Scribd
Upload
26 views

Transform Methods in Image Processing

This document outlines two lectures on transform methods in image processing. Lecture 1 introduces various transforms including the Discrete Fourier Transform (DFT), Discrete Cosine Transform (DCT), and Discrete Wavelet Transform (DWT). It discusses how these transforms represent images in the frequency domain and provides examples of their applications in problems like image restoration, edge detection, and compression. Lecture 2 will cover filtering, filter banks, and applications like JPEG compression and fingerprint matching. MATLAB demonstrations are used to illustrate key concepts from signal and image processing like the DFT, convolution, and analysis/synthesis waveforms.

Uploaded by

vishwanath
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
26 views

Transform Methods in Image Processing

This document outlines two lectures on transform methods in image processing. Lecture 1 introduces various transforms including the Discrete Fourier Transform (DFT), Discrete Cosine Transform (DCT), and Discrete Wavelet Transform (DWT). It discusses how these transforms represent images in the frequency domain and provides examples of their applications in problems like image restoration, edge detection, and compression. Lecture 2 will cover filtering, filter banks, and applications like JPEG compression and fingerprint matching. MATLAB demonstrations are used to illustrate key concepts from signal and image processing like the DFT, convolution, and analysis/synthesis waveforms.

Uploaded by

vishwanath
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 32

0

50

50

100

100

150

150

200

200

250

250
0

50

100

150

200

250

300

350

50

100

150

200

250

300

350

Transform Methods in Image


Processing
Mount Holyoke Mathematics Seminar
S. Allen Broughton - Spring Term, 2001
http://www.rose-hulman.edu/~brought/Epubs/mhc/mhctransimage.html

Background - 1

based on image processing and compression


courses taught for several years
taught to juniors & seniors in math, computer
science, physics/applied optics, and
electrical/computer engineering
course for Imaging Systems Certificate
collaboration of faculty from above departments

Background - 2

matrix algebra and Fourier series base, DSP


helpful
can get by with matrix algebra only
Matlab heavily used, Maple to a lesser
extent
course notes under continuous revision
http://www.rosehulman.edu/~brought/courses/ma490mip/
3

Outline of Talks

Lecture 1- Introduction, various transforms

Discrete Fourier Transform DFT and FFT


Discrete Cosine transform DCT (pictures only)
Discrete Wavelet transform DWT (pictures only)
motivating application: motion fields by digital image
correlation

Lecture 2- Convolution/Filtering, Filter Banks to


Wavelets

filtering and convolution


DIC application
filter banks
motivating application: JPEG and fingerprint image
compression

Lecture 1 - 1

Some image processing problems


MATLAB demos
signal and image models
time domain representation of signals and images
colour vs black and white
vector space models
transforms and processing

Lecture 1 - 2

frequency domain representation


Discrete Fourier Transform (DFT) of
signals and images
matrix representation
analysis and synthesis waveforms
computational issues

Image Processing Problems

restoration: deblurring photos, e.g., Hubble


telescope images
edge detection: medical imaging
denoising: part of restoration
compression: FBI finger print problem, JPEG
image storage, transmission, and retrieval
digital image correlation: detecting motion with
before and after photos
7

Matlab Examples

restoration - ansmid3.m
edge detection - edgedet.m
denoising - dftdemo.m

Image and Signal Models - 1

A (general) signal is a function on a time or


spatial domain (or combination) D=Dt
X : Dt R or X : Dt C,
in general
X L2(Dt )
energy of a signal is (proportional to) XX
Hermitian inner product
9

Image and Signal Models - 2

continuous models (domain is continuous)


audio: Dt = T (real line, time)
audio: Dt = T (interval on real line, time,
periodic signal)
image: Dt = R (rectangle in plane)
movie: Dt = R T (spatial time)
scientific data: Dt = V (3D volume)
scientific data: Dt can be quite an arbitrary set
10

Image and Signal Models - 3

discrete models
audio: Dt = Z (integers, bi-infinite, uniform
discrete sampling)
audio: Dt = ZN (integers mod N, finite, discrete
sampling, periodic signal, wrap around)
image: Dt = Z m Zn (rectangle in plane,
computer screen with wrap around scrolling )
movie: Dt = (Zm Zn ) ZN
11

Image and Signal Models - 4

continuous models
used for analysis and modeling physical processes
more difficult for students on the first pass, need
calculus/real analysis
discretize for computation

discrete models
concentrate on finite duration, discretely sampled
signals and images
theory simpler, need only linear algebra
well adapted for computation
12

Image and Signal Models - 5

signals are vectors in RN or CN = L2(ZN ), as


a convolution algebra
images are matrices in MR(m,n) or MC(m,n)
= L2(Zm Zn ), as a convolution algebra
signals and images are also quantized (nonlinear process)
8 bit sound has 256 values
24 bit colour has 3 colours at 256 levels each
13

Colour

8-bit pseudo-colour
256 color levels
colour lookup table

24 bit true colour


three colour matrices R, G, B
256 levels for each colour

assume images are monochrome for simplicity


show MATLAB example - eightbit.m

14

Vector & Matrix Models

good model for (additive) noise

Y=X+E (signal + noise)

superposition of basic signals and images is


a vector space process
image processing as a linear operator
X AX (signals)
X AXB (images)
A and B are matrices of the appropriate size
15

General Transform Process


1
~
~
T
process
T
X
X Y Y

T is some linear transform


the middle process is usually non-linear
examples of the middle process are quantization
(for compression), deblurring. truncation (for
denoising)
Matlab demo dftdemo.m
Matlab demo ansmid3.m again
16

Frequency Domain
Representation/Transforms - 1

Now comes the design part of the transform


process
1.
2.
3.

select some set of parameters of interest, they will


form a parameter space Df
for audio signals this is called the frequency domain
and the parameter selected is frequency
for each point Df in the frequency domain pick a
pure wave form E that typifies this parameter
value

17

Frequency Domain
Representation/Transforms - 2
4.
by

Define

X : D f C
X ( ) = X E

5. X ( ) is interpreted as the energy content at the


frequency
6. X is the frequency domain representation of the
signal X or the ???? transform of X
7. T : X X
is the ???? transform
8. T 1 : X X is the inverse ???? transform
18

Frequency Domain
Representation/Transforms - 3

Often the transform parameters are created in


the following way
1.

2.

3.

There is a lie group action on the space of signals


derived, in part from a geometric action on the time
or spatial domain, indeed the domain may be a group
The frequency domain is the space of irreducible
representations (Fourier transform) or may be the
group (continuous wavelet transform, windowed
Fourier transform)
pure wave forms may be irreducible characters
(Fourier transform) or nicely constructed signals on
the space using the group action
19

Frequency Domain
Representation/Transforms - 4

Example 1: standard Fourier transform

X is a signal on the reals Dt = R


frequency domain Df = R
E = exp(i t)

X ( ) =

X
(
t
)
e

it

dt

20

Frequency Domain
Representation/Transforms - 5

Example 2: Discrete Fourier Transform of a 1D


signal

X = [X(0), ... , X(N-1)]t a finite discretely sampled


signal on Dt = ZN
frequency domain Df = ZN
EN,k(r) = exp(2i kr/N), r ZN

X ( k ) = X E N , k =

N 1

r =0

r
X ( r ) exp( 2 i ( k ))
N

Matlab demo dft1demo3


21

Frequency Domain
Representation/Transforms - 6

Example 3: Discrete Fourier Transform of


an image

X = [X(i,j)] is an mn image in matrix form


Dt = Z m Z n
frequency domain Df = Z m Zn (vertical,
horizontal) frequencies
Em,n,k,l = E m,k (En,l)t as matrices
Matlab demo wavefft2cband.m (polar
decomposition)
22

Frequency Domain
Representation/Transforms - 7

Example 4: Windowed Fourier transform


of a signal

X is a signal on R
g is a windowing function of compact
support or a Gaussian, of norm 1
G = R2 acts via (p,q)X = eiqtX(t-p)
pure wave forms are eiqtg(t-p)

X ( p, q ) =

X
(
t
)
e

iqt

g (t p )dt
23

Frequency Domain
Representation/Transforms - 8

Example 5: Continuous Wavelet


transform of a signal

X is a signal on R
is an appropriate mother wavelet function, of
norm1, compact support or a derivative of a
Gaussian
G = ax+b group acts via (a,b)X = X((t-b)/a)
pure wave forms are a-1/2((t-b)/a)

X ( a, b) =
a

t b
X
(
t
)

(
)
dt

24

Frequency Domain
Representation/Transforms - 9

Maple worksheet wft.mws


Maple worksheet cwt.mws

25

Matrix Representation
of DFT -1

(Ek)* is the Hermitian transpose of Ek


as a matrix product
X ( k ) = E k* X
Set FN = [E0, ..., EN-1]*

X = FN X

FN(r,s) = zrs , z=exp(-i/N)

(FN)* FN = NIN thus we have the inversion


formula
*
N

1
X = F X
N

26

Matrix Representation
of DFT -2

The 2D DFT of an image is given by


doing 1D transforms to all columns and
then all rows
X FmX(Fn)t = (FmX)(Fn)t

27

Analysis and Synthesis


Waveforms - 1

Suppose Y= MX is any matrix transform


on 1D signals
Inverse transforms is X=M-1Y
E0 , ..., EN-1 columns of Mt (transposed
rows of M), are the analysis or
decomposition waveforms
Fourier coefficients of X are given by
the matrix product (Er)tX
28

Analysis and Synthesis


Waveforms - 2

F0 ,...,FN-1 columns of M-1 , are the


synthesis or reconstruction waveforms
the original signal
X=M-1Y= [F0 ,...,FN-1 ]Y
is a linear combination of the synthesis
waveforms and the coefficients are the
Fourier coefficients
29

Analysis and Synthesis


Waveforms - 3

Matlab demos of analysis and synthesis

waveformsDFT.m
analsynDCT.m
analsynDWT.m

30

Computational Issues

DFT, DCT and DWT all have fast


computational algorithms

fast algorithm for DFT is known as FFT


(Cooley-Tukey algorithm)
sample at 2b points to get speed
performance in O(Nlog(N)) not O(N3)
Matlab demo fftperform.m

31

End of Lecture 1

Questions, Questions, Questions!!!

32

You might also like