Professional Documents
Culture Documents
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING - 2
This Session
4
DIGITAL IMAGE PROCESSING - 2
Linear Transforms
• For a given image and transform (or set of basis functions), both the
orthogonality of the basis functions and the coefficients of the resulting
transform are computed using inner products
• All of an image’s transforms are equivalent in the sense that they contain
the same information and total energy
• They are reversible and differ only in the way that the information and
energy is distributed among the transform’s coefficients
6
DIGITAL IMAGE PROCESSING - 2
Fourier Transforms
7
DIGITAL IMAGE PROCESSING - 2
Why Frequency Domain?
8
DIGITAL IMAGE PROCESSING - 2
Time - Frequency Representations
• Time Domain
• Frequency Domain
9
DIGITAL IMAGE PROCESSING - 2
Time - Frequency Representations Cont..
• Time Domain:
10
DIGITAL IMAGE PROCESSING - 2
Example
Consider
x(t)=cos(210t)+cos(225t)+cos(250t)+cos(2 100t)
11
DIGITAL IMAGE PROCESSING - 2
Example
12
DIGITAL IMAGE PROCESSING - 2
Example
Now consider a non statioanry signal
The interval 0 to 300 ms has a 100 Hz sinusoid,
the interval 300 to 600 ms has a 50 Hz sinusoid,
the interval 600 to 800 ms has a 25 Hz sinusoid,
and finally the interval 800 to 1000ms has a 10Hz sinusoid
13
DIGITAL IMAGE PROCESSING - 2
Example
Its Fourier Transform is
FT of nonstationary signal
FT of stationary signal
14
Ft
DIGITAL IMAGE PROCESSING - 2
FFT of stationary and nonstationary signals
• Other than the ripples, and the difference in amplitude, both the
spectra are almost identical.
16
DIGITAL IMAGE PROCESSING - 2
Joint Time-Frequency
17
DIGITAL IMAGE PROCESSING - 2
Joint Time-Frequency
• Music
Score
• Ideal
Time/frequ
ency
behaviour
18
DIGITAL IMAGE PROCESSING - 2
Joint Time-Frequency
Time-frequency Plane
• Since the support of a function can be defined as the set of points where
the function is nonzero, Heisenberg’s uncertainity principle tells us that it
is impossible for a function to have finite support in both time and
frequency.
• This is called the Heisenberg-Gabor inequality that places a lower bound
on the area of the Heisenberg cell revealing that and cannot both
be arbitrarily small.
20
DIGITAL IMAGE PROCESSING - 2
Joint Time-Frequency Analysis (JTFA)
• Gabor
23
DIGITAL IMAGE PROCESSING - 2
Short Time Fourier Transform(STFT)
f w (t ) = f (t ) w(t a)
25
DIGITAL IMAGE PROCESSING - 2
STFT Cont..
26
DIGITAL IMAGE PROCESSING - 2
Time Frequency Tile of STFT
Ω
∆t
Ωc ∆Ω
t
tc
27
DIGITAL IMAGE PROCESSING - 2
• Perfect frequency resolution in the FT is due to the fact that the window used in the
FT is its kernel, the ejt function, which lasts at all times from minus infinity to plus
infinity. But time information is lost.
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING - 2
This Session
• STFT Cont..
• Introduction to Wavelet transforms
4
DIGITAL IMAGE PROCESSING - 2
• Perfect frequency resolution in the FT is due to the fact that the window used in the
FT is its kernel, the ejt function, which lasts at all times from minus infinity to plus
infinity. But time information is lost.
Disadvantages of STFT
Choice of Window
• It would be appropriate to have a narrow window if frequency
is high and wide window if signal changes slowly
• Multiresolution feature
9
DIGITAL IMAGE PROCESSING - 2
Wavelet Transforms
10
DIGITAL IMAGE PROCESSING - 2
History of Wavelets
11
DIGITAL IMAGE PROCESSING - 2
Wavelet Transforms
12
DIGITAL IMAGE PROCESSING - 2
Wavelets (Introduction)
• Unlike Fourier transform, whose basis functions are
sinusoids, basis function of wavelet transforms are
non sinusoidal
• Based on small waves, called wavelets, of limited duration.
• Fourier transform provides only frequency
information, but wavelet transform provides time-
frequency information.
• Wavelets lead to a multiresolution analysis of signals.
• Multiresolution analysis: Representation of a signal
(e.g., images) in more than one resolution/scale.
• Features that might go undetected at one resolution may
be easy to spot in another.
13
DIGITAL IMAGE PROCESSING - 2
Multiresolution
14
DIGITAL IMAGE PROCESSING - 2
Image pyramids
15
DIGITAL IMAGE PROCESSING - 2
Image pyramids
16
DIGITAL IMAGE PROCESSING - 2
Image pyramids
17
DIGITAL IMAGE PROCESSING - 2
Image pyramids
18
DIGITAL IMAGE PROCESSING - 2
Image pyramids
19
DIGITAL IMAGE PROCESSING - 2
Image pyramids
20
DIGITAL IMAGE PROCESSING - 2
Approximation and Residual Pyramids
21
DIGITAL IMAGE PROCESSING - 2
Wavelet Theory
22
DIGITAL IMAGE PROCESSING - 2
Wavelet Theory Cont..
23
DIGITAL IMAGE PROCESSING - 2
Wavelet Theory Cont..
24
DIGITAL IMAGE PROCESSING - 2
Time-frequency tiles
25
DIGITAL IMAGE PROCESSING - 2
frequency
|****************************************
|* * * * * * * * * * * * * * *
|* * * * * * *
|* * * *
|* *
--------------------------------------------> time
29
DIGITAL IMAGE PROCESSING - 2
Wavelet Transforms
• Wavelet means small wave. So wavelet analysis is about
analysing signal with short duration finite energy functions
• The transformation is called wavelet transform
• The wavelet is made to scan the signal with varying
translations and dilations
• The wavelet plot shows correlation between wavelet at some
scale & location of the signal
30
DIGITAL IMAGE PROCESSING - 2
Wavelet transforms
• Wavelet function (basis) is given by:
1 t b
ab (t )
a a
• Wavelet transform is defined as
W (a, b) f (t ) a ,b (t )dt
t
1 t b
f (t ) dt
t a a
32
DIGITAL IMAGE PROCESSING - 2
Continuous Wavelet Transform
• CWT is given by
1 t b
W (a, b) f (t ) dt
t a a
33
DIGITAL IMAGE PROCESSING - 2
Discrete Wavelet transforms
34
DIGITAL IMAGE PROCESSING - 2
35
DIGITAL IMAGE PROCESSING - 2
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING - 2
This Session
4
DIGITAL IMAGE PROCESSING - 2
Time-frequency tiles
5
DIGITAL IMAGE PROCESSING - 2
frequency
|****************************************
|* * * * * * * * * * * * * * *
|* * * * * * *
|* * * *
|* *
--------------------------------------------> time
9
DIGITAL IMAGE PROCESSING - 2
Time-frequency Localization
11
DIGITAL IMAGE PROCESSING - 2
Wavelet transforms
• Wavelet function (basis) is given by:
1 t b
ab (t )
a a
W (a, b) f (t ) a ,b (t )dt
t
1 t b
f (t ) dt
t a a
13
DIGITAL IMAGE PROCESSING - 2
Sampling Continuous Wavelet Transform
• CWT is given by
1 t b
W (a, b) f (t ) dt
t a a
14
DIGITAL IMAGE PROCESSING - 2
Discrete Wavelet transforms
15
DIGITAL IMAGE PROCESSING - 2
Sub band Coding
16
DIGITAL IMAGE PROCESSING - 2
Decomposition using DWT
17
DIGITAL IMAGE PROCESSING - 2
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING - 2
This Session
4
DIGITAL IMAGE PROCESSING - 2
Sampling Continuous Wavelet Transform
• CWT is given by
1 t b
W (a, b) f (t ) dt
t a a
5
DIGITAL IMAGE PROCESSING - 2
Discrete Wavelet transforms
6
DIGITAL IMAGE PROCESSING - 2
Sub band Coding
7
DIGITAL IMAGE PROCESSING - 2
Decomposition using DWT
8
DIGITAL IMAGE PROCESSING - 2
Multiresolution Expansions
• Series Expansions
Any function can be expressed as linear combination of basis
functions
where
Dual function of
* Complex conjugate operation
9
DIGITAL IMAGE PROCESSING - 2
Multiresolution Expansions
• k is integer index
• ak is real valued expansion coefficient
• k(t) is set of real valued functions of t called expansion set
10
DIGITAL IMAGE PROCESSING - 2
Multiresolution Expansions
11
DIGITAL IMAGE PROCESSING - 2
Spanning set
12
DIGITAL IMAGE PROCESSING - 2
Wavelet Expansion
13
DIGITAL IMAGE PROCESSING - 2
14
DIGITAL IMAGE PROCESSING - 2
Additional Characteristics
k (t ) (t k ); k Z , L2 ( R )
16
DIGITAL IMAGE PROCESSING - 2
Scaling function
Let us define,
k (t ) (t k ); k Z , L2 ( R )
17
DIGITAL IMAGE PROCESSING - 2
Expansion using Scaling function
• A 2-D family of functions is generated from basic scaling function
j ,k (t ) 2 j / 2 (2 j t k ); j , k Z
Whose span over k is j Span k (2 j t ) int k
• If
f (t) j
f (t) å akj k (2 j t k)
k
• For j>0 the span can be larger thus representing finer
details
• For j<0 the span is smaller the scaling function is wider &
translated in larger steps thus representing coarse
information.
19
DIGITAL IMAGE PROCESSING - 2
Scaling function
20
DIGITAL IMAGE PROCESSING - 2
Multiresolution Analysis (MRA)
j j1j Z
21
DIGITAL IMAGE PROCESSING - 2
Nested Vector Space
V3
V2
V1
V0
j j1j Z 22
DIGITAL IMAGE PROCESSING 2
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING 2
This Session
4
DIGITAL IMAGE PROCESSING 2
Scaling function
Let us define,
k (t ) (t k ); k Z , L2 ( R )
5
DIGITAL IMAGE PROCESSING 2
Expansion using Scaling function
• A 2-D family of functions is generated from basic scaling function
j ,k (t ) 2 j / 2 (2 j t k ); j , k Z
Whose span over k is j Span k (2 j t ) int k
• If
f (t) j
f (t) akj k (2 j t k)
k
• For j>0 the span can be larger thus representing finer
details
• For j<0 the span is smaller the scaling function is wider &
translated in larger steps thus representing coarse
information.
7
DIGITAL IMAGE PROCESSING 2
Scaling function
8
DIGITAL IMAGE PROCESSING 2
Multiresolution Analysis (MRA)
j j1j Z
9
DIGITAL IMAGE PROCESSING 2
Nested Vector Space
V3
V2
V1
V0
j j1j Z 10
DIGITAL IMAGE PROCESSING 2
Scaling and Wavelet Function
11
DIGITAL IMAGE PROCESSING 2
Scaling and Wavelet Function
V3
W1 W2
W0
V2
V1
V0
V j 1 V j W j
13
DIGITAL IMAGE PROCESSING 2
Signal Representation using Scaling and Wavelet Function
L ( R ) V j 0 W j 0 W j 0 1...
2
14
DIGITAL IMAGE PROCESSING 2
Approximation and Residual Pyramids
15
DIGITAL IMAGE PROCESSING 2
Types of Wavelet families -Mother Wavelets
• Haar Wavelets
• Coiflet Wavelets
• Bi-orthogonal Wavelets
16
DIGITAL IMAGE PROCESSING 2
Haar Wavelets
18
DIGITAL IMAGE PROCESSING 2
Daubechies Mother Wavelets
19
DIGITAL IMAGE PROCESSING 2
Other Mother Wavelets
Symlet Wavelet
Scaling Functions:
• Integer translations and dyadic scalings of a scaling function
21
DIGITAL IMAGE PROCESSING 2
1.
j 0,0 (x)
2.
j1,0 (x)
3.
j1,1 (x)
22
DIGITAL IMAGE PROCESSING 2
Also
1
j 0,0 (x) j1,0 (x)
2
1
j1,1 (x)
2
23
DIGITAL IMAGE PROCESSING 2
Example
24
DIGITAL IMAGE PROCESSING 2
Multiresolution Expansions
2. ,
3. ,
25
DIGITAL IMAGE PROCESSING 2
Example
1.25
0.75
26
DIGITAL IMAGE PROCESSING 2
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING 2
This Session
• 1D,2D DWT
• DWT applied to images
4
DIGITAL IMAGE PROCESSING 2
Continuous Wavelet Transforms
W (a, b) f (t ) a ,b (t )dt
t
1 t b
f (t ) dt
t a a
• or every a,b we have a wavelet coefficient representing how
much the scaled wavelet is similar to the function at location
t=b
5
DIGITAL IMAGE PROCESSING 2
Properties of Wavelet
6
DIGITAL IMAGE PROCESSING 2
Properties of Wavelet
Regularity conditions:
• Regularity conditions are imposed on the wavelet functions
in order to make the wavelet transform decrease quickly with
decreasing scale s.
8
DIGITAL IMAGE PROCESSING 2
More Properties of Wavelets
9
DIGITAL IMAGE PROCESSING 2
Selection of Basis Function
10
DIGITAL IMAGE PROCESSING 2
Approximation and Detail Coefficients of DWT
11
DIGITAL IMAGE PROCESSING 2
Signal Representation using Scaling and Wavelet Function
12
DIGITAL IMAGE PROCESSING 2
Haar Scaling Function
ìï 1 ;0 £ x £ 1
(x) í
ïî 0 ;Otherwise
13
DIGITAL IMAGE PROCESSING 2
Multiresolution Expansions
• Scaling functions
• Dilation equation for scaling function
14
DIGITAL IMAGE PROCESSING 2
Multiresolution Expansions
Wavelet functions
15
DIGITAL IMAGE PROCESSING 2
Haar Wavelets
⎩(x)⎩
⎩⎩ 0, otherwise.
⎩
⎩ 1, 0⎩ x<1/2,
⎩0, otherwise.
⎩1, if0 x<⎩ 1
Thus approximation filter coefficients
⎩(x)2⎩(x1)
17
DIGITAL IMAGE PROCESSING 2
The Haar Transform
• Haar wavelet is the only one wavelet basis, which holds the
properties of orthogonal, (anti-) symmetric and compactly
supported.
18
DIGITAL IMAGE PROCESSING 2
The Haar Wavelet Filters
19
DIGITAL IMAGE PROCESSING 2
Transformation using Haar Wavelet
20
DIGITAL IMAGE PROCESSING 2
DWT of an Image
21
DIGITAL IMAGE PROCESSING 2
DWT of an Image
22
DIGITAL IMAGE PROCESSING 2
DWT of an Image
23
DIGITAL IMAGE PROCESSING 2
Wavelet Decomposition
24
DIGITAL IMAGE PROCESSING 2
4 – level Wavelet Decomposition of Lena
25
DIGITAL IMAGE PROCESSING 2
DWT
26
DIGITAL IMAGE PROCESSING 2
Image Pyramids
27
DIGITAL IMAGE PROCESSING 2
Image Pyramids
28
DIGITAL IMAGE PROCESSING 2
Image Pyramids
29
DIGITAL IMAGE PROCESSING 2
Image Pyramids
30
DIGITAL IMAGE PROCESSING 2
Image Pyramids
31
DIGITAL IMAGE PROCESSING 2
Image Pyramids
32
DIGITAL IMAGE PROCESSING 2
Subband Coding
33
DIGITAL IMAGE PROCESSING 2
2-Band Subband Coding
34
DIGITAL IMAGE PROCESSING 2
4-Band Subband Coding
35
DIGITAL IMAGE PROCESSING 2
Comparison of FFT,CWT,DWT
• For applications where the frequency values play a very vital role and
where approximations cannot be tolerated FFT can be preferred to
wavelet transforms.
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING 2
This Session
4
DIGITAL IMAGE PROCESSING 2
DWT of an Image
5
DIGITAL IMAGE PROCESSING 2
DWT of an Image
Third Level
Decomposition
Second Level
Decomposition
DC value
7
DIGITAL IMAGE PROCESSING 2
Wavelet Decomposition
8
DIGITAL IMAGE PROCESSING 2
4 – level Wavelet Decomposition of Lena
9
DIGITAL IMAGE PROCESSING 2
DWT
10
DIGITAL IMAGE PROCESSING 2
Applications of DWT
13
DIGITAL IMAGE PROCESSING 2
Image Compression using DWT
14
DIGITAL IMAGE PROCESSING 2
Image Compression using DWT
Disadvantage of DCT:
• Fixed block size: does not consider irregular shapes of real
images
• Introduces blocking artifacts at low bit rates
15
DIGITAL IMAGE PROCESSING 2
Image Compression using DWT
17
DIGITAL IMAGE PROCESSING 2
DWT of an Image
DC value
18
DIGITAL IMAGE PROCESSING 2
Wavelet Coding: Compression using DWT
21
DIGITAL IMAGE PROCESSING 2
Selection of Wavelet
symlets 22
Cohen- Daubechies-Feauveau biorthogonal wavelets.
DIGITAL IMAGE PROCESSING 2
Performance Analysis of Wavelet basis Functions
23
DIGITAL IMAGE PROCESSING 2
Performance Analysis of Wavelet basis Functions
24
DIGITAL IMAGE PROCESSING 2
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
1
DIGITAL IMAGE PROCESSING 2
This Session
4
DIGITAL IMAGE PROCESSING 2
DWT
5
DIGITAL IMAGE PROCESSING 2
Image Compression using DWT
6
DIGITAL IMAGE PROCESSING 2
Wavelet Coding: Compression using DWT
7
DIGITAL IMAGE PROCESSING 2
Selection of Wavelet
8
DIGITAL IMAGE PROCESSING 2
Decomposition Level Selection
9
DIGITAL IMAGE PROCESSING 2
Decomposition Level Selection
13
DIGITAL IMAGE PROCESSING 2
Quantizer Design
16
DIGITAL IMAGE PROCESSING 2
Compression Standards
• If the image has more than one component, such as the red, green, and
blue planes of a color image, each component is shifted individually.
20
DIGITAL IMAGE PROCESSING 2
JPEG 2000 Algorithm
22
DIGITAL IMAGE PROCESSING 2
JPEG 2000 Algorithm
23
DIGITAL IMAGE PROCESSING 2
JPEG 2000 Algorithm
24
DIGITAL IMAGE PROCESSING 2
JPEG 2000 Algorithm
26
DIGITAL IMAGE PROCESSING 2
JPEG 2000 Algorithm
Consider 4 JPEG-2000 approximations of Lena.
Each row contains a result after compression and reconstruction, the scaled difference between the
result and the original image, and a zoomed portion of the reconstructed image.
level of compression, C = 52
level of compression, C = 52
29
THANK YOU
Dr. Shruthi M L J
Department of Electronics &
Communication Engineering
shruthimlj@pes.edu
+91 8147883012
31