Computer Vision

Spring 2012 15-385,-685

Instructor: S. Narasimhan

Wean Hall 5409
T-R 10:30am 11:50am
Frequency domain analysis and
Fourier Transform

Lecture #4
How to Represent Signals?
Option 1: Taylor series represents any function using
polynomials.




Polynomials are not the best - unstable and not very
physically meaningful.

Easier to talk about signals in terms of its frequencies
(how fast/often signals change, etc).
Jean Baptiste Joseph Fourier (1768-1830)
Had crazy idea (1807):
Any periodic function
can be rewritten as a
weighted sum of Sines and
Cosines of different
frequencies.
Dont believe it?
Neither did Lagrange,
Laplace, Poisson and
other big wigs
Not translated into
English until 1878!
But its true!
called Fourier Series
Possibly the greatest tool
used in Engineering
A Sum of Sinusoids
Our building block:


Add enough of them to
get any signal f(x) you
want!

How many degrees of
freedom?

What does each control?

Which one encodes the
coarse vs. fine structure of
the signal?
) +| ex Asin(
Fourier Transform
We want to understand the frequency e of our signal. So, lets
reparametrize the signal by e instead of x:

) +| ex Asin(
f(x) F(e)
Fourier
Transform
F(e) f(x)
Inverse Fourier
Transform
For every e from 0 to inf, F(e) holds the amplitude A and phase
| of the corresponding sine

How can F hold both? Complex number trick!
) ( ) ( ) ( e e e iI R F + =
2 2
) ( ) ( e e I R A + =
) (
) (
tan
1
e
e
|
R
I

=
Time and Frequency
example : g(t) = sin(2pi f t) + (1/3)sin(2pi (3f) t)
Time and Frequency
=
+
example : g(t) = sin(2pi f t) + (1/3)sin(2pi (3f) t)
Frequency Spectra
example : g(t) = sin(2pi f t) + (1/3)sin(2pi (3f) t)
=
+
Frequency Spectra
Usually, frequency is more interesting than the phase
=
+
=
Frequency Spectra
=
+
=
Frequency Spectra
=
+
=
Frequency Spectra
=
+
=
Frequency Spectra
=
+
=
Frequency Spectra
=
1
1
sin(2 )
k
A kt
k
t

Frequency Spectra
Frequency Spectra
FT: Just a change of basis
.
.
.
*
=
M * f(x) = F(e)
IFT: Just a change of basis
.
.
.
*
=
M
-1
* F(e) = f(x)
Fourier Transform more formally
Arbitrary function Single Analytic Expression
Spatial Domain (x) Frequency Domain (u)
Represent the signal as an infinite weighted sum
of an infinite number of sinusoids
( ) ( )
}

= dx e x f u F
ux i t 2
(Frequency Spectrum F(u))
1 sin cos = + = i k i k e
ik
Note:
Inverse Fourier Transform (IFT)
( ) ( )
}

= dx e u F x f
ux i t 2
Also, defined as:
( ) ( )
}

= dx e x f u F
iux
1 sin cos = + = i k i k e
ik
Note:
Inverse Fourier Transform (IFT)
( ) ( )
}

= dx e u F x f
iux
t 2
1
Fourier Transform
Fourier Transform Pairs (I)
angular frequency ( )
iux
e

Note that these are derived using

angular frequency ( )
iux
e

Note that these are derived using

Fourier Transform Pairs (I)
Fourier Transform and Convolution
h f g - =
( ) ( )
}

= dx e x g u G
ux i t 2
( ) ( )
} }

= dx d e x h f
ux i
t t t
t 2
( ) | | ( )
( )
| |
} }

= dx e x h d e f
x u i u i t t t t
t t t
2 2
( ) | | ( ) | |
} }

= ' '
' 2 2
dx e x h d e f
ux i u i t t t
t t
Let
Then
( ) ( ) u H u F =
Convolution in spatial domain
Multiplication in frequency domain

Fourier Transform and Convolution

h f g - =
FH G=
fh g =
H F G - =
Spatial Domain (x) Frequency Domain (u)
So, we can find g(x) by Fourier transform
g
= f
- h
G
=
F

H
FT FT IFT
Properties of Fourier Transform
Spatial Domain (x) Frequency Domain (u)
Linearity ( ) ( ) x g c x f c
2 1
+ ( ) ( ) u G c u F c
2 1
+
Scaling ( ) ax f |
.
|

\
|
a
u
F
a
1
Shifting ( )
0
x x f
( ) u F e
ux i
0
2t
Symmetry ( ) x F ( ) u f
Conjugation
( ) x f
-
( ) u F
-
Convolution ( ) ( ) x g x f - ( ) ( ) u G u F
Differentiation
( )
n
n
dx
x f d
( ) ( ) u F u i
n
t 2
frequency ( )
ux i
e
t 2
Note that these are derived using
Properties of Fourier Transform
Example use: Smoothing/Blurring
We want a smoothed function of f(x)
( ) ( ) ( ) x h x f x g - =
H(u) attenuates high frequencies in F(u) (Low-pass Filter)!
Then
( ) ( )
(

=
2
2
2
2
1
exp o tu u H
( ) ( ) ( ) u H u F u G =
to 2
1
u
( ) u H
( )
(

=
2
2
2
1
exp
2
1
o
o t
x
x h
Let us use a Gaussian kernel
o
( ) x h
x
Does not look anything like what we have seen
Magnitude of the FT
Image Processing in the Fourier Domain
Image Processing in the Fourier Domain
Does not look anything like what we have seen
Magnitude of the FT
Convolution is Multiplication in Fourier Domain
*
f(x,y)
h(x,y)
g(x,y)
|F(s
x
,s
y
)|
|H(s
x
,s
y
)|
|G(s
x
,s
y
)|
Low-pass Filtering
Let the low frequencies pass and eliminating the high frequencies.
Generates image with overall
shading, but not much detail
High-pass Filtering
Lets through the high frequencies (the detail), but eliminates the low
frequencies (the overall shape). It acts like an edge enhancer.

Boosting High Frequencies
Most information at low frequencies!
Fun with Fourier Spectra
Next Class
Image resampling and image pyramids

Horn, Chapter 6