Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
35 views

Lecture 07

This document provides an outline and recap of topics related to the Fourier transform and linear filters. It discusses how the Fourier transform allows reducing differentiation to multiplication through its commutative properties. It provides examples of elementary periodic functions with different frequencies that could be combined linearly. It also gives some Python tips for a first image processing project involving the discrete Fourier transform and shifting.

Uploaded by

zaid
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
35 views

Lecture 07

This document provides an outline and recap of topics related to the Fourier transform and linear filters. It discusses how the Fourier transform allows reducing differentiation to multiplication through its commutative properties. It provides examples of elementary periodic functions with different frequencies that could be combined linearly. It also gives some Python tips for a first image processing project involving the discrete Fourier transform and shifting.

Uploaded by

zaid
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 18

Image Processing,

Retrieval, and Analysis (I)


Prof. Christian Bauckhage
Outline
Lecture 07

Recap

The Fourier Transform and Linear Filters


Motivation
Tips for the 1st Project
Derivation of the FT: Fourier Series
Recap

homogeneous coordinates
I using 3D homogeneous coordinates, affine transformations
in the 2D plane can be expressed as matrix operations,
since

x = Mu + t

can be written as
    
x M t u
=
1 00 1 1
Recap

rotation about an arbitrary point p


I rotation about the origin
    
px cos ϕ sin ϕ 0 pu
py  =  − sin ϕ cos ϕ 0 pv 
1 0 0 1 1

I rotation about p = [a, b]T


      
px 1 0 a cos ϕ sin ϕ 0 1 0 −a pu
py  = 0 1 b  − sin ϕ cos ϕ 0 0 1 −b pv 
1 0 0 1 0 0 1 0 0 1 1
Recap

example
Recap

polar coordinates
I 2D points may also be referenced using polar coordinates
y y

py
p p

px x x
Recap

the (r, ϕ)-plane

y ϕ

496 π/2

0 x 0 r
0 730 0 882.6
The Fourier Transform and Linear Filters

motivation
I get to know a different point of view for looking at g(x)
I understand it and make use of its characteristics
example 1

g(x, y) log |G(µ, ν)|


The Fourier Transform and Linear Filters

example 2

g(x, y) log |G(µ, ν)|


The Fourier Transform and Linear Filters

introductory observation
I it is one of the marvels of mathematics that difficult
operations can often be reduced to simpler ones
The Fourier Transform and Linear Filters

introductory observation
I it is one of the marvels of mathematics that difficult
operations can often be reduced to simpler ones

example: logarithm
I the relation

ln(ab) = ln a + ln b

allows for reducing multiplication to addition


The Fourier Transform and Linear Filters

example (cont.): logarithm


I the following commutative diagram does hold
ln
a, b −−−−→ ln a, ln b
 
·y
 +
y
ln−1
ab ←−−−− ln a + ln b
The Fourier Transform and Linear Filters

example: Fourier transform


I there are more abstract transformations that allow for reducing
differentiation to multiplication
I every square integrable function f (x) can be expressed as
Z
∞
1
f (x) = √ F(ω) eiωx dω
2π
−∞

I here, the function F(ω) is defined as


Z
∞
1
F(ω) = √ f (x) e−iωx dx
2π
−∞

and is called the Fourier transform of f (x)


The Fourier Transform and Linear Filters

example (cont.): Fourier transform


I using an operator notation, we write

F(ω) = F f (x)

and

f (x) = F−1 F(ω)


The Fourier Transform and Linear Filters

example (cont.): Fourier transform


I we find that the FT of the derivative of f (x) to be
Z
∞
0 iω
Ff (x) = √ f (x) e−iωx dx = iωFf (x) = iωF(ω)
2π
−∞

so that it indeed reduces differentiation to multiplication


and we have
F
f (x) −−−−→ F(ω)
 
d 
dx y y·iω
F −1
f 0 (x) ←−−−− iωF(ω)
The Fourier Transform and Linear Filters

python tips for the 1st project

fft fftshift
→ →

note:

I when you read an image and convert it to an array, make sure it


is in float format
I while g(x, y) is a real valued function, G(µ, ν) is complex valued;
only for the sake of visualization does the figure show log|G(µ, ν)|
The Fourier Transform and Linear Filters

example
I elementary periodic functions (each of period 1 but of different
frequencies 2πk) and a possible linear combination

0 1 0 1

sin 2πkx, k = 1 sin 2πkx, k = 2

0 1 0 1

P
sin 2πkx, k = 3 k αk sin(νk x + φk )
The Fourier Transform and Linear Filters

terminology
I the expression

X
n
αk sin(νk x + φk )
k=1

involves
I amplitude αk
I frequency νk
I phase φk

You might also like