Sie sind auf Seite 1von 39

Wavelets and Multiresolution

By
Dr. Mahua Bhattacharya

ABV–INDIAN INSTITUTE OF INFORMATION TECHNOLOGY AND MANAGEMENT GWALIOR (M.P.), INDIA


INSTRUCTOR

►Dr. Mahua Bhattacharya


Associate Professor
ABV-IIITM Gwalior
Email Id- mahuabhatta@gmail.com
OBJECTIVE

➢ Fourier Transform
➢ Wavelet Transform
➢ Image Pyramid and System Block Diagram
➢ Subband Coding
➢ Multilevel Decomposition
➢ Discrete Fourier Transform using Haar Functions.
➢ Wavelets in MATLAB.
➢ Application
3
FOURIER TRANSFORM

Fourier Transform decomposes a function of time i.e. a signal, into frequencies


that make it up and breaks the signal into an alternative representation
characterized by sine and cosines.
➔ The Fourier Transform shows that any waveform can be rewritten as the
sum of sinusoidal functions.

4
LIMITATION OF FOURIER TRANSFORM

➢ Fourier Transform only gives what


frequency components exist in a signal.

➢ Fourier Transform cannot tell at what


time the frequency components occur.

5
WAVELET TRANSFORM

“The wavelet transform is a tool that cuts up data, functions or operators into different
frequency components, and then studies each component with a resolution matched to its
scale.”
➔ A Wavelet Transform is the representation of a function by wavelets.

6
“ A wavelet is a mathematical
function which is used to divide
a given function or continuous
time signal into different scale
components.

7
FOURIER VS. WAVELET

● FFT, basis functions: sinusoids


● Wavelet transforms: small waves, called wavelet
● FFT can only offer frequency information
● Wavelet: frequency + temporal information
● Fourier analysis doesn’t work well on discontinuous, “bursty” data like music,
video, power, earthquakes,…

8
FOURIER VS. WAVELET

➢ Fourier
❏ Loses time (location) coordinate completely
❏ Analyses the whole signal
❏ Short pieces lose “frequency” meaning

➢ Wavelets
❏ Localized time-frequency analysis
❏ Short signal pieces also have significance
❏ Scale = Frequency band
9
CONTINUOUS WAVELET TRANSFORM

Continuous wavelet transform is a function of 2 variable -


➢ τ- translation
➢ S - scale (frequency -1)
Where ψ(t)is the transform function, and is called the Mother wavelet.
.

10
MOTHER WAVELETS

11
STEPS TO COMPUTE CONT. WAVELET
TRANSFORM

Step 1.Take a mother wavelet and compare it to section at the start of the original
signal and calculate a correlation coefficient C.

Step 2. Shift the wavelet to the right and repeat step 1 until the whole signal is covered.

12
STEPS TO COMPUTE CONT. WAVELET
TRANSFORM

Step 3. Scale (stretch) the wavelet and repeat steps 1 through 2.

Step 4. Repeat steps 1 through 3 for all scales.

13
IMAGE PYRAMID AND SYSTEM BLOCK DIAGRAM

➢ Normally, we used to work with an image of constant size. But in some occasions,
we need to work with images of different resolution of the same image.

➢ For example, while searching for something in an image, like face, we are not sure at
what size the object will be present in the image. In that case, we will need to create
a set of images with different resolution and search for object in all the images.

➢ These set of images with different resolution are called Image Pyramids (because
when they are kept in a stack with biggest image at bottom and smallest image at
top look like a pyramid). 14
IMAGE PYRAMID AND SYSTEM BLOCK DIAGRAM

Total number of elements in a P+1 level pyramid for P>0 is

A pyramidal image structure 15


System Block Diagram for creating it
TYPES OF IMAGE PYRAMID

There are two kinds of Image Pyramids.


➢ Gaussian Pyramid
➢ Laplacian Pyramids
1) Gaussian Pyramid -
Higher level (Low resolution) in a Gaussian Pyramid is formed by removing
consecutive rows and columns in Lower level (higher resolution) image. Then each
pixel in higher level is formed by the contribution from 5 pixels in underlying level
with gaussian weights. By doing so, a M×N image becomes M/2×N/2 image.
16
AN EXAMPLE OF GAUSSIAN PYRAMID

YYYYYYYYYYYYYYYYYYYYYYYn An image pyramid of 4 levels.

17
TYPES OF IMAGE PYRAMID

2) Laplacian Pyramids -

➢ Laplacian Pyramids are formed from the Gaussian Pyramids. Laplacian


pyramid images are like edge images only. Most of its elements are
zeros.
➢ They are used in image compression. A level in Laplacian Pyramid is
formed by the difference between that level in Gaussian Pyramid and
expanded version of its upper level in Gaussian Pyramid.

18
AN EXAMPLE OF LAPLACIAN PYRAMID

19
Image Pyramid and System Block Diagram (Steps)

1. Compute reduced-resolution approximation of input image. Done by filtering


(approximation filter) and downsampling.
2. Upsample output of (1) and interpolate. This creates a prediction image.
3. Compute the differences between the prediction in (2) and the input image in (1).
This difference is called the prediction residual and can later be used to
reconstruct the original image.

20
ANOTHER EXAMPLE

21
SUBBAND CODING

An image is decomposed into a set of band-limited components, called


subbands, which can be reassembled to reconstruct the original image without
error.
➢ Two-band filter bank is used for one-dimensional subband coding and
decoding.
➢ Filter banks satisfying ˆx(n)=x(n) are called biorthogonal.

22
SUBBAND CODING

hvbjfskb nnnnnnnnnnnnnn

Two band filter bank Spectral properties of two band filter

23
ONE- DIMENSIONAL SUBBAND CODING

● The sampled input signal is convolved in a bank of M bandpass analysis filters with
frequency responses Hi(ejw) and down-sampled by factors ki that correspond to the
bandwidths of Hi(ejw).
● The frequency responses and down-sampling factors are usually chosen such that
the number of samples before and after subband decomposition is the same, i.e.,

24
Diagram of a one-dimensional M-channel subband coding scheme.

25
TWO DIMENSIONAL WAVELET TRANSFORM

26
MULTILEVEL DECOMPOSITION

27
AN EXAMPLE OF DECOMPOSITION (SUBBAND CODING)

The resultant four


subbands are
a) Approximation
b) Vertical detail
c) Horizontal detail
d) Diagonal details

28
TWO- DIMENSIONAL SUBBAND CODING

Most two-dimensional
subband
decompositions are
built by cascading one-
dimensional subband
filter banks.

29
2-D 4-BAND FILTER BANK

30
DISCRETE WAVELET TRANSFORM (DWT)

➢ Continuous wavelet transform: calculate wavelet coefficient at every possible


scale and shift.
➢ Discrete wavelet transform: choose scale and shift on powers of two (dyadic
scale and shift)

➔ “Subset” of scale and position based on power of two rather than every
“possible” set of scale and position in continuous wavelet transform
➔ Behaves like a filter bank: signal in, coefficients out .
➔ Down-sampling necessary (twice as much data as original signal).

31
DISCRETE WAVELET TRANSFORM (DWT)

32
HAAR TRANSFORM (DWT WITH HAAR FUNCTIONS)

● The basis functions of the Haar transform are the simplest known
orthonormal wavelets.
● These are used as filter banks in subband coding.
● The Haar wavelets are oldest and simplest known orthonormal wavelets
defined as T=HFH
where F: NXN image matrix,
H: NxN transformation matrix.
● Haar basis functions hk(z) are defined over the continuous, closed interval
[0,1] for k=0,1,..N-1 where N=2n .
33
HAAR BASIS FUNCTIONS

34
AN EXAMPLE OF MULTILEVEL DECOMPOSITION

Discrete wavelet transform


using Haar basis functions.

Approximations (64×64,128×128
and 256×256) that can be
obtained from wavelet transform.
35
WAVELETS IN MATLAB

1. Start MATLAB.
2. Run wavemenu (part of wavelet toolbox).
3. Select Wavelet-2D.
4. Load the image(preferably in *.mat format)
5. Experiment with different families, types, levels and results of different
compression levels.
6. Look at the families of wavelets by going to Display→Wavelet Display.

36
AN EXAMPLE

37
Applications

▰ DWT can be used to separate an image into lower resolution approximation


image as well as horizontal , vertical and diagonal detail components.
▰ The whole process can be repeated to compute multiple scale wavelet
decomposition.
▰ Multiresolution can be used for this down-sampled version of the original
image since there will be fewer local optima for the optimization methods to
get stuck in.

38
THANKS!
Any questions?
Find me at mahuabhatta@gmail.com

39

Das könnte Ihnen auch gefallen