Fourier Transformation( 1-D Continuous
Signal)
Fourier Transformation:
Let f(x) is a continuous function of some variable then the Fourier transformation of
f(x) is F(u)
Here f(x) must be continuous & integralable
=F(u)
Inverse Fourier Transformation:
F(u) is a Fourier transform of signal f(x) so after inverse Fourier transformation of
F(u) we get f(x)
=F(x)
Fourier Transformation ( 1-D Continuous
Signal)
Fourier Transformation Pair
F(u) → Fourier Transform of signal f(x)
F(x) → Original Signal or Inverse Fourier Transform of F(u)
• Here F(u) is a complex function contains real part & imaginary part
F(u) = R(u) + jI(u)
|F(u)|
We have
Fourier Spectrum:
The phase angle:
Power Spectrum :
Fourier Transformation ( 2-D Continuous
Signal)
• Forward Fourier Transformation:
Let f(x,y) is 2 dimensional signal with 2 variable
• Inverse (Backward) Fourier Transformation:
Fourier Transformation ( 2-D Continuous
Signal)
• Fourier Spectrum:
• Phase angle :
• Power Spectrum:
2-D Discrete Fourier Transformation Forw
ard
• 2D discrete Fourier Transformation:
• Let we have an Image of size MxN then F(u,v) is the F T of image
f(x,y)
Where variable u = 0, 1, 2, …., M-1 and v = 0, 1, 2, …., N-1
• Inverse (Backward) Fourier Transformation :
Where variable x = 0, 1, 2, …., M-1 and y = 0, 1, 2, …., N-1
2-D Discrete Fourier Transformation
• For a square image i.e. M = N and the Fourier Transformation Pair is
as follows
Discrete F T Result
Original Image Transformed Image
Properties of Fourier Transformation
• Seperability
• Translation
• Periodicity
• Conjugate
• Rotation
• Distributive
• Scaling
• Convolution
• Corelation
Seperability
• The separbility property says that we can do 2D Fourier
transformation as two 1 D Fourier Transformation
• Inverse Fourier Transform
X represent row
of image so x is
fixed
Fourier
Transformation
along row
Seperability Cont…
• 2D Inverse Fourier transformation can also be viewed as two 1 D Inverse
Fourier Transformation
IDFT along rows
IDFT along columns
• Advantage of Seperability:
• Operation become much simpler and less time complexity
Seperability Concept
Translation
Translation of x and y by x0 and y0 respectively.
• Fourier Transform
Magnitude of Additional Phase
FT
remains same
Translation Cont..
• Inverse Fourier Transform
• Here sift x0, y0 does not change Fourier spectrum but it add some
phase sift diff
Periodicity
• Periodicity property says that the Discrete Fourier Transform and
Inverse Discrete Fourier Transform are periodic with a period N
Proof:
So we can say that Discrete Fourier Transform is periodic
with N
Conjugate
• If f(x,y) is a real valued function then
F(u,v) = F* (-u, -v)
• Where F* indicate it complex conjugate
• Now Fourier Spectrum
|F(u,v)| = |F(-u,-v)|
• This property help to visualize Fourier Spectrum
Rotation
• Let x = rcosθ and y = sinθ
• u = wcosø and v = sinø
• Then we have
f(x,y) = f(r,θ) in Spatial Domain
F(u,v) = F(w, ø) in Frequency Domain
• Now Rotated Image is f(r, θ + θ0 ) and
f(r, θ + θ0 ) ↔ F(w, ø + ø0)
• F(w, ø + ø0) is F T of Rotated image
Rotation Concept
Distributivity
• DFT is distributive over addition but not on multiplication
Scaling
• If a and b are two scaling quantity then
a f(x,y) ↔ a F(u,v)
• If f(x,y) is multiplied by scalar quantity a then its F T is also
multiplied by same scalar quantity
• Scaling Individual dimension
Convolution & Correlation
• Convolution:
• Convolution in spatial domain is equivalent to multiplication in
frequency domain and vice versa
• Correlation:
• Where f* and F* indicate conjugates of f and F Correlation &
Correlation
Fast Fourier Transformation
• A 2D Fourier transform
• Has complexity O( )
• For a 1D Discrete F T complexity become O ( )
• Where we take for simplification. We have N = 2N no.
of input and we assume N = 2M
Fast Fourier Transformation
• Re-write F(u) as
• We take
• Total complexity reduces to N
