
…
…
… … .
re-arrange terms
![Page 88: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/88.jpg)
Standard Haar wavelet decomposition (cont’d)
(1) row-wise Haar decomposition:
…
detailaverage
…
…… … .
…
… … . …
row-transformed result
![Page 89: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/89.jpg)
Standard Haar wavelet decomposition (cont’d)
(2) column-wise Haar decomposition:
…
detailaverage
…
…… … .
……
…… … .
…
row-transformed result column-transformed result
![Page 90: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/90.jpg)
Example
……
…… … .
row-transformed result
…
… … .
re-arrange terms
![Page 91: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/91.jpg)
Example (cont’d)
……
…… … .
column-transformed result
![Page 92: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/92.jpg)
2D Haar basis for standard decomposition
To construct the standard 2D Haar wavelet basis, consider all possible outer products of 1D basis functions.
φ0,0(x)
ψ0,0(x)
ψ0,1(x)
ψ1,1(x)
V2=V0+W0+W1
Example:
![Page 93: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/93.jpg)
2D Haar basis for standard decomposition
To construct the standard 2D Haar wavelet basis, consider all possible outer products of 1D basis functions.
φ00(x), φ00(x) ψ00(x), φ00(x) ψ01(x), φ00(x)
( ) ( )ji jix x ( ) ( )j
i jix x
![Page 94: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/94.jpg)
2D Haar basis of standard decomposition
( ) ( )ji jix x
( ) ( )ji jix x
V2
![Page 95: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/95.jpg)
Non-standard Haar wavelet decomposition
• Alternates between operations on rows and columns.
(1) Perform one level decomposition in each row (i.e., one step of horizontal pairwise averaging and differencing).
(2) Perform one level decomposition in each column from step 1 (i.e., one step of vertical pairwise averaging and differencing).
(3) Repeat the process on the quadrant containing averages only (i.e., in both directions).
![Page 96: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/96.jpg)
Non-standard Haar wavelet decomposition (cont’d)
x x x … xx x x … x
… … .x x x . . . x
one level, horizontalHaar decomposition:
……
… … .
…
……
… … .
one level, vertical Haar decomposition:
…
Note: averaging/differencingof detail coefficients shown as
![Page 97: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/97.jpg)
Non-standard Haar wavelet decomposition (cont’d)
one level, horizontalHaar decompositionon “green” quadrant
one level, vertical Haar decompositionon “green” quadrant
……
… … .
……
re-arrange terms
……
…… … .
…
![Page 98: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/98.jpg)
Example
……
… … .
……
re-arrange terms
![Page 99: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/99.jpg)
Example (cont’d)
……
…… … .
![Page 100: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/100.jpg)
2D Haar basis for non-standard decomposition
• Defined through 2D scaling and wavelet functions:
( , ) ( ) ( )
( , ) ( ) ( )( , ) ( ) ( )( , ) ( ) ( )
x y x y
x y x yx y x yx y x y
000 ( , ) ( , )
( , ) 2 (2 ,2 )
( , ) 2 (2 ,2 )
( , ) 2 (2 ,2 )
j j j jkf
j j j jkf
j j j jkf
x y x y
x y x k y f
x y x k y f
x y x k y f
![Page 101: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/101.jpg)
2D Haar basis for non-standard decomposition (cont’d)
000 ( , ) ( , )
( , ) 2 (2 ,2 )
( , ) 2 (2 , 2 )
( , ) 2 (2 , 2 )
j j j jkf
j j j jkf
j j j jkf
x y x y
x y x k y f
x y x k y f
x y x k y f
LL
LH: intensity variations along columns (horizontal edges)
HL: intensity variations along rows (vertical edges)HH: intensity variations along diagonals
LL: average Detail coefficients
• Three sets of detail coefficients (i.e., subband encoding)
![Page 102: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/102.jpg)
2D Haar basis for non-standard decomposition (cont’d)
( ) ( )ji jix x
( ) ( )ji jix x
V2
![Page 103: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/103.jpg)
Forward/Inverse DWT(using textbook’s notation)
H LHV HLD HH
LL
i=H,V,D
![Page 104: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/104.jpg)
2D DWT using filter banks (analysis)
H LHV HLD HH
LH HH
LL HL
![Page 105: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/105.jpg)
Illustrating 2D wavelet decomposition
The wavelet transform can be applied again on the lowpass-lowpass version of the image, yielding seven subimages.
LL LH
HL HH
LL
HH
HL
LH
![Page 106: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/106.jpg)
2D IDWT using filter banks (synthesis)
H LHV HLD HH
![Page 107: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/107.jpg)
Wavelets Applications• Noise filtering• Image compression
– Fingerprint compression• Image fusion• Recognition
G. Bebis, A. Gyaourova, S. Singh, and I. Pavlidis, "Face Recognition by Fusing Thermal Infrared and Visible Imagery", Image and Vision Computing, vol. 24, no. 7, pp. 727-742, 2006.
• Image matching and retrieval Charles E. Jacobs Adam Finkelstein David H. Salesin, "Fast
Multiresolution Image Querying", SIGRAPH, 1995.
![Page 108: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/108.jpg)
Image Querying Using Wavelets
“query by content” or “query by example”
Typically, the K best matches are reported.
![Page 109: wavelet](https://reader035.fdocuments.net/reader035/viewer/2022070520/577cc9d61a28aba711a4c0c8/html5/thumbnails/109.jpg)
Fast Multiresolution Image Querying
painted low resolution target
queries