Professional Documents
Culture Documents
Lecture 3. Filtering in The Frequency Domain
Lecture 3. Filtering in The Frequency Domain
Spring 2008
New Mexico Tech
Outline
► Fourier Transform
03/21/22 2
Fourier Series and Fourier Transform: History
► Jean Baptiste Joseph Fourier, French mathematician and physicist
(03/21/1768-05/16/1830) http://en.wikipedia.org/wiki/Joseph_Fourier
Permanent
Orphaned: at nine
Secretary of the
French Academy of
Egyptian expedition Sciences: 1822
with Napoleon I:
1798 Théorie analytique
Governor of Lower de la chaleur :
Egypt 1822
(The Analytic
Theory of Heat)
03/21/22 3
Fourier Series and Fourier Transform: History
► Fourier Series
Any periodic function can be expressed as the sum of
sines and /or cosines of different frequencies, each
multiplied by a different coefficients
► Fourier Transform
Any function that is not periodic can be expressed as the
integral of sines and /or cosines multiplied by a weighing
function
03/21/22 4
Fourier Series: Example
03/21/22 5
Preliminary Concepts
| C | R 2 I 2 and arctan( I / R)
C | C | (cos j sin )
Using Euler's formula,
C | C | e j
03/21/22 6
Fourier Series
where
2 n
1 T /2 j t
cn f (t )e T
dt for n 0, 1, 2,...
T T / 2
03/21/22 7
Impulses and the Sifting Property (1)
A unit impulse of a continuous variable t located
at t =0, denoted (t ), defined as
if t 0
(t )
0 if t 0
and is constrained also to satisfy the identity
(t )dt 1
( x) 1
x
The sifting property
x
f ( x) ( x x0 ) f ( x0 )
x
f ( x) ( x) f (0)
03/21/22 9
Impulses and the Sifting Property (3)
03/21/22 10
Fourier Transform: One Continuous Variable
03/21/22 11
Fourier Transform: One Continuous Variable
W /2
F ( ) f (t )e j 2 t
dt Ae j 2t dt
W /2
A j 2 t W /2 A
e e jW e jW
j 2 W /2 j 2 W
sin(W )
AW
(W )
03/21/22 12
Fourier Transform: Impulses
e j 2 0
=1
The Fourier transform of a unit impulse located at t t0 :
F ( ) (t t0 )e j 2t dt
e j 2t0
=cos(2 t0 ) j sin (2 t0 )
03/21/22 13
Fourier Transform: Impulse Trains
Impulse train sT (t ), sT (t ) (t nT )
n
where
2 n
1 T / 2 j t
cn
T T / 2
sT (t )e T
dt
03/21/22 14
Fourier Transform: Impulse Trains
1 0 1
= e n
Tj 2 ( T) t n
e T
dt ( )
T
2 n n 2 n n
( j T t ) 1 ( j t )e j 2t du
1
sT (t ) c n e T e T T
n 2 nt T n
j
e T
03/21/22 15
Fourier Transform: Impulse Trains
Let S ( ) denote the Fourier transform of the
periodic impulse train S T (t )
2 n
1 j t
S ( ) S T (t ) e T
T n
1 j 2Tn t
e
T n
1 n
=
T n
(
T
)
03/21/22 16
Fourier Transform and Convolution
The convolution of two functions is denoted
by the operator
f (t ) h (t ) f ( ) h(t ) d
f (t ) h(t ) f ( )h(t )d e j 2t dt
= f ( ) h(t )e j 2t dt d
= f ( ) H ( )e j 2 d
=H ( ) f ( )e j 2 d
=H ( ) F ( )
03/21/22 17
Fourier Transform and Convolution
f (t ) h(t ) H ( ) F ( )
f (t )h(t ) H ( ) F ( )
03/21/22 18
Fourier Transform of Sampled Functions
f (t ) f (t ) sT (t )
n
f (t ) (t nT )
03/21/22 19
Fourier Transform of Sampled Functions
F ( ) f (t ) f (t ) sT (t ) F ( ) ? S ( )
F ( ) F1( ) S ( ) n F ( ) S ( ) d
S ( )
1T n
(
T
)
n
= F ( ) ( )d
T
n T
1 n
=
T n
F ( ) (
T
)d
1 n
=
T n
F (
T
)
03/21/22 20
Question
1. Continuous
2. Discrete
03/21/22 21
Fourier Transform of Sampled Functions
► A bandlimited signal is a signal whose Fourier transform
is zero above a certain finite frequency. In other words, if
the Fourier transform has finite support then the signal is
said to be bandlimited.
x(t ) sin(2 ft )
03/21/22 22
Fourier Transform of Sampled Functions
max max
Over-sampling
1
2 max
F ( ) T
1
n Critically-sampling
T
F (
T
)
n 1
2max
T
under-sampling
1
2 max
03/21/22 T 23
Nyquist–Shannon sampling theorem
1
► Sufficient separation is guaranteed if 2max
T
Sampling theorem: A continuous, band-limited function
can be recovered completely from a set of its samples if
the samples are acquired at a rate exceeding twice the
highest frequency content of the function
03/21/22 24
Nyquist–Shannon sampling theorem
?
f (t ) F ( )e j 2t d
03/21/22 25
Aliasing
03/21/22 26
Aliasing
03/21/22 27
Aliasing
03/21/22 28
Function Reconstruction from Sampled Data
F ( ) H ( ) F ( )
f (t ) 1
F ( )
1
H ( ) F ( )
h(t ) f (t )
f (t ) f (nT )sinc (t nT ) / nT
n
03/21/22 29
The Discrete Fourier Transform (DFT) of One
Variable
M 1
F ( ) f ( x)e j 2 x / M , 0,1,..., M 1
x 0
1 M 1
f ( x)
M
F
0
( ) e j 2 x / M
, x 0,1, 2,..., M 1
03/21/22 30
2-D Impulse and Sifting Property: Continuous
if t z 0
The impulse (t , z ), (t , z )
0 otherwise
and
(t , z )dtdz 1
03/21/22 31
2-D Impulse and Sifting Property: Discrete
1 if x y 0
The impulse ( x, y ), ( x, y )
0 otherwise
x y
f ( x, y ) ( x, y ) f (0, 0)
and
x y
f ( x, y ) ( x x0 , y y0 ) f ( x0 , y0 )
03/21/22 32
2-D Fourier Transform: Continuous
F ( , ) f (t , z ) e j 2 ( t z )
dtdz
and
f (t , z ) f ( , )e j 2 ( t z )
d d
03/21/22 33
2-D Fourier Transform: Continuous
F ( , ) f (t , z )e j 2 ( t z ) dtdz
T /2 Z /2
Ae j 2 ( t z ) dtdz
T /2 Z /2
sin(T ) sin( T )
ATZ T
T
03/21/22 34
2-D Sampling and 2-D Sampling Theorem
2 D impulse train:
sT Z (t , z ) (t mT , z nZ )
m n
03/21/22 35
2-D Sampling and 2-D Sampling Theorem
03/21/22 37
Aliasing in Images: Example
In an image system, the
number of samples is fixed at
96x96 pixels. If we use this
system to digitize checkerboard
patterns …
Under-sampling
03/21/22 38
Aliasing in Images: Example
Re-sampling
03/21/22 39
Aliasing in Images: Example
Re-sampling
03/21/22 40
Moiré patterns
http://en.wikipedia.org/wiki/Moiré_pattern
03/21/22 41
Moiré patterns
03/21/22 42
Moiré patterns
03/21/22 43
Moiré patterns
A moiré pattern
formed by
incorrectly down-
sampling the
former image
03/21/22 44
2-D Discrete Fourier Transform and Its
Inverse
DFT:
M 1 N 1 j 2 ( x / M y / N )
F ( , ) f ( x, y )e
x 0 y 0
IDFT:
j 2 ( x / M y / N )
1 M 1 N 1
f ( x, y )
MN
F ( , )e
x 0 y 0
03/21/22 45
Properties of the 2-D DFT
relationships between spatial and frequency intervals
03/21/22 46
Properties of the 2-D DFT
translation and rotation
f ( x, y )e j 2 ( 0 x / M 0 y / N ) F ( 0 , 0 )
and
j 2 ( x0 / M y0 / N )
f ( x - x0 , y - y0 ) F ( , )e
f ( x, y ) f ( x k1M , y ) f ( x, y k 2 N ) f ( x k1M , y k 2 N )
f ( x ) e j 2 ( 0 x / M ) F ( 0 )
0 M / 2, f ( x)(1) x F ( M / 2)
f ( x, y )(1) x y F ( M / 2, N / 2)
03/21/22 48
Properties of the 2-D DFT
periodicity
03/21/22 49
Properties of the 2-D DFT
Symmetry
03/21/22 50
Properties of the 2-D DFT
Fourier Spectrum and Phase Angle
03/21/22 51
03/21/22 52
03/21/22 53
Example: Phase Angles
03/21/22 54
Example: Phase Angles and The Reconstructed
03/21/22 55
2-D Convolution Theorem
1-D convolution
M 1
f ( x ) h( x ) f ( m) h( x m)
m 0
2-D convolution
M 1 N 1
f ( x, y ) h ( x, y ) f (m, n)h( x m, y n)
m 0 n 0
Mirroring h
about the
origin
Translating
the mirrored
function by x
Computing
the sum for
each x
03/21/22 57
An Example of Convolution
It causes the
wraparoun
d error
It can be
solved by
appending
zeros
03/21/22 58
Zero Padding
P ≥A+B-1
03/21/22 59
Zero Padding
► Let f(x,y) and h(x,y) be two image arrays of sizes A×B and
C×D pixels, respectively. Wraparound error in their
convolution can be avoided by padding these functions
with zeros
f ( x, y ) 0 x A -1 and 0 y B -1
f p ( x, y )
0 A x P or B y Q
h ( x, y ) 0 x C -1 and 0 y D -1
h p ( x, y )
0 C x P or D y Q
Here P A C 1; Q B D 1
03/21/22 60
Summary
03/21/22 61
Summary
03/21/22 62
Summary
03/21/22 63
Summary
03/21/22 64
The Basic Filtering in the Frequency Domain
03/21/22 65
The Basic Filtering in the Frequency Domain
03/21/22 66
The Basic Filtering in the Frequency Domain
03/21/22 67
The Basic Filtering in the Frequency Domain
03/21/22 68
The Basic Filtering in the Frequency Domain
03/21/22 69
Zero-Phase-Shift Filters
1
g ( x, y ) {H (u , v) F (u , v)}
F (u , v) R (u , v) jI (u , v )
1
g ( x, y ) H (u, v) R(u, v) jH (u, v) I (u, v)
Filters affect the real and imaginary parts equally,
and thus no effect on the phase.
These filters are called zero-phase-shift filters
03/21/22 70
Examples: Nonzero-Phase-Shift Filters
Even small
Phasechanges
angle is in the phase angle can ishave
Phase angle
dramaticmultiplied
(usually by undesirable) effects on the
multiplied by filtered
0.5 0.5
output
03/21/22 71
Summary:
Steps for Filtering in the Frequency Domain
1. Given an input image f(x,y) of size MxN, obtain the
padding parameters P and Q. Typically, P = 2M and Q = 2N.
g p ( x, y ) real 1 G (u , v) ( 1) x y
03/21/22 73
An Example:
Steps for Filtering in the Frequency Domain
03/21/22 74
Correspondence Between Filtering in the
Spatial and Frequency Domains (1)
03/21/22 75
Correspondence Between Filtering in the
Spatial and Frequency Domains (2)
Let H (u ) denote the difference of Gaussian filter
- u 2 /212 - u 2 /2 22
H (u) Ae Be
with A B and 1 2
03/21/22 76
Correspondence Between Filtering in the
Spatial and Frequency Domains (3)
03/21/22 77
Correspondence Between Filtering in the
Spatial and Frequency Domains: Example
600x600
03/21/22 78
Correspondence Between Filtering in the
Spatial and Frequency Domains: Example
03/21/22 79
Generate H(u,v)
h ( x, y ) 0 x 2 and 0 y 2
h p ( x, y )
0 3 x 602 or 3 y 602
03/21/22 80
Generate H(u,v)
03/21/22 81
Image Smoothing Using Filter Domain Filters:
ILPF
03/21/22 82
Image Smoothing Using Filter Domain Filters:
ILPF
03/21/22 83
ILPF Filtering Example
03/21/22 84
ILPF
Filtering
Example
03/21/22 85
The Spatial Representation of ILPF
03/21/22 86
Image Smoothing Using Filter Domain Filters:
BLPF
Butterworth Lowpass Filters (BLPF) of order n and
with cutoff frequency D0
1
H (u , v)
1 D(u, v) / D0
2n
03/21/22 87
03/21/22 88
The Spatial Representation of BLPF
03/21/22 89
Image Smoothing Using Filter Domain Filters:
GLPF
By letting D0
D 2 ( u , v )/2 D0 2
H (u, v) e
03/21/22 90
Image Smoothing Using Filter Domain Filters:
GLPF
03/21/22 91
03/21/22 92
Examples of smoothing by GLPF (1)
03/21/22 93
Examples of smoothing by GLPF (2)
03/21/22 94
Examples of smoothing by GLPF (3)
03/21/22 95
Image Sharpening Using Frequency Domain
Filters
H HP (u , v) 1 H LP (u , v)
03/21/22 96
Image Sharpening Using Frequency Domain
Filters
03/21/22 97
03/21/22 98
The Spatial Representation of Highpass
Filters
03/21/22 99
Filtering Results by IHPF
03/21/22 100
Filtering Results by BHPF
03/21/22 101
Filtering Results by GHPF
03/21/22 102
Using Highpass Filtering and Threshold for
Image Enhancement
BHPF
(order 4 with a cutoff
frequency 50)
03/21/22 103
The Laplacian in the Frequency Domain
H (u, v) 4 2 (u 2 v 2 )
H (u , v) 4 2 (u P / 2) 2 (v Q / 2) 2 )
4 2 D 2 (u, v)
The Laplacian image
2 f ( x, y ) 1 H (u , v) F (u, v)
Enhancement is obtained
g ( x, y ) f ( x, y ) c 2 f ( x, y ) c -1
03/21/22 104
The Laplacian in the Frequency Domain
1 1 4 2 D 2 (u, v) F (u, v)
03/21/22 105
The Laplacian in the Frequency Domain
03/21/22 106
Unsharp Masking, Highboost Filtering and
High-Frequency-Emphasis Fitering
g mask ( x, y ) f ( x, y ) f LP ( x, y )
g ( x, y ) 1 1 k * 1 H LP (u, v) F (u , v)
1 1 k * H HP (u, v) F (u, v)
03/21/22 107
Unsharp Masking, Highboost Filtering and
High-Frequency-Emphasis Fitering
1
g ( x, y ) k
1 k2 * H HP (u , v) F (u , v)
k1 0 and k2 0
03/21/22 108
Gaussian Filter
D0=40
High-Frequency-Emphasis Filtering
Gaussian Filter
K1=0.5, k2=0.75
03/21/22 109
Homomorphic Filtering
f ( x , y ) i ( x , y ) r ( x, y )
f ( x, y )
= i ( x, y ) r ( x, y ) ?
z ( x, y ) ln f ( x, y ) ln i ( x, y ) ln r ( x, y )
Z (u, v) Fi (u, v) Fr (u , v )
03/21/22 110
Homomorphic Filtering
S (u , v) H (u, v) Z (u, v)
H (u, v) Fi (u, v) H (u, v) Fr (u, v)
s ( x, y ) 1 S (u, v)
1 H (u, v) Fi (u, v) H (u, v) Fr (u, v)
1 H (u, v) Fi (u, v) 1 H (u, v) Fr (u, v)
i '( x, y) r '( x, y )
03/21/22 111
Homomorphic Filtering
H (u , v) ( H L ) 1 e
c D 2 ( u ,v )/ D02
L
03/21/22 113
L 0.25
Homomorphic
H 2
Filtering
c 1
D0 80
03/21/22 114
Homomorphic Filtering
03/21/22 115
Selective Filtering
Non-Selective Filters:
operate over the entire frequency rectangle
Selective Filters
operate over some part, not entire frequency rectangle
• bandreject or bandpass: process specific bands
• notch filters: process small regions of the frequency
rectangle
03/21/22 116
Selective Filtering:
Bandreject and Bandpass Filters
H BP (u , v) 1 H BR (u , v)
03/21/22 117
Selective Filtering:
Bandreject and Bandpass Filters
03/21/22 118
Selective Filtering:
Notch Filters
Zero-phase-shift filters must be symmetric about the origin.
A notch with center at (u0, v0) must have a corresponding
notch at location (-u0,-v0).
where H k (u, v) and H - k (u, v) are highpass filters whose centers are
at (uk , vk ) and (-uk , -vk ), respectively.
03/21/22 119
Selective Filtering:
Notch Filters
Q
H NR (u, v) H k (u, v) H k (u, v)
k 1
where H k (u, v) and H -k (u, v) are highpass filters whose centers are
at (uk , vk ) and (-uk , -vk ), respectively.
A Butterworth notch
reject filter D 0 =3
and n=4 for all
notch pairs
03/21/22 121
Examples:
Notch Filters
(2)
03/21/22 122
03/21/22 123