12/25/2023

Contents
Lecture No. 10
1. Background
Image Enhancement 2. Introduction to the Fourier Transform and
the Frequency Domain

in the Frequency Domain 1. Smoothing Frequency-Domain Filters

2. Sharpening Frequency Domain Filters

CIS-322

Background Background
• Fourier series
: any periodic function can be expressed as the sum
of sines and/or cosines of different frequencies,
each multiplied by a different coefficient.
• Fourier transform
: aperiodic even function can be expressed as the Sum of these functions
integral of sines and/or cosines multiplied by a
weighing function.

Introduction to the Fourier Transform The One-Dimensional Fourier

and the Frequency Domain Transform and its Inverse
1. The Fourier transform in one and two  The continuous case:
dimensions • Fourier transform,F (u ),of a single variable f (x).

F (u )   f ( x)e  j 2 uxdx

• Inverse Fourier transform
1. A discrete formulation of the continuous 
f ( x)   F (u )e j 2 uxdu
transform and some of its properties 
• Fourier transform of two variables
 
F (u , v)    f ( x, y )e  j 2 ( ux vy ) dxdy
 
• Inverse Fourier transform
 
f ( x, y )    F (u, v)e j 2 ( ux vy ) dudv
 

1
 12/25/2023

The One-Dimensional Fourier The One-Dimensional Fourier

Transform and its Inverse Transform and its Inverse
 The discrete case: • F (u )in polar coordinate
• Discrete Fourier Transform F (u )  F (u ) e  j (u )
M 1
1
F (u ) 
M
 f ( x)e j 2ux/ M
x 0
for u  0,1,, M  1 where F (u )  [ R 2 (u )  I 2 (u )]1/ 2 : the magnitude or spectrum

• Inverse transform of the Fourier tr ansform

M 1  I (u ) 
f ( x)   F (u )e  (u )  tan 
j 2ux / M 1
for x  0,1,, M  1 e j  cos   j sin   : the phase angle or phase spectrum
u 0
 R(u ) 
• The concept of frequency domain of the Fourier tr ansform
P(u )  F (u )  R 2 (u )  I 2 (u ) : power spectrum
2
M 1
1
F (u ) 
M
 f ( x)[cos 2ux / M  j sin 2ux / M ]
x 0
(square of the Fourier spectrum)

Two dimension Fourier Two dimension Fourier Transform

Transform and its inverse and its inverse
• Fourier transform pair for a function f(x,y) of • Fourier Spectrum
two variables:


F (u, v)  R 2 (u, v)  I 2 (u, v) 
1/ 2

{ f ( x, y )}  F (u, v)    f ( x, y) exp[  j 2 (ux  vy)]dxdy

 • Phase
and  I (u, v) 
   tan 1  
 {F (u, v)}  f ( x, y )    F (u, v) exp[ j 2 (ux  vy)]dudv
1  R(u, v) 
 • Power Spectrum
where u,v are the frequency variables.
P(u , v)  F (u , v)  R 2 (u , v)  I 2 (u , v)
2

A 2D Fourier Transform: a square function

Two dimension Fourier Transform
f(x,y)
Consider a square function in the xy plane: and its inverse
f(x,y) = rect(x) rect(y)
y F {f(x,y)}
The 2D Fourier Transform splits into the x
product of two 1D Fourier Transforms:
f(x,y)
F (2){f(x,y)}
F {f(x,y)} = sinc(kx/2) sinc(ky/2)

y
This picture is an optical determination x
of the Fourier Transform of the
square function!

2
 12/25/2023

Discrete Fourier Transform Discrete Fourier Transform

• A continuous function f(x) is discretized into
• Where x assumes the discrete values
a sequence:
(0,1,2,3,…,M-1) then
{ f ( x0 ), f ( x0  x), f ( x0  2x),..., f ( x0  [ N  1]x)}
f ( x)  f ( x0  xx)
by taking N or M samples x units apart.
• The sequence {f(0),f(1),f(2),…f(M-1)}
denotes any M uniformly spaced samples
from a corresponding continuous function.

Discrete Fourier Transform Discrete Fourier Transform

• The discrete Fourier transform pair that • To compute F(u) we substitute u=0 in the exponential
term and sum for all values of x
applies to sampled functions is given by:
• We repeat for all M values of u
1
M 1 • It takes M*M summations and multiplications
F(u) 
M
 f (x)exp[ j2ux / M] For u=0,1,2,…,M-1
x 0
M 1
and 1
F(u) 
M
 f (x)exp[ j2ux / M] For u=0,1,2,…,M-1
M 1 x 0
 f (x)   f (u)exp[ j2ux / M] For x=0,1,2,…,M-1
u 0 • The Fourier transform and its inverse always exist!




The Two-Dimensional DFT and

Discrete Fourier Transform
its Inverse
• The values u = 0, 1, 2, …, M-1 correspond to • The discrete Fourier transform of a function(image) f ( x, y)
of size M*N
samples of the continuous transform at
1 M 1 N 1
values 0, u, 2u, …, (M-1)u. F (u, v)   f ( x, y)e j 2 (ux/ M vy / N )
MN x0 y 0
• The inverse transform
• i.e. F(u) represents F(uu), where: M 1 N 1
f ( x, y )   F (u, v)e j 2 (ux / M  vy / N )
1 u 0 v 0
u  • In the polar coordinate,
Mx same expression as in the previous section



3
 12/25/2023

The Two-Dimensional DFT and Example of Computation of DFT

its Inverse • x[n] is defined as:
x[0]=1, x[1]=2, x[2]=2, x[3]=1, x[n]=0 otherwise
• The value of the transform at (u, v)  (0,0) so N=4 points
N 1 2kn
1 M 1 N 1 • from the definition j
X k   x[n]e N k  0,1,..., N  1
F (0,0)   f ( x, y) : the average of f ( x, y)
MN x0 y 0 • The 4-point DFT follows n  0
2kn
• If f ( x, y ) is real, its Fourier transform is conjugate 3
X k   x[n]e
j
4
k  0,1,...,3
• that is
symmetric n 0
jk j 3 k
F (u, v)  F * (u,v) X k  x[0]  x[1]e

2
 x[2]e  jk  x[3]e

2
k  0,...,3
jk j 3 k
From this, the spectrum of the Fourier transform is 
 2e  jk  1e


• eventuallyX k  1  2e k  0,...,3
2 2

symmetric
 6e j 0 k 0  6 k 0
 j 5
(1  j ) k  1
F (u, v)  F (u,v) 
Xk  
2e 4 k 1 
or Xk 
 0e j 0 k 2  0 k 2
j 3
 k 3 
(1  j ) k  3

4
2e

The Two-Dimensional DFT and

its Inverse