Discrete Orthogonal Transforms and Its Applications: Dr. J. Sheeba Rani

Discrete Orthogonal Transforms and its
applications
Presented by
Dr. J. Sheeba Rani
Asst Professor/Avionics
Indian Institute of Space Science and Technology,

Dept. of Space, Trivandrum.
Indian Institute of Space Science and Technology 04/27/2020
Presentation Outline
• Need for Transforms.
•What are Orthogonal Transforms? Why not Non
Orthogonal?
•What are the requirements of Orthogonal
transforms and Why it is required
•Brief Discussion on types of Discrete Orthogonal
Transforms
•Applications
•Current Research Scope
Indian Institute of Space Science and Technology

Why Transforms?
• Better image processing/signal processing
• To obtain a further information from the signal that is not

readily available in the raw signal.
• Fast computation: convolution vs. Multiplication
• Efficient Storage and transmission
• Transformed data as a measurement/feature for high level

processing
• Spatial Frequency Information , Multiresolution for further

analysis

What Transform does?
Transforms tries to approximate a function f(x) by another function
gn(x) which consists of a sum over N orthogonal functions F(x)
weighted by some coefficients an.
N
f ( x)  g N ( x)   ai  i ( x)
i 0
1/ 2
 b

f ( x)  g N ( x)     f ( x)  g N ( x) dx 
2
 Min !
2
a 
... a good choice for the basis functions F(x) are orthogonal functions.
What are orthogonal functions? Two functions f and g are said to
be orthogonal in the interval [a,b] if
b
fi gi
 f ( x) g ( x)dx  0
a
Linear Transform posses…
• A linear transformation R from Rn to Rn is orthogonal if and only if the
vectors form an orthonormal basis of Rn
• An nxn matrix A is orthogonal if and only if its columns form an
orthonormal bases of Rn
• A linear transformation T from Rn to Rn is called orthogonal if it
preserves the lengths of vectors.
•  
||T(x)|| = ||x||
• 
• If T(x) is an orthogonal transformation then we say that A is an
orthonormal matrix.
Example
Let x be a one dimensional vector of length2 x1. A is the orthogonal matrix.
I T(x) is the transformed output.
The transformed output be of length 2 x1.
T(x)=Ax; is orthogonal if AAT=I

Significance of Orthogonal
Transforms
•Orthogonal transforms tend to de-correlate the
components of a given signal .
•Orthogonal transforms tend to redistribute the energy
contained in the signal so that most of energy is contained in
a small number of components.
•Good for efficient storage and transmission.
•Inverse
• transform exist
• Orthonormal, complete basis (sort of) shift- and rotation
invariant
•Fast Computation

Broad Applications
Signal Processing
Speech signal analysis
Image Processing Familiar application is Image Compression
JPEG 2000
Object recognition
Image compression
Communication
Signal Analysis
Signal Correlation
Data compression
Spectroscopy
Multiresolution Analysis

Types of Orthogonal Transforms
Type I DFT
Sub-Optimal
SINUSOIDAL
Discrete
(DCT)
Orthogonal
Transform Type II
Optimal (KLT)
NON
SINUSOIDAL

Some names
KLT Transform (Optimal)

Discrete Fourier transform
Discrete Cosine Transform
Discrete Sine Transform
Discrete Tchebichef Transform
Discrete Haar Transform
Discrete Walsh Hadamard Transform
Discrete Wavelet Transform
Discrete Krawthouk Transform
Discrete Wavelet Transform

KL TRANSFORM (OPTIMAL)
•It de-correlates the transform coefficients,
•packs the most energy in few coefficients,
•minimizes the mean—square error (mse) between the
reconstructed and originalimages and
•minimizes the total entropy compared to any other transform.

• PCA is the widely used subspace projection technique for
dimensionality reduction. PCA basis vectors are computed
from a set of training images .
• As a first step ,the average image z is computed and
subtracted from the training images creating a set of data
samples.
i1,i2,…..in ε L-1
• These data samples are then arrayed in a matrix X, with one

column per sample image. The covariance matrix is
calculated using
C = XXT


C is the sample covariance matrix for the training images.
Theprincipal components of the covariance matrix are
computed by solving
R1(C)R=A
Where A is the diagonal matrix of eigenvalues and R is the

matrix of orthonormal eigenvectors.
‘N’ eigenvectors associated with the largest eigenvalues are
used to define the subspace, where N is the desired subspace
dimensionality.i.e U = [u1,u2 ,….uD], containing D eigenvectors
of C with largest corresponding eigenvalues. Here, D<NP.
A feature vector K of dimensionality D is then derived from a
feature space X using
K = UT X

Steps to create an Eigen space. 1. Centre data: Each of the training images must be centered.
2. Subtracting the mean image from each of the training images centers the training
images.
3. The mean image is a column vector such that each entry is the mean of all
corresponding pixels of the training images. Create data matrix: Once the training
images are centered, they are combined into a data matrix of size NxP, where P is the
number of training images and each column is a single image.
4. Create covariance matrix: The data matrix is multiplied by its transpose to create a
covariance matrix .
5. Compute the eigenvalues and eigenvectors.
6. Order eigenvectors: Order the eigenvectors V according to their corresponding
eigenvalues from high to low. Keep only the eigenvectors associated with non-zero
eigenvalues. This matrix of eigenvectors is the eigenspace V , where each column of
V is an eigenvector.
7. Project training images. Each of the centered training images is projected into the
eigenspace. To project an image into the eigenspace, calculate the dot product of the
image with each of the ordered eigenvectors.
8. Therefore, the dot product of the image and the first eigenvector will be the first value
in the new vector. The new vector of the projected image will contain as many values
as eigenvectors.

Example
x=[100,200,150,125,150,160]
y=[130,120,140,200,140,170] Image
z=[90,150,200,50,150,160] Data
w=[120,210,160,140,130,200]
Step 1 Mean=[110,170,162.5,128.75,142.5,172.5] Centre data
Dat =
-10.0000 20.0000 -20.0000 10.0000 Subtracting the mean image

30.0000 -50.0000 -20.0000 40.0000 from each of the training
Step 2
-12.5000 -22.5000 37.5000 -2.5000 images centers the training
-3.7500 71.2500 -78.7500 11.2500 images.
7.5000 -2.5000 7.5000 -12.5000
-12.5000 -2.5000 -12.5000 27.5000

Covar1 =
Create data matrix: Once
1.0e+004 * the training images are
Step 3 &4 centered, combine them
0.1000 -0.0500 -0.1100 0.3150 -0.0400 0.0600 into a data matrix of size
-0.0500 0.5400 -0.0100 -0.1650 -0.0300 0.1100 NxP, where P is the
-0.1100 -0.0100 0.2075 -0.4537 0.0275 -0.0325 number of training images
0.3150 -0.1650 -0.4537 1.1419 -0.0938 0.1163 and each column is a
-0.0400 -0.0300 0.0275 -0.0938 0.0275 -0.0525 single image.
0.0600 0.1100 -0.0325 0.1163 -0.0525 0.1075
Step 5 Eigen vector and eigen values val =
1.0e+004 *
vec =
-0.0000 0 0 0 0 0
-0.0348 0.6520 0.6324 0.3361 0.0233 0.2455 0 -0.0000 0 0 0 0
0.0778 -0.0759 0.2047 -0.1809 0.9429 -0.1566 0 0 0.0000 0 0 0
0.1842 -0.5932 0.5293 0.4373 -0.1514 -0.3461 0 0 0 0.1062 0 0
0.0756 -0.4318 0.1321 -0.0541 0.0669 0.8849 0 0 0 0 0.5634 0
-0.8405 -0.1580 0.3437 -0.3678 -0.1005 -0.0715 0 0 0 0 0 1.4548
-0.4967 -0.0758 -0.3774 0.7245 0.2700 0.0856

Vect =
Step 6
0.3361 0.0233 0.2455
-0.1809 0.9429 -0.1566
0.4373 -0.1514 -0.3461
-0.0541 0.0669 0.8849
Select the eigen vector which has highest eigen values
-0.3678 -0.1005 -0.0715
0.7245 0.2700 0.0856
x11=Vect'*x1' Step 6 KL coefficient

Step 7
x11 =
y11 =
-25.8662
25.5668 1.1837 z11 =
-7.7518 -38.9276
5.7384 w11 =
83.5378
-34.4008
-86.0463 18.9441
47.7615
10.2603

Test Image
Test1=[100,160,201,60,155,160]
Tcentre =
-10.0000 -10.0000 184.5000 -68.7500 12.5000 -12.5000
Tcoef=Vect'*Tcentre'
Tcoef =
69.0110
-46.8159
-127.5451

implementation of KLT involves
determination of eigenvalues and corresponding eigenvectors
of the covariance matrix
There is no general algorithm that enables its fast computation
The next equivalent optimal

transform is
Discrete Cosine transform
Limitations of 1D PCA
2D image matrices must be previously transformed to 1D
vector.
The size of the covariance matrix is very high and
2D Subspace
computation time is high.

Analysis
To overcome this problem

2D PCA
2D LDA
Disadvantages of this methods
Requirement of huge feature matrix representation.
To overcome this Daoqiang et al. proposed
2 directional 2 Dimensional PCA. Diagonal PCA method
PCA method uses fewer coefficients for recognition
purposes but two dimensional methods consume least
computation time and more features for recognition.

DFT
A Discrete time /spatial domain signals can be transformed into discrete
frequency/multi-resolution domain using forward transform equation Fourier
Transform
convert
spatial/time
domain to
frequency
domain
a(u,n) and b(u,n) are the Orthogonal Basis function. For a Discrete Fourier Transform
The Orthogonal Basis function is defined as
One Dimensional Two Dimensional

Drawn using In Matrix Notation
Matlab) Y=Ax
X=A-1Y
Inference
I=ones(25,25);
F=fft2(I);
subplot(3,2,1) Fourier Transform
imshow(I) gives the frequency
subplot(3,2,2) information in the
imshow(real(F)) transformed domain
subplot(3,2,3)
imshow(imag(F))
I1=[1 -1 2 -1;1 1 3 -4;2 -3 Application

1 -4; 1 1 1 1] Image Enhancement
F1=fft2(I1); Image Filtering
subplot(3,2,4) Image segmentation
imshow(I1) Feature Extraction
subplot(3,2,5)
imshow(real(F1))
subplot(3,2,6)
imshow(imag(F1))

e of
c
c an
i gnifi onal
s
t he o siti
is e p
t
a . h
t a rp
h
W ase rve , sh s
ph rese tion es
a
It p orm nuiti
inf conti
dis

Original Image Fourier Transform Magnitude
Fourier Transform Phase Reconstructed Magnitude
Reconstructed Phase Reconstructed Magnitude and

Phase

Two Dimensional Implementation

DFT –Fast Convolution
04/27/2020
Discrete Cosine Transform
• Used in JPEG and MPEG

• Another Frequency Transform, with Different
Set of Basis Functions
04/27/2020
The DCT is a widely used frequency transform because it
Discrete Cosine
closely approximates the optimal KLT transform

Transform
KLT is is signal-dependent, and there is no general algorithm

for its fast computation.
The DCT possess data-independent basis functions and

several algorithms for fast implementation.
The DCT provides a good trade-off between energy packing

ability and computational complexity.
The energy packing property of DCT is superior to that of any

other unitary transform.

Orthogonal 2D DCT basis functions The result is an N x
N matrix C(v; u) containing DCT coefficients.
DCT Transform

04/27/2020
I=DCT2(Original image) % matlab

04/27/2020
Truncation effect in DFT and DCT

04/27/2020
DCT versus DFT
•DCT has a higher compression efficiency than DFT
does
•DCT avoids the generation of spurious spectral
components
The transform which minimizes the MSE

is the one that packs the maximum
amount of variance in the first few
coefficients

What are Wavelets?
• Wavelets are functions that “wave” above and

below the x-axis, have
• (1) varying frequency,
• (2) limited duration,
• (3) an average value of zero.
• This is in contrast to sinusoids, used by FT,
which have infinite energy.

Wavelet Transform
• Convert a signal into a series of wavelets
• Provide a way for analyzing waveforms,
bounded in both frequency and duration
• Allow signals to be stored more efficiently than
by Fourier transform
• Be able to better approximate real-world
signals
• Well-suited for approximating data with sharp
discontinuities
• Good for Multiresolution analysis
04/27/2020
STATIONARITY OF SIGNAL
 Stationary Signal
 Signals with frequency content unchanged
in time
 All frequency components exist at all times
 Non-stationary Signal
 Frequency changes in time
 One example: the “Chirp Signal”
04/27/2020
STATIONARITY OF SIGNAL
2 Hz + 10 Hz + 20Hz 3 600
2 500
Magnitud
Magnitud
1 400
Stationary 0 300
e
-1 200
-2 100
-3 0
0 0.2 0.4 0.6 0.8 1 0 5 10 15 20 25
Time Frequency (Hz)
0.0-0.4: 2 Hz + 1 250
0.8
0.4-0.7: 10 Hz + 0.6 200
Magnitud
Magnitud
0.7-1.0: 20Hz 0.4
0.2 150
Non- 0
e
e
Stationary -0.2
-0.4
100
-0.6 50
-0.8
-1 0
0 0.5 1 0 5 10 15 20 25
Time Frequency (Hz)
04/27/2020
CHIRP SIGNALS
Frequency: 2 Hz to 20 Hz Different in Time Domain Frequency: 20 Hz to 2 Hz
1 150 1 150
0.8 0.8
0.6 0.6
0.4 0.4
Magnitud
Magnitud
100 100
Magnitud
Magnitud
0.2 0.2
0 0
e
e
e
e
-0.2 -0.2
50 50
-0.4 -0.4
-0.6 -0.6
-0.8 -0.8
-1 0 -1 0
0 0.5 1 0 5 10 15 20 25 0 0.5 1 0 5 10 15 20 25
Time Frequency (Hz) Time Frequency (Hz)
Same in Frequency Domain
Atwhat
At whattime
timethe
thefrequency
frequencycomponents
componentsoccur?
occur? FT
FTcan
cannot
nottell!
tell!
04/27/2020
Limitation of FT for Non Stationary
Signals
• FT Only Gives what Frequency Components Exist in the
Signal
• The Time and Frequency Information can not be Seen at
the Same Time
• Time-frequency Representation of the Signal is Needed
We need to know whether and also when an information

occurred or happened
Earlier solution is Short Time Fourier Transform (STFT)
04/27/2020
SFORT TIME FOURIER TRANSFORM
(STFT)
• Dennis Gabor (1946) Used STFT
– To analyze only a small section of the signal at a time -- a
technique called Windowing the Signal.
• The Segment of Signal is Assumed Stationary
• A 3D transform
STFTX    t , f     x t   *  t  t    e  j 2 ft dt
t
 t  : the window function
A function of time
and frequency
04/27/2020
Limitation of STFT
• Unchanged Window
• Dilemma of Resolution
– Narrow window -> poor frequency resolution
– Wide window -> poor time resolution
• Heisenberg Uncertainty Principle
– Cannot know what frequency exists at what time intervals
04/27/2020
MULTIRESOLUTION ANALYSIS (MRA)
• Wavelet Transform
– An alternative approach to the short time Fourier transform to
overcome the resolution problem
– Similar to STFT: signal is multiplied with a function
• Multiresolution Analysis
– Analyze the signal at different frequencies with different
resolutions
– Good time resolution and poor frequency resolution at high
frequencies
– Good frequency resolution and poor time resolution at low
frequencies
– More suitable for short duration of higher frequency; and longer
duration of lower frequency components
04/27/2020
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).
f (t )   ak k (t )
k

There are many different wavelets:
Haar Morlet Daubechies

  jk (t )
(dyadic/octave grid)
 jk (t )  2 j/2
  2 t k
j
j
scale/frequency
localization
time localization
Continuous Wavelet Transform
(CWT)
Translation parameter, Scale parameter Normalization
measure of time (measure of frequency) constant
1 * t   
CWTx  , s   x  , s    x  t      dt
Forward
s  s 
CWT:
Translation
(The location of Scale
the window)
Mother Wavelet
Continuous wavelet Scale = 1/j = 1/Frequency

transform of the signal f(t) Mother wavelet
(window)
04/27/2020
SCALE
• Scale
– S>1: dilate the signal
– S<1: compress the signal
• Low Frequency -> High Scale -> Non-detailed
Global View of Signal -> Span Entire Signal
• High Frequency -> Low Scale -> Detailed View
Last in Short Time
• Only Limited Interval of Scales is Necessary
04/27/2020
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.
04/27/2020
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.
04/27/2020
CWT: Main Steps (cont’d)
4. Scale the wavelet and repeat steps 1 through 3.
5. Repeat steps 1 through 4 for all scales.
04/27/2020
Coefficients of CTW Transform
• Wavelet analysis produces a time-scale view of the input signal or image.
1  t  
C ( , s )   f  t 

 dt
s t  s 
04/27/2020
RESOLUTION OF TIME & FREQUENCY
Better time
resolution;
Poor
frequency
resolution
Frequenc
y
Better
frequency
resolution;
Poor time Time
resolution • Each box represents a equal portion
• Resolution in STFT is selected once for entire
analysis
04/27/2020
Comparison with other Transformations
From http://www.cerm.unifi.it/EUcourse2001/Gunther_lecturenotes.pdf, p.10
04/27/2020
FT vs WT
weighted by F(u)
weighted by C(τ,s)
04/27/2020
Discrete Wavelet Transform (DWT)
• Provides sufficient information both for analysis and synthesis
• Reduce the computation time sufficiently
• Easier to implement
• Analyze the signal at different frequency bands with different
resolutions
• Decompose the signal into a coarse approximation and detail
information
a jk   f (t ) *jk (t ) (forward DWT)

t
f (t )   a jk jk (t ) (inverse DWT)

k j
where  jk (t )  2 j / 2  2 j t  k 
04/27/2020
DFT vs DWT
FT expansion: one parameter basis
or f (t )   a l l (t )
WT expansion l
two parameter basis
f (t )   a jk jk (t )
k j

Multiresolution Representation
using  (t ) jk
f (t )
fine
details
narrower, small translations
f (t )   a jk jk (t )
k j j
coarse
Indian Institute of Space Science and Technology details
04/27/2020
Multiresolution
Representation using  jk (t )
f (t )
fine
details
f (t )   a  jk (t )
jk j
k j
coarse
04/27/2020
Multiresolution
f (t )
fine
details
wider, large translations
f (t )   a jk jk (t ) j
k j
coarse
04/27/2020
Multiresolution
high resolution
f (t )
(more details)
fˆ1 (t )
j
fˆ2 (t )
…
low resolution
fˆs (t ) (less details)
f (t )   a jk  jk (t )
k j

Wavelet expansion (cont’d)
• f(t) is written as a linear combination of φ(t-k)

and ψ(2jt-k) :
f (t )   ck (t  k )   d jk (2 t  k ) j
k k j
scaling function wavelet function
Note: in Fourier analysis, there are only two possible values of k ( i.e., 0 and
π/2); the values j correspond to different scales (i.e., frequencies).
04/27/2020
1D Haar Wavelets
• Haar scaling and wavelet functions:
φ(t) ψ(t)
computes average computes details
04/27/2020
Wavelet Transforms in Two Dimension
The two-dimensional FWT  the analysis filter.
04/27/2020
Wavelet Transforms in Two Dimension
W ( j , m, n) WH ( j , m, n)
W ( j  1, m, n)
WV ( j , m, n) WD ( j , m, n)
two-dimensional decomposition
W ( j, m, n) WH ( j , m, n )
W ( j  1, m, n)
WV ( j , m, n) WD ( j , m, n)
Two-scale of two-dimensional decomposition

04/27/2020
RECONSTRUCTION
A Process After decomposition
or analysis.
Also called synthesis
Reconstruct the signal from the wavelet coefficients

Where wavelet analysis involves filtering and down
sampling, the wavelet reconstruction process consists
of up sampling and filtering
04/27/2020
Wavelet Transform using Haar Wavelet

Wavelet Applications
Astronomy, acoustics, nuclear engineering,

sub-band coding, signal and image
processing, neurophysiology, music,
magnetic resonance imaging, speech
discrimination, optics, fractals, turbulence,
earthquake-prediction, radar, human
vision, and pure mathematics applications
04/27/2020
Application- Image Compression
Wavelet Entropy
image Quantization bitstream
coding coding
 W j (m, n) 
• Quantization q j (m, n)  sign( y j ( m, n)) 
  j


– uniform scalar quantization
– separate quantization step-sizes for each subband
• Entropy coding
– Huffman coding
– Arithmetic coding
04/27/2020
Simulation Result
DCT-based Wavelet-based
Original image image compression image compression
CR = 11.2460 CR = 10.3565
RMS = 4.1316 RMS = 4.0104
04/27/2020
7 Simulation Result
CR = 27.7401 CR = 26.4098
RMS = 6.9763 RMS = 6.8480
04/27/2020
7 Simulation Result
CR = 53.4333 CR = 51.3806
RMS = 10.9662 RMS = 9.6947
04/27/2020
Research Scope
• Analyzing the potential and characteristics of

orthogonal functions and utilizing in various
applications
• Developing fast computational algorithms for
implementation
• Developing new transforms using orthogonal
polynomials and deriving its invariant
characteristics
• For object recognition, image analysis and
understanding

Thank You

Reference
R. C. Gonzolez, R. E. Woods, "Digital Image Processing second

edition", Prentice Hall, 2002.
R. C. Gonzolez, R. E. Woods, S. L. Eddins, "Digital Image
Processing Using Matlab", Prentice Hall, 2004.
T. Acharya, A. K. Ray, "Image Processing: Principles and
Applications", John Wiley & Sons, 2005.
B. E. Usevitch, 'A Tutorial on Modern Lossy Wavelet Image
Compression: Foundations of JPEG 2000', IEEE Signal
Processing Magazine, vol. 18, pp. 22-35, Sept. 2001.

In many practical applications, such as radar, sonar, and communications, the signals
under consideration are usually time-varying, that is, their frequencies are varying
with time.
the phase of an arbitrary time-varying signal can be well approximated by a

polynomial of sufficient order. This kind of time-varying signals with polynomial
phase, also known as the polynomial-phase signals
LocalpolynomialFouriertransform:Areviewonre
cent developmentsandapplications Xiumei Li a,
GuoanBi b, , SrdjanStankovic c,
AbdelhakM.Zoubir d Signal processing
The PPSs have been reported to be used in many

differentareas,suchasbiomedical engineering [10], imageprocessing [11],
imageandaudio watermarking [12], motionestimationinvideosequence
[13,14], communications [15–17], sonar [18] and radar applications
[2,19].

which describe how the frequency content of a
signal evolves over the time and help us to
obtain more detailed information of the time-

varying signals. There exist many different TFRs
generally falling into two categories, which are
linear TFRs and nonlinear TFRs [3,20,22].
Linear TFRs mainly include the STFT and the
wavelet transform.
Nonlinear TFRs include the WVD, the

ambiguity function (AF), smoothed versions of
the WVD, and Cohen’s class


Discrete Orthogonal Transforms and Its Applications: Dr. J. Sheeba Rani

Uploaded by

Copyright:

Available Formats

Discrete Orthogonal Transforms and Its Applications: Dr. J. Sheeba Rani

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Discrete Orthogonal Transforms and Its Applications: Dr. J. Sheeba Rani

Uploaded by

Copyright:

Available Formats

Discrete Orthogonal Transforms and its

Indian Institute of Space Science and Technology,

Indian Institute of Space Science and Technology

• To obtain a further information from the signal that is not

• Fast computation: convolution vs. Multiplication

• Efficient Storage and transmission

• Transformed data as a measurement/feature for high level

• Spatial Frequency Information , Multiresolution for further

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

KLT Transform (Optimal)

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

• These data samples are then arrayed in a matrix X, with one

Indian Institute of Space Science and Technology

Where A is the diagonal matrix of eigenvalues and R is the

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Step 1 Mean=[110,170,162.5,128.75,142.5,172.5] Centre data

-10.0000 20.0000 -20.0000 10.0000 Subtracting the mean image

Indian Institute of Space Science and Technology 04/27/2020

Step 5 Eigen vector and eigen values val =

Indian Institute of Space Science and Technology 04/27/2020

x11=Vect'*x1' Step 6 KL coefficient

Indian Institute of Space Science and Technology 04/27/2020

-10.0000 -10.0000 184.5000 -68.7500 12.5000 -12.5000

Indian Institute of Space Science and Technology 04/27/2020

The next equivalent optimal

computation time is high.

To overcome this problem

Indian Institute of Space Science and Technology

One Dimensional Two Dimensional

Indian Institute of Space Science and Technology

I1=[1 -1 2 -1;1 1 3 -4;2 -3 Application

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Fourier Transform Phase Reconstructed Magnitude

Reconstructed Phase Reconstructed Magnitude and

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

• Used in JPEG and MPEG

closely approximates the optimal KLT transform

KLT is is signal-dependent, and there is no general algorithm

The DCT possess data-independent basis functions and

The DCT provides a good trade-off between energy packing

The energy packing property of DCT is superior to that of any

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

Indian Institute of Space Science and Technology

The transform which minimizes the MSE

Indian Institute of Space Science and Technology

• Wavelets are functions that “wave” above and

Indian Institute of Space Science and Technology

Same in Frequency Domain

We need to know whether and also when an information