0% found this document useful (0 votes)
8 views115 pages

Unit-II Frequency Domain

The document discusses the history of the United States and how it became a world power after World War II. It describes events like the Marshall Plan and how the US rebuilt war-torn Europe and established new global institutions and alliances.

Uploaded by

Nitish Chandra
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
8 views115 pages

Unit-II Frequency Domain

The document discusses the history of the United States and how it became a world power after World War II. It describes events like the Marshall Plan and how the US rebuilt war-torn Europe and established new global institutions and alliances.

Uploaded by

Nitish Chandra
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 115

Digital Image Processing

Chapter 4: Image Enhancement in


the Frequency Domain
Background
 The French mathematiian Jean
Baptiste Joseph Fourier
 Born in 1768
 Published Fourier series in 1822
 Fourier’s ideas were met with
skepticism
 Fourier series
 Any periodical function can be
expressed as the sum of sines and/or
cosines of different frequencies, each
multiplied by a different coefficient
 Fourier transform
 Functions can be expressed as the
integral of sines and/or cosines
multiplied by a weighting function
 Functions expressed in either a Fourier
series or transform can be
reconstructed completely via an inverse
process with no loss of information

 ( x, y )  1
2 2 2 2 ( x 2  y 2 )  ( u 2  v 2 ) / 2 2
A2 e  Ae
 Applications
 Heat diffusion
 Fast Fourier transform (FFT) developed
in the late 1950s
Introduction to the Fourier Transform
and the Frequency Domain

 The one-dimensional Fourier


transform and its inverse
 Fourier transform

F (u )   f ( x )e  j 2 ux
dx

 Inverse Fourier transform

f ( x)   F (u )e j 2 ux
du

 Two variables

Fourier transform
 
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
 
 Discrete Fourier transform (DFT)
 Original variable
f ( x), x  0,1,2,..., M  1
 Transformed variable
F (u ), u  0,1,2,..., M  1
M 1
1
F (u ) 
M
 f ( x )e
x 0
 j 2 ux / M
,

u  0,1,2,..., M  1
M 1
f ( x)   F (u )e j 2 ux / M
,
u 0

x  0,1,2,..., M  1
 DFT
 The discrete Fourier transform and its
inverse always exist
 f(x) is finite in the book
 Sines and cosines
j
e  cos   j sin 
M 1
1
F (u ) 
M
 f ( x)[cos 2 ux / M  j sin 2 ux / M ]
x 0
 Time domain
x
 Time components
f (x)
 Frequency domain
u
 Frequency components
F (u )
 Fourier transform and a glass prism
 Prism
 Separates light into various color
components, each depending on its
wavelength (or frequency) content
 Fourier transform
 Separates a function into various
components, also based on frequency
content
 Mathematical prism
 Polar coordinates
j ( u )
F (u )  F (u ) e
 Real part
R (u )
 Imaginary part
I (u )
 Magnitude or spectrum

 
1
F (u )  R (u )  I (u )
2 2 2

 Phase angle or phase spectrum

 I (u ) 
 (u )  tan 1

 R (u ) 
 Power spectrum or spectral density
2
P (u )  F (u )  R (u )  I (u )
2 2
 Samples

f ( x)  f ( x0  xx)

F (u )  F (uu )
1
u 
Mx
 Some references
 http://local.wasp.uwa.edu.au/~pbourk
e/other/dft/
 http://homepages.inf.ed.ac.uk/rbf/HIP
R2/fourier.htm
 Examples
 test_fft.c
 fft.h
 fft.c
 Fig4.03(a).bmp
 test_fig2.bmp
 The two-dimensional DFT and its
inverse
M 1 N 1
1
F (u , v) 
MN
 f ( x, y)e
x 0 y 0
 j 2 ( ux / M  vy / N )
,

u  0,1,2,..., M  1
v  0,1,2,..., N  1
 Spatial, or image variables: x, y
 Transform, or frequency variables:
u, v
M 1 N 1
f ( x, y )   F (u , v)e j 2 ( ux / M  vy / N )
,
u 0 v 0

x  0,1,2,..., M  1
y  0,1,2,..., N  1
 Magnitude or spectrum

 
1
F (u , v)  R (u , v)  I (u , v)
2 2 2

 Phase angle or phase spectrum


 I (u , v) 
 (u, v)  tan 
1

 R (u , v) 
 Power spectrum or spectral density
2
P (u , v)  F (u , v)  R (u , v)  I (u , v)
2 2
 Centering


 f ( x, y )(1) x y
  F (u  M / 2, v  N / 2)
 Average gray level
 F(0,0) is called the dc component of
the spectrum
M 1 N 1
1
F (0,0) 
MN
 f ( x, y)
x 0 y 0
 Conjugate symmetric
 If f(x,y) is real

F (u , v)  F * (u ,v)
F (u , v)  F (u ,v)
 Relationships between samples in
the spatial and frequency domains
1 1
u  v 
Mx Ny
 The separation of spectrum zeros in the
u-direction is exactly twice the
separation of zeros in the v direction
 Filtering in
the
frequency
domain
 Strong edges that run approximately at
+45 degree, and -45 degree
 The inclination off horizontal of the long
white element is related to a vertical
component that is off-axis slightly to
the left
 The zeros in the vertical frequency
component correspond to the narrow
vertical span of the oxide protrusions
 Basics of filtering in the frequency
domain
1. Multiply the input image by ( 1)
 x y

to center the transform


 2. Compute F(u,v)
 3. Multiply F(u,v) by a filter function
H(u,v)
 4. Compute the inverse DFT
 5. Obtain the real part
6. Multiply the result by ( 1)
x y

 Fourier transform of the output
image
G (u , v)  H (u , v) F (u , v)
 zero-phase-shift filter
 Real H(u,v)

 I (u , v) 
 (u, v)  tan 1

 R (u , v) 
 Inverse Fourier transform of G(u,v)

 G (u , v)
1

 The imaginary components of the


inverse transform should all be zero
 When the input image and the filter
function are real
 Set F(0,0) to be zero, a notch filter
0 if (u , v)  ( M / 2, N / 2)
H (u , v)  
1 otherwise
 Lowpass filter
 Pass low frequencies, attenuate high
frequencies
 Blurring
 Highpass filter
 Pass high frequencies, attenuate low
frequencies
 Edges, noise
 Convolution theorem

f ( x, y ) * h ( x, y ) 
M 1 N 1
1
MN
 f (m, n)h( x  m, y  n)
m 0 n 0

f ( x, y ) * h( x, y )  F (u , v) H (u , v)
f ( x, y )h( x, y )  F (u , v) * H (u , v)
 Impulse function of strength A
A ( x  x0 , y  y0 )
M 1 N 1

 s( x, y) A ( x  x , y  y )  As( x , y )
x 0 y 0
0 0 0 0

M 1 N 1

 s( x, y) A ( x, y)  As(0,0)
x 0 y 0
M 1 N 1
1
F (u , v) 
MN
  ( x, y)e
x 0 y 0
 j 2 ( u x / M  v y / N )

1

MN
f ( x, y ) * h ( x, y ) 
M 1 N 1
1

MN m 0 n 0
 (m, n)h( x  m, y  n)

1
 h ( x, y )
MN
f ( x, y ) * h( x, y )  F (u , v) H (u , v)
 ( x, y ) * h( x, y )   ( x, y )H (u , v)
h( x, y )  H (u , v)
 Gaussian filter
 u 2 / 2 2
H (u )  Ae
2 2 2 x 2
h( x)  2 Ae
 Highpass filter
 u 2 / 2 12  u 2 / 2 2 2
H (u )  Ae  Be
2 2 12 x 2 2 2 2 2 x 2
h( x)  2  1 Ae  2  2 Be

A  B and  1   2
Smoothing Frequency-Domain Filterers

G (u , v)  H (u , v) F (u , v)
 Ideal lowpass filters

1 if D(u, v)  D 0
H (u , v)  
0 if D(u, v)  D 0

D(u , v)  (u  M / 2)  (v  N / 2)
2

2 1/ 2
 Cutoff frequency

D0
 Total image power
M 1 N 1
PT   P (u , v)
u 0 v 0
 Portion of the total power
 
  100 P(u, v) / PT 
u v 
 Blurring and ringing properties
 Filter

G (u , v)  H (u , v) F (u , v)
 Convolution

g ( x, y )  h ( x, y ) * f ( x, y )
 h ( x, y ) : Spatial filter
 H (u , v) was multiplied by (1) u v
 Then the inverse DFT
 The real part of the inverse DFT was
multiplied by ( 1) x y
 The filter h( x, y )
 A dominant component at the origin
 Concentric, circular components about
the center component --- ringing
 The radius of the center component
and the number of circles per unit
distance from the origin are inversely
proportional to the value of the cutoff
frequency of the ideal filter.
 Butterworth lowpass filters
1
H (u , v) 
1  D(u , v) / D0 
2n


D(u , v)  (u  M / 2)  (v  N / 2)
2

2 1/ 2

 H (u , v)  0.5 when D(u, v)  D0


 Butterworth lowpass filters
 Order 1: No ringing
 Order 2: Imperceptible ringing
 Higher order: Ringing becomes a
significant factor
 Gaussain lowpass filters
 D 2 ( u ,v ) / 2 2
H (u , v)  e

D(u , v)  (u  M / 2)  (v  N / 2) 2

2 1/ 2

 When   D 0
 D 2 ( u ,v ) / 2 D0 2
H (u , v)  e
 No ringing
 Additional examples of lowpass
filters
 Machine perception
 Printing and publishing
 Satellite and aerial images
Sharpening Frequency Domain Filters

 Highpass filter
H hp (u , v)  1  H lp (u , v)
 Spatial filter: h ( x, y )
 H (u , v) was multiplied by (1) u v
 Then the inverse DFT
 The real part of the inverse DFT was

multiplied by ( 1) x y
 Ideal highpass filters

0 if D(u, v)  D 0
H (u , v)  
1 if D(u, v)  D 0

D(u , v)  (u  M / 2)  (v  N / 2)
2

2 1/ 2
 Butterworth highpass filters
1
H (u , v) 
1  D0 / D(u , v)
2n


D(u , v)  (u  M / 2)  (v  N / 2)
2

2 1/ 2
 Gaussian highpass filters
 D 2 ( u ,v ) / 2 D0 2
H (u , v)  1  e

D(u , v)  (u  M / 2)  (v  N / 2)
2

2 1/ 2
 The Laplacian in the frequency
domain
 d n f ( x) 
 n   ( ju ) n
F (u )
 dx 
  2 f ( x, y )  2 f ( x, y ) 
  
  x 2
 y 2

 ( ju ) 2 F (u , v)  ( jv ) 2 F (u , v)
 (u 2  v 2 ) F (u , v)
 2
f ( x, y )  (u 2
 v ) F (u , v)
2

H (u , v)  (u 2  v 2 )

 After centering


H (u , v)   (u  M / 2)  (v  N / 2) 2 2

 Inverse Fourier transform
 f ( x, y ) 
2

1

  (u  M / 2)  (v  N / 2) F (u , v)
2 2
 
 Fourier-transform pair

 f ( x, y ) 
2

 
 (u  M / 2)  (v  N / 2) F (u , v)
2 2
 Subtracting the Laplacian from the
original image

g ( x, y )  f ( x, y )   2 f ( x, y )

g ( x, y ) 
 
1 1  (u  M / 2) 2  (v  N / 2) 2 F (u , v) 
 Unsharp masking, high-boost
filtering, and high-frequency
emphasis filtering
 Highpass filtering

f hp ( x, y )  f ( x, y )  f lp ( x, y )
 High-boost filtering

f hb ( x, y )  Af ( x, y )  f lp ( x, y )
f hb ( x, y )  ( A  1) f ( x, y )  f ( x, y )  f lp ( x, y )
f hb ( x, y )  ( A  1) f ( x, y )  f hp ( x, y )

 Frequency domain

H hb (u , v)  1  H lp (u , v)

H hb (u , v)  ( A  1)  H hp (u , v)
 High-frequency emphasis

H hfe (u , v)  a  bH hp (u , v)
 where a  0 and b  a
Homomorphic Filtering
 Illumination and reflectance
components
f ( x, y )  i ( x, y ) r ( x, y )
 Derivations
 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 ( x, y )  ln f ( x, y )
 ln i ( x, y ) ln r ( x, y )
 Or

Z (u , v)  Fi (u , v)  Fr (u , v)
 Frequency domain
S (u , v)  H (u , v) Z (u , v)
 H (u , v) Fi (u , v)  H (u , v) Fr (u , v)
 Spatial domain
s ( x, y )   S (u, v)
1

  H (u , v) Fi (u , v)
1

 H (u , v) Fr (u , v)
1
i ' ( x, y )   H (u, v) Fi (u, v)
1

r ' ( x, y )   H (u , v) Fr (u , v)
1

s ( x, y )  i ' ( x, y )  r ' ( x, y )
g ( x, y )  e e
s ( x, y )
e
i '( x , y ) r '( x , y )

 i0 ( x, y )r0 ( x, y )
 Decrease the contribution made by the
low frequencies (illumination)
 Amplify the contribution made by high
frequencies (reflectance)
 Simultaneous dynamic range
compression and contrast enhancement

H (u , v)  ( H   L ) 1  e  c ( D 2 ( u ,v ) / D0 2 )
  L
Implementation
 Translation
j 2 ( u 0 x / M  v0 y / N )
f ( x, y )e  F (u  u0 , v  v0 )
 j 2 ( ux0 / M  vy0 / N )
f ( x  x0 , y  y0 )  F (u , v)e

 When u0  M / 2 and v0  N / 2
j 2 ( u 0 x / M  v0 y / N ) j ( x  y )
e e
x y
 (1)
x y
f ( x, y )(1)  F (u  M / 2, v  N / 2)
u v
f ( x  M / 2, y  N / 2)  F (u , v)(1)
 Distributivity and scaling

 f1 ( x, y )  f 2 ( x, y )   f1 ( x, y )  f 2 ( x, y )

af ( x, y )  aF (u , v)
1
f (ax, by )  F (u / a, v / b)
ab
 Rotation
 Polar coordinates
x  r cos  y  r sin 
u  w cos  v  w sin 
 Rotating f ( x, y ) by an angle  0
rotates F (u , v ) by the same angle

f (r ,  )  F ( w,  )
f (r ,    0 )  F ( w,    0 )
 Periodicity and conjugate symmetry
 Periodicity property

F (u , v)  F (u  M , v) 
F (u , v  N )  F (u  M , v  N )
f ( x, y )  f ( x  M , y ) 
f ( x, y  N )  f ( x  M , y  N )
 Conjugate symmetry

F (u , v)  F (u ,v)
*

 Symmetry of the spectrum

F (u , v)  F (u ,v)
F (u , v)  F (u , v)
F (u , v)  F (u ,v)
 Separability
M 1 N 1
1
F (u , v) 
MN
 f ( x, y)e
x 0 y 0
 j 2 ( ux / M  vy / N )

M 1 N 1
1 1

M
e
x 0
 j 2 ( ux / M )

N
 f ( x, y )e
y 0
 j 2 ( vy / N )

M 1
1

M
 f ( x, v )e
x 0
 j 2 ( ux / M )
 where
N 1
1
F ( x, v ) 
N
 f ( x, y )e
y 0
 j 2 ( vy / N )

 We can compute the 2-D transform by


first computing a 1-D transform along
each row of the input image, and then
computing a 1-D transform along each
column of this intermediate result
 Computing the inverse Fourier
transform using a forward transform
algorithm
M 1
1
F (u ) 
M
 f ( x )e
x 0
 j 2 ux / M
,

u  0,1,2,..., M  1
M 1
f ( x)   F (u )e j 2 ux / M ,
u 0

x  0,1,2,..., M  1
 Calculate

M 1
1 * 1
M
f ( x) 
M
F
u 0
*
(u )e  j 2 ux / M
 Inputting F * (u ) into an algorithm
designed to compute the forward
transform gives the quantity
1 *
f ( x)
M
 2-D
1
f ( x, y ) 
*

MN
M 1 N 1
1

MN u 0 v 0
*
F (u , v)e  j 2 ( ux / M  vy / N )
 More on periodicity: the need for
padding
 Convolution: Flip one of the functions
and slide it pass the other
M 1
1
f ( x )  h( x ) 
M
 f ( m) h( x  m)
m 0

You might also like