2D Fourier Transform

Definition

F (u , v) = F { f ( x, y )} = f ( x, y ) e i 2 ( u x + v y ) dx dy

where x, y are rectangular spatial coordinates and u, v are spatial frequencies.

We can rewrite this as two 1D transforms:

F (u, v) = f ( x, y ) e i 2 u x dx e i 2 v y dy

i.e. for each y take a 1D FT with respect to x , then take a 1D FT with respect to y.

Polar Coordinates

Since
u x + v y = r (cos cos + sin sin )
= r cos( ),
then the Fourier transform in polar coordinates is
2
F ( , ) = F { f (r , )} = f (r , )e i 2 r cos( ) r dr d
0 0
and the inverse transform is
2
f (r , ) = F 1 {F ( , )} = F ( , )e i 2 r cos( ) d d
0 0

Separability in Polar Coordinates

A function f (r , ) is separable in polar coordinates if it can be written in the form
f (r , ) = f r (r ) f ( ) .

Suppose such an f is circularly symmetric, with f ( ) = 1 . As a specific example, we will

consider a 2D circular (or cylinder) function:
1 r <1

circ(r ) = 1 / 2 r = 1
0 r >1

The Fourier transform of a circularly symmetric function is

F ( , ) = 2 r f r (r ) J 0 (2 r ) dr .
0
This is also known as the Hankel transform of order zero and as the Fourier-Bessel
transform. The function J 0 is the zero order Bessel function of the first kind defined as
1 2 ia cos( )
2 0
J 0 (a) = e d .
It oscillates like a damped cosine.

Continuing with our specific example, the Fourier transform of circ(r ) is

F {circ(r )} = 2 0 r circ(r ) J 0 (2 r ) dr = 2 0 r J 0 (2 r ) dr .
1

Substitute r '= 2 r , and dr '= 2 dr to find:

2 J 1 (2 )
F {circ(r )} =
1
2 2 0 0 x J 0 ( x) dx = J 1 ( )
r ' J ( r ' ) dr ' = , since

0

where J 1 is the first order Bessel function of the first kind, similar to a damped sinusoid.

The function somb( ) or sombrero (also known as Mexican hat, Bessinc, and jinc) is
defined as
2 J (2 )
somb( ) = 1 ,
2
and is pictured on the right side below. Thus, the Fourier transform of circ(r ) is
proportional to a sombrero function of , the radial coordinate in frequency space:
F {circ(r )} = somb( ) .