EECE/CS 253 Image Processing: Richard Alan Peters II

EECE\CS 253 Image Processing
Lecture Notes: The 1&2-Dimensional

Lecture Notes Fourier Transforms
Richard Alan Peters II

Department of Electrical Engineering and
Computer Science
Fall Semester 2012
This work is licensed under the Creative Commons Attribution-Noncommercial 2.5 License. To view a copy of this license, visit http://creativecommons.org/licenses/by-nc/2.5/ or
send a letter to Creative Commons, 543 Howard Street, 5th Floor, San Francisco, California, 94105, USA.
Signal:
A measurable phenomenon that changes over time
or throughout space.
image
sound
code
01101000101101110110010110001
2013-03-08
1999-2013 by Richard Alan Peters II
Signals: Space-Time vs. FrequencyDomain Representation

Space/time representation: a graph of the
measurements with respect to a point in time and/or
positions in space.
Fact: signals undulate (otherwise theyd contain
no information).
Frequency-domain representation: an exact
description of a signal in terms of its undulations.
2013-03-08
Origin of Sounds
2013-03-08
The mechanical vibrations of an object in an atmosphere.

Vibrations: internal elastic motions of the material.
The surface of the object undulates causing compressions
and rarefactions in the air which propagate through the
air away from the surface.
An object vibrates with different modes.
A mode is a vibratory pattern with a distinctive shape
part of the object surface moves out while another part
moves in a standing wave.
Vibratory Modes / Standing Waves: Examples
displacement from rest position
internal pressure
string modes
pipe modes
2013-03-08
Note that
the modes
are all
sinusoids.
Note that
the negatives
of these also
will occur
Sound Waves:
Emerge from the superposition of the modes.
pipe sound
string sound
2013-03-08
Sound Waves:
The vibratory modes

add up to one complex
motion that pushes
the air around the
vibrating object
Odd-order
harmonics
pipe sound
2013-03-08
string sound
Even-order
harmonics
Emerge from the superposition of the modes.
Fact: Any Real Signal has

a Frequency-Domain
Representation
Odd-order harmonics
sq (t ) =
2p
1
sin
2
n
1
t
+
(
)
+
2
n
1
l
n=-
The modes shown (blue)

sum to the rippling square
wave (black).
As the number of modes
in the sum becomes large,
it approaches a square
wave (red).
2013-03-08
Frequency-Domain Representation
Any periodic signal can be described by a sum of sinusoids.
sq (t ) =
2p
1
n
t
sin
2
1
+
(
)
n=- 2n + 1
The sinusoids are called

basis functions.
The multipliers are called
Fourier coefficients.
2013-03-08
sq (t ) =
2p
1
n
t
sin
2
1
+
(
)
n=- 2n + 1

basis functions.
2013-03-08
Basis
functions
10
sq (t ) =
2p
1
n
t
sin
2
1
+
(
)
n=- 2n + 1

basis functions.
2013-03-08
The Fourier
coefficients
(of a square
wave).
11
Example: Partial Sums of a Square Wave

The limit of the
given sequence
of partial sums1
is exactly a
square wave
1 sine
2 sines
4 sines
the limit as n
approaches
infinity of the
sum of n sines.
1
8 sines
2013-03-08
16 sines
32 sines
12
Anatomy of a Sinusoid
f (t)
2p
f (t ) = A sin t - f
l
2013-03-08
13
The Inner Product: a Measure of Similarity

The similarity between functions f and g on the interval (- / 2, / 2)
can be defined by
l /2
f,g =
f (t ) g * (t ) dt
-l /2
where g * (t ) is the complex conjugate of g (t ).
This number, called the inner product of f and g , can also be

thought of as the amount of g in f or as the projection of f onto g .
If f and g have the same energy, then their inner product is
maximal if f = g . On the other hand if f , g = 0, then f and g
have nothing in common.
2013-03-08
14
Inner Products
a function, f
pointwise product f(t)g(t)
1024
g is a component of f
f (t ) g (t )dt 0.12
2013-03-08
15
Inner Products
a function, f
pointwise product f(t)h(t)
1024
h is a not a comp. of f
f (t )h(t )dt 0
2013-03-08
16
Inner Product of a Periodic Function and a Sinusoid
3 different
representations
2013-03-08
17
real number results

yield the amplitude
of that sinusoid in
the function.
2013-03-08
18
Complex number result

yields the amplitude and
phase of that sinusoid in
the function.
2013-03-08
19
The Fourier Series
f (t ) = A0 +
n=1
is the decomposition of a -periodic

signal into a sum of sinusoids.
2p n
2p n
An cos
t + Bn sin
t
l
l
periodic : $l such that f (t nl ) = f (t )
2
An =
l
2
Bn =
l
2013-03-08
l /2
-l /2
l /2
-l /2
The representation of a
function by its Fourier
Series is the sum of sinusoidal basis functions
multiplied by coefficients.
2p n
f (t ) cos
t - jn dt for n 0
Fourier coefficients are

generated by taking the
inner product of the
function with the basis.
2p n
f (t ) sin
t - jn dt for n 0
The basis functions

correspond to modes
of vibration.
20
can also be written in terms

of complex exponentials
The Fourier Series

f (t ) =
Ce
n
n=-
i 2lpn t
Cn e
2 pn
+
+fn
l
n=-
2pn
2pn
+
+
+
cos
f
sin
f
C
t
C
t
n
n
n
n
l
l
n=-
Cn = Cn e
+ ifn
l /2
2 pn
-i
t
1
l
dt
= f (t ) e
l -l /2
l /2
2p n
2p n
1
= f (t ) cos
t + fn + Cn sin
t + fn dt
l
l -l /2
2013-03-08
i = -1
Cn = Cn e+ifn
eix = cos x + i sin x
f (t + nl ) = f (t )
for all integers n
21
The Fourier Series
Contd. on next page.
Relationship between the real and the complex Fourier Series
f (t ) = [ An cos wnt + Bn sin wnt ], where wn =

n=0
2pn
l
l /2
l /2
2
= f (h ) cos wn h d h cos wnt + f (h ) sin wn h d h sin wn t
l n=0 -l /2
-l /2
l /2
2
= f (h )[cos wn h cos wnt + f (h ) sin wn h sin wnt ] d h
l n=0 -l /2
l /2
The sine-plus-cosine
1
= f (h ) cos (wn h - wnt ) d h
form results from the
l n=0 -l /2
projection of f onto a
cosine that is in phase
with the current time.
2013-03-08
22
Relationship between the real and the

complex Fourier Series (contd.)
Claim:
0=
sin (w h - w t ).
n
n=-
l /2
Therefore:
-l /2
Thus:
f (h ) sin (wn h - wnt ) d h = 0.
n=-
l /2
1
-i f (h ) sin (wn h - wn t ) d h = 0.
l n=- -l /2
Then add
zero to the
equation at
the end of
the previous
page:
l /2
l /2
1
1
f (t ) = f (h ) cos (wn h - wnt ) d h - i f (h ) sin (wn h - wnt ) d h .

l n=- -l /2
l n=- -l /2
2013-03-08
23

l /2
l /2
1
1
f (t ) = f (h ) cos (wn h - wnt ) d h - i f (h ) sin (wn h - wnt ) d h

l n=- -l /2
l n=- -l /2
l /2
1
= f (h )[cos wn (h - t ) - i sin wn (h - t )] d h
l n=- -l /2
l /2
1
= f (h ) e-iwn (h-t )d h
l n=- -l /2
l /2
Then some algebraic

manipulations lead to
the result.
2 pn
2 pn
h
-i
+i
t
1
l
l
h
h
f
e
d
e
=
(
)
n=- l -l /2
Ce
n
n=-
2013-03-08
i 2lpn t
n=-
ifn
Cn e e
i 2lpn t
Cn e
t f
2 pn
+
+ n
l
n=-
24

l /2
l /2
1
1
f (t ) = f (h ) cos (wn h - wnt ) d h - i f (h ) sin (wn h - wnt ) d h

l n=- -l /2
l n=- -l /2
l /2
1
= f (h )[cos wn (h - t ) - i sin wn (h - t )] d h
l n=- -l /2
l /2
1
= f (h ) e-iwn (h-t )d h
l n=- -l /2
l /2
Then some algebraic

manipulations lead to
the result.
2 pn
2 pn
h
-i
+i
t
1
l
l
h
h
f
e
d
e
=
(
)
n=- l -l /2
Ce
n
n=-
2013-03-08
i 2lpn t
n=-
ifn
Cn e e
i 2lpn t
Cn e
t f
2 pn
+
+ n
l
n=-
25
Why are Fourier Coefficients Complex Numbers?

f (t ) =
Cn e
+i
2 pn
t
l
where Cn = Cn e +ifn .
n=-
Cn represents the
amplitude, A=|Cn|, and
relative phase, , of
that part of the original
signal, f (t), that is a
sinusoid of frequency
n = 2n / .
2013-03-08
intensity
0
0
frequency, = 1/
26
What about real + imaginary?
F (w ) = ( N2A co s )[d (w + N l ) + d (w - N l )]
+ i ( N2A sin )[- d (w + N l ) + d (w - N l )]
The FS of a cosine is a pair of

impulses with complex amplitudes
2013-03-08
27
The real and imaginary

parts at the positive
frequency, N/ ...
Real + Imaginary to
Magnitude & Phase
form a magnitude, NA/2, and a phase, 0.

2013-03-08
28
The real and imaginary

parts at the negative
frequency, -N/ ...
Real + Imaginary to
Magnitude & Phase
form a magnitude, NA/2, and a phase, -0.

2013-03-08
29
Fourier Series of a Square Wave

Time-domain
signal
Fourier
magnitude
Fourier
phase
2013-03-08
30
The Fourier Transform

is the decomposition of a nonperiodic signal into a
continuous sum* of sinusoids.
F ( w ) = F (w ) e
iF(w )
= f (t ) e i 2 p w t dt
-
= f (t )[cos ( 2pw t ) + i sin ( 2pw t )] dt

-
f (t ) = F (w ) e
-i 2 p w t
d w = F (w ) e-i(2 p w t +F(w)) d w
-
= F (w ) [cos ( 2pw t ) - i sin (2pw t )] d w

-
= F (w ) [cos ( 2pw t +F (w )) - i sin (2pw t +F (w ))] d w

-
2013-03-08
i.e., an integral.
31
Mammals Use the FT in Hearing
2013-03-08
32
The Discrete Fourier Transform

A discrete signal, {hk k = 0,1, 2, , N 1 }, of finite length N can be repre -
sented as a weighted sum of N sinusoids, e i 2 k n /N n = 0,1, 2, , N 1

through
N 1
hk = H n e i 2 k n /N
n= 0
where the set, { H n n = 0,1, 2, , N 1}, are the Fourier coefficients

defined as the projection of the original signal onto sinusoid, n, given by :
1 N 1
Hn =
hk e +i 2 k n /N
N k=0
2013-03-08
33
The Two-Dimensional Fourier Transform

Primary Uses of the FT in Image Processing:
2013-03-08
Explains why down-sampling can add distortion to

an image and shows how to avoid it.
Useful for certain types of noise reduction,
deblurring, and other types of image restoration.
For feature detection and enhancement, especially
edge detection.
34
The Fourier Transform: Discussion

The expressions
F (w ) = f (t ) e-i 2 p w t dt = f (t ) , e+ i 2 p w t
-
continuous signals
defined over all
real numbers
and
1
Hn =
N
N -1
he
-i 2 p k n / N
= hk , e
+i 2 p k n /N
n= 0
discrete signals
with N terms or
samples.
for the Fourier coefficients are inner products which can be

thought of as measures of the similarity between the functions
f (t ) and e + i 2 t for t ( , ) or between the sequences
N -1
+ i 2 p k n / N N -1
{ hk } k =0 and { e
} k =0 .
2013-03-08
35
The Fourier Transform:

Discussion (contd.)
In the context of inner products, the complex exponentials
{ e-i2pw t
} and { e-i 2pk n/N
w and w (- , )
, -2, -1, 0,1, 2,
are called orthogonal sets since they have the property:
e-i 2 p w 1 t , e-i 2p w 2 t = e-i 2 p w 1 t e+ i 2 p w 2 t dt = { 0,
, if w 1= w 2
if w 1 w 2
-i 2 p j n / N
-i 2 p k n / N
, e
= e-i 2 p j n /N e+i 2 p k n /N = { 0,
N -1
c , if j = k
if j k
n= 0
The function
sets are called
orthogonal
basis sets
They are called basis sets since for any function1, f (t), of a real variable
there exists a complex-valued function F(w), and for any sequence1, hk , there
exist complex numbers, Hn , such that
f (t ) = F (w ) e
-i 2 p w t
2013-03-08
N -1
d w and hk = H n e-i 2p k n /N .
n= 0
1 with
finite energy.
36

Discussion (contd.)
Consider the 2-dimensional functions
{e
-i 2 p ( ux + vy )
u , v, x, y }
and
{e
jm
-i 2 p ( M
+ kn
)
N
j , m 0, ..., M - 1, k , n 0,..., N- 1
These are, likewise, orthogonal:

e-i 2 p(u1x+v1 y) , e-i 2 p (u2 x+v2 y)
e-i 2 p (u1x+v1 y) e+ i 2 p (u2 x+v2 y) dxdy
- -
=
j m k n
-i 2 p 1 + 1
M
N
j m k n
-i 2 p 2 + 2
M
N
, e
M -1 N -1
m= 0 n= 0
( jMm + kNn) e+i 2p( jMm + kNn)
-i 2 p
= e
=
2013-03-08
, if u1 =u2 and v1 =v2

0, otherwise
c , if j1 = j2 and k1 = k2
0, otherwise
.
37

Discussion (contd.)
Therefore
{e
- i 2 p ( ux + vy )
u , v, x, y } and
{e
-i2p(
jm
M
kn
N
j, k , m , n, M
are orthogonal basis sets. This suggests that function f (x,y) defined on the
real plane, and sequence {{ hmn }} for integers m and n have analogous
Fourier representations,

f ( x, y ) =
M - 1N - 1
F (u , v) e+i 2 p ( ux+vy ) dudv and hmn = H jk e
jm kn
+i 2 p +
M
N
j= 0 k = 0
- -
where the Fourier coefficients are given by

F (u , v) =
M -1 N -1
jm kn
-i 2 p +
M
N
f ( x, y) e-i 2 p ( ux+vy ) dxdy and H jk = hmn e

m=0 n=0
- -
(True for finite energy functions f (x,y) and {{ hmn }}.)

2013-03-08
38
Continuous Fourier Transform
The continuous Fourier

transform assumes a
continuous image exists
in a finite region of an
infinite plane.
The BoingBoing Bloggers
2013-03-08
39
Discrete Fourier Transform

The discrete Fourier
transform assumes a
digital image exists on a
closed surface, a torus.
2013-03-08
40
Discrete Fourier Transform

The discrete Fourier
transform assumes a
digital image exists on a
closed surface, a torus.
2013-03-08
41
The 2D Fourier Transform of a Digital Image

Let I(r,c) be a single-band (intensity) digital image with R
rows and C columns. Then, I(r,c) has Fourier representation
R-1 C -1
I ( r , c ) = I ( v ,u ) e
vr uc
+ i 2 p +
R C
u =0 v=0
where
R-1 C -1
these complex
exponentials are
2D sinusoids.
vr uc
-i 2 p +
R C
I (v,u ) = RC1 I(r ,c) e

r =0
c=0
are the R x C Fourier coefficients.

2013-03-08
42
What are 2D sinusoids?

To simplify the situation assume R = C = N. Then
i 2
vr
R
+ uc
C
=e
2
(vr + uc)
N
=e
2
( r sin + c cos)
N
where
v = w sin , u = w cos , w = v 2 + u 2 , and

Write
Note: since images are indexed

by row & col with r down and c
to the right, is positive in the
clockwise direction.
N
= ,
w
Then by Eulers relation,
i 2 p l1 (r sin + c cos)
= tan -1 ( uv ) .
= cos [ 2lp (r sin + c cos )] i sin [ 2lp (r sin + c cos )].

2013-03-08
43
What are 2D sinusoids? (contd.)

Both the real part of this,
Re e
} = + cos[
2p
l
(r sin + c cos )]
2p
l
(r sin + c cos )]
and the imaginary part,
Im e
} = sin [
are sinusoidal gratings of unit amplitude, period and direction .

Then
w
2pw
is the radian frequency, and
the frequency, of the wavefront
N
N
and =
2013-03-08
N
is the wavelength in pixels in the wavefront direction.
w
44
2D Sinusoids:
I ( r , c) =
2p
A
cos (r sin + c cos ) + j + 1
l
... are plane waves with

grayscale amplitudes,
periods in terms of lengths, ...
orientation
= phase shift
2013-03-08
45
2D Sinusoids:
2013-03-08
... specific orientations,

and phase shifts.
46
The Fourier Transform of an Image

c
I
2013-03-08
Re[F{I}]
Im[F{I}]
47
Points on the Fourier Plane

If R=C=N the point at
column freq. u and row
freq. v represents a
sinusoid with freq. and
orientation .
If R C then = 1/
where is the length of
vector (C/u, R/v) and the
wavefront orientation is
= tan-1[(v/R)/(u/C)].
2013-03-08
48
Points on the Fourier Plane (of a Digital Image)

In the Fourier transform of an RC digital image,
positions u and v indicate the number of repetitions
of the sinusoid in those directions. Therefore the
wavelengths along the column and row axes are
u = Cu
and v =
R
v
-v direction
pixels,
and the wavelength in the wavefront direction is
wf
More about
this later
(pp. 66-87).
of a digital image
-1
2
2 2
= RC (uR) + (vC ) .
wf =
v
R
, and
1
2
2
(uR) + (vC ) cycles.
RC
(0,0)
The wavefront direction is given by
wf = tan -1 ( v ) = tan -1 ( vu CR ).
u
2013-03-08
row freq.
column freq.
u direction
u = Cu , v =
- direction
The frequency is the fraction of the sinusoid

traversed over one pixel,
Note that the wave

front direction =
only if R=C.
49
Points on the Fourier Plane
Note that is the wavefront

direction only if R=C.
x
This point represents this particular sinusoidal grating
2013-03-08
50
The Value of a Fourier Coefficient

is a complex
number with a
real part and an
imaginary part.
If you represent
that number as a
magnitude, A, and
a phase, ,
..these represent the amplitude
and offset of the sinusoid with
frequency and direction .*
2013-03-08
*See
p. 49.
51
The Value of a Fourier Coefficient

The magnitude and phase
representation makes
more sense physically
since the Fourier magnitude, A (,), at point (,)
represents the amplitude
of the sinusoid
and the phase, (,),
represents the offset of the
sinusoid relative to origin.
2013-03-08
52
The Fourier Coefficient at (u,v)

So, the point (u,v) on the
Fourier plane
represents a sinusoidal
grating of frequency
and orientation .*
The complex value, F(u,v),
of the FT at point (u,v)
represents the amplitude, A, and the phase
offset, , of the sinusoid.
2013-03-08
*See
p. 49.
53
The Sinusoid from the Fourier Coeff. at (u,v)
Note that the wave

front direction =
only if R=C.
2013-03-08
54
FT of an Image (Magnitude + Phase)
I
2013-03-08
log{|F{I}|2+1}
[F{I}]
55
FT of an Image (Real + Imaginary)
I
2013-03-08
Re[F{I}]
Im[F{I}]
56
The Power Spectrum

The power spectrum of a signal is the square of
the magnitude of its Fourier Transform.
For display,
the log of
the power
spectrum is
often used.
I (u ,v) = I (u ,v) I*(u ,v)

= [ Re I (u ,v) + i Im I (u ,v)][ Re I (u ,v) - i Im I (u ,v)]
2
= [ Re I (u ,v)] + [ Im I (u ,v)] .
At each location (u,v) it indicates the squared intensity of the

frequency component with period l = 1 / u 2 + v 2 and orientation
q = tan -1 (v / u ).
For display in Matlab:

PS = fftshift(2*log(abs(fft2(I))+1));
2013-03-08
57
On the Computation of the Power Spectrum

The power spectrum (PS) is defined by PS( I ) = F {I(u, v )} .
We take the base-e logarithm of the PS in order to view it. Otherwise its dynamic range could be
too large to see everything at once. We add 1 to it first so that the minimum value of the result is
0 rather than infinity, which it would be if there were any zeros in the PS. Recall that
log( f 2) = 2log( f ).
Multiplying by 2 is not necessary if you are generating a PS for viewing, since you'll probably
have to scale it into the range 0-255 anyway. It is much easier to see the structures in a Fourier
plane if the origin is in the center. Therefore we usually perform an fftshift on the PS before it is
displayed.
>> PS = fftshift(log(abs(fft2(I))+1));
>> M = max(PS(:));
>> image(uint8(255*(PS/M)));
2
If the PS is being calculated for later computational use -- for example the autocorrelation of a
function is the inverse FT of the PS of the function -- it should be calculated by
>> PS = abs(fft2(I)).^2;
2013-03-08
58
The Uncertainty Relation

space
frequency
FT
If x y is the extent of
the object in space and
if u v is its extent in
frequency then,
x y u v
space
frequency
FT
2013-03-08
1
16 2
A small object in space

has a large frequency
extent and vice-versa.
59
The Uncertainty Relation

small extent
IFT
small extent
large extent
space
IFT
2013-03-08
A symmetric pair of
lines in the frequency
domain becomes a
sinusoidal line in the
spatial domain.
large extent
frequency
small extent
Recall: a symmetric
pair of impulses in the
frequency domain
becomes a sinusoid in
the spatial domain.
small extent
space
large extent
frequency
large extent
60
The Fourier Transform of an Edge
edge
2013-03-08
Power Spectrum
Phase Spectrum
61
The Fourier Transform of a Bar
bar
2013-03-08
Power Spectrum
Phase Spectrum
62
Coordinate Origin of the FFT
Center =
(floor(R/2)+1, floor(C/2)+1)
Even
Odd
Even
Odd
Image Origin
Image Origin
Weight Matrix Origin
Weight Matrix Origin
After FFT shift
After FFT shift
After IFFT shift
After IFFT shift
2013-03-08
63
Matlabs fftshift and ifftshift

I = ifftshift(J):
J = fftshift(I):
origin
origin
from FFT2
or ifftshift
after fftshift
J ( R/2 +1, C/2 +1) I (1,1)
I (1,1) J ( R/2 +1, C/2 +1)
where x = floor(x) = the largest integer smaller than x.

2013-03-08
64
Matlabs fftshift and ifftshift

J = fftshift(I):
I (1,1) J ( R/2 +1, C/2 +1)

2
I = ifftshift(J):
J ( R/2 +1, C/2 +1) I (1,1)
where x = floor(x) = the largest integer smaller than x.

2013-03-08
65
Points on the Fourier Plane (of a Digital Image)

In the Fourier transform of an RC digital image,
positions u and v indicate the number of repetitions
of the sinusoid in those directions. Therefore the
wavelengths along the column and row axes are
u = Cu
and v =
R
v
-v direction
pixels,
and the wavelength in the wavefront direction is
wf
of a digital image
-1
2
2 2
= RC (uR) + (vC ) .
wf =
v
R
, and
1
2
2
(uR) + (vC ) cycles.
RC
(0,0)
The wavefront direction is given by
wf = tan -1 ( v ) = tan -1 ( vu CR ).
u
2013-03-08
row freq.
column freq.
u direction
u = Cu , v =
- direction
The frequency is the fraction of the sinusoid

traversed over one pixel,
Note that the wave

front direction =
only if R=C.
66
Geometrical Derivation of Wavelength

Since the wavelength of a horizontal* wave
is R/v and that of a vertical is C/u, the line
segment, h, that connects the two distances
is parallel to the wavefront. The wavelength
is the altitude of the triangle w.r.t. h (the
perpendicular to h that intersects the origin).
The area of the triangle, one half of base
times height, is independent of the leg that is
taken to be the base. Equate the expression
with base C/u to that with base h, to find
w.r.t R, C, v, u, & h. Then replace h with its
expression as a function of R, C, v, & u to
get the final expression.
*The equivalue lines are horizontal in a wave with a vertical wave front and vice versa.
2013-03-08
67
Coordinates and Directions in the Fourier Plane

decreasing rows
(-r,-c)
(-r,+c)
(-r,-c)
>0
<0
<0
(+r,-c)
(-r,+c)
increasing cols
(+r,+c)
decreasing cols
(+r,-c)
>0
(+r,+c)
increasing rows
Since rows increase down and columns to the right, slopes and
angles are opposite those of a right-handed coordinate system.
2013-03-08
68
Inverse FFTs of Impulses

horizontal is the
wavefront direction.
fftshifted
highest-possible-frequency horizontal sinusoid (C is even)

2013-03-08
69

vertical is the
fftshifted
highest-possible-frequency vertical sinusoid (R is even)

2013-03-08
70

a checker-board
pattern.
fftshifted
highest-possible-freq horizontal+vertical sinusoid (R & C even)

2013-03-08
71

horizontal is the
fftshifted
lowest-possible-frequency horizontal sinusoid

2013-03-08
72

vertical is the
fftshifted
lowest-possible-frequency vertical sinusoid

2013-03-08
73

negative diagonal is
the wavefront direction.
fftshifted
lowest-possible-frequency negative diagonal sinusoid

2013-03-08
74

positive diagonal is
the wavefront direction.
fftshifted
lowest-possible-frequency positive diagonal sinusoid

2013-03-08
75
Frequencies and Wavelengths in the Fourier Plane

512 columns
384 rows
+u direction
+v direction
Note this
and this.
frequencies: (u,v) = (4,3); wavelengths: (u, v) = (128,128)

How can that be?
2013-03-08
76
u = C / u
384 rows
512 columns
u = # of complete cycles
in the horizontal direction
frequencies: (u,v) = (1,0); wavelength: u= 512

2013-03-08
77
v = # of complete
cycles in the
vertical direction
v = R / v
384 rows
512 columns
frequencies: (u,v) = (0,1); wavelength: v= 384

2013-03-08
78
u = C / u
384 rows
512 columns

2013-03-08
79
v = # of complete
cycles in the
vertical direction
v = R / v
384 rows
512 columns

2013-03-08
80
u = C / u
384 rows
512 columns

2013-03-08
81
v = # of complete
cycles in the
vertical direction
v = R / v
384 rows
512 columns

2013-03-08
82
In the Fourier plane of a

square image, the orientation
of the line through the point
pair = the orientation of the
wave front in the image. Not
so for a non-square image.
In the F plane the angle is -45

in this image its about -53
(yellow line). Thats because
the fraction of R covered by
one pixel is 4/3 the fraction of
C covered by one pixel.
384 rows
512 columns
Also as a result, the

wavelength is 102.4.
frequencies: (u,v) = (3,3); wavelengths: (u, v) = (170 ,128)

2013-03-08
83
In general the slope of the

wavefront direction in the
image is given by (v/R) / (u/C).
Therefore its angle is
Frequencies
and
vC Wavelengths in the Fourier Plane
= tan ,
-1
wf
uR
512 columns
wf
wf
384 rows
wf
and the wavelength is:
frequencies: (u,v) = (3,3);

2013-03-08
-1
2
2 2
wavelengths: wf(=u, RC
v)(uR
= )(170
,128)
+ (vC
) .
84
384 rows
512 columns
frequencies: (u,v) = (3,3); wavelengths: (u, v) = (170 ,128)

2013-03-08
85
384 rows
512 columns

2013-03-08
86
The ratio R/C = in this image. Therefore at

frequency (4,3) the wave front angle is
-1 3 512
-1 3 4
= tan -1 (in
tan
tan
1
45
,Fourier
=
=
=
)
Frequencies
and
Wavelengths
the
wf
4 384
4 3
Plane
512 columns
wf
wf
384 rows
wf
and the wavelength is

wf
2013-03-08
2
2
= 384 512 (3 384) + (4 512)
1
2
83.67,
87
Power Spectrum of an Image
2013-03-08
88
Relationship between Image and FT

phase
power spectrum
power spectrum
2013-03-08
phase
89
Features in the FT and in the Image

Lines in
the Power
Spectrum
are
perpendicular to
lines in the
image.
2013-03-08
90
Fourier Magnitude and Phase
2013-03-08
91
Fourier Magnitude
log
F {I}
2013-03-08
92
Fourier Phase
F { I}
2013-03-08
93
Q: Which contains more visually relevant

information; magnitude or phase?
original image
2013-03-08
Fourier log
magnitude
Fourier phase
94
Magnitude Only Reconstruction
2013-03-08
95
Phase Only Reconstruction
2013-03-08
96

EECE/CS 253 Image Processing: Richard Alan Peters II

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

EECE/CS 253 Image Processing: Richard Alan Peters II

Uploaded by

Copyright:

Available Formats

EECE\CS 253 Image Processing

Lecture Notes: The 1&2-Dimensional

Richard Alan Peters II

1999-2013 by Richard Alan Peters II

Signals: Space-Time vs. FrequencyDomain Representation

1999-2013 by Richard Alan Peters II

The mechanical vibrations of an object in an atmosphere.

Vibratory Modes / Standing Waves: Examples

displacement from rest position

1999-2013 by Richard Alan Peters II

Emerge from the superposition of the modes.

1999-2013 by Richard Alan Peters II

The vibratory modes

1999-2013 by Richard Alan Peters II

Emerge from the superposition of the modes.

Fact: Any Real Signal has

The modes shown (blue)

1999-2013 by Richard Alan Peters II

The sinusoids are called

1999-2013 by Richard Alan Peters II

The sinusoids are called

1999-2013 by Richard Alan Peters II

The sinusoids are called

1999-2013 by Richard Alan Peters II

Example: Partial Sums of a Square Wave

1999-2013 by Richard Alan Peters II

The Inner Product: a Measure of Similarity

where g * (t ) is the complex conjugate of g (t ).

This number, called the inner product of f and g , can also be

1999-2013 by Richard Alan Peters II

pointwise product f(t)g(t)

1999-2013 by Richard Alan Peters II

pointwise product f(t)h(t)

1999-2013 by Richard Alan Peters II

Inner Product of a Periodic Function and a Sinusoid

1999-2013 by Richard Alan Peters II

Inner Product of a Periodic Function and a Sinusoid

real number results

1999-2013 by Richard Alan Peters II

Inner Product of a Periodic Function and a Sinusoid

Complex number result

1999-2013 by Richard Alan Peters II

The Fourier Series

is the decomposition of a -periodic

periodic : $l such that f (t nl ) = f (t )

Fourier coefficients are

The basis functions

1999-2013 by Richard Alan Peters II

can also be written in terms

The Fourier Series

1999-2013 by Richard Alan Peters II

The Fourier Series

Contd. on next page.

Relationship between the real and the complex Fourier Series

f (t ) = [ An cos wnt + Bn sin wnt ], where wn =

1999-2013 by Richard Alan Peters II

Relationship between the real and the

Contd. on next page.

f (h ) sin (wn h - wnt ) d h = 0.

f (t ) = f (h ) cos (wn h - wnt ) d h - i f (h ) sin (wn h - wnt ) d h .

1999-2013 by Richard Alan Peters II

Relationship between the real and the

f (t ) = f (h ) cos (wn h - wnt ) d h - i f (h ) sin (wn h - wnt ) d h