Filtering in the Frequency Domain

1 / 35
2 / 35
• Compute Fourier transform of image
• Multiply the result by a filter transfer function (or simply filter).
• Take the inverse transform to produce the enhanced image.

• Summary:
• G(u,v) = H(u,v) F(u,v)

• Filtered Image =  1 G(u,v)

 3 / 35
 Spatial & Frequency Domain

●
The most fundamental relation ship between spatial and frequency
domain is established by a well known result called convolution
theorem.
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)

●
Filters in the spatial and frequency domain form a FT pair, i.e. given a
filter in the frequency domain we can get the corresponding one in the
spatial domain by taking its inverse FT.

4 / 35
5 / 35
6 / 35
●
Types of enhancement that can be done:
●
Lowpass filtering: reduce the high-frequency content
-- blurring or smoothing
●
Highpass filtering: increase the magnitude of high-frequency
components
-- sharpening.

7 / 35
8 / 35
 Lowpass Filtering in the Frequency Domain

●
Edges, noise contribute significantly to the high-frequency
content of the FT of an image.
●
Blurring/smoothing is achieved by reducing a specified
range of high-frequency components:

G (u , v) H (u , v) F (u, v)

9 / 35
 Smoothing in the Frequency Domain

G (u , v) H (u , v) F (u, v)
●
a) Ideal
●
b) Butterworth (parameter: filter order)
●
c) Gaussian

10 / 35
 Ideal Filter (Lowpass)
●
A 2-D ideal low-pass filter:
1 if D (u, v) D0
H (u, v) 
0 if D (u , v)  D0
● where D0 is a specified nonnegative quantity and D(u,v) is the
distance from point (u,v) to the center of the frequency
rectangle.
●
Center of frequency rectangle: (u,v) = (M/2 , N/2)
●
Distance from any point to the center (origin) of the FT:
D(u , v )=√(u 2 +v 2 ) 11 / 35
Ideal Filter (Lowpass)

12 / 35
 Ideal Filter (Lowpass)
●
Ideal:
● all frequencies inside a circle of radius D0 are passed with
no attenuation
●
all frequencies outside this circle are completely
attenuated.
●
Cutoff-frequency: the point of transition between
H(u,v)=1 and H(u,v)=0 (D0)
●
To establish cutoff frequency loci, we typically compute
circles that enclose specified amounts of total image
power PT. 13 / 35
 Ideal Filter (Lowpass)
●
PT is obtain by summing the components of power
spectrum P(u,v) at each point for u up to M-1 and v up to
N-1.
●
A circle with radius r, origin at the center of the frequency
rectangle encloses a percentage of the power which is
given by the expression

100[ ∑ ∑ P (u ,v )/ PT ]
u v
●
The summation is taken within the circle r

14 / 35
Ideal Filter (Lowpass)

15 / 35
16 / 35
17 / 35
 Butterworth Filter (Lowpass)
●
This filter does not have a sharp discontinuity establishing
a clear cutoff between passed and filtered frequencies.
1
H (u , v) 
1  [ D(u, v) / D0 ]2 n
●
To define a cutoff frequency locus: at points for which
H(u,v) is down to a certain fraction of its maximum value.
● When D(u,v) = D0, H(u,v) = 0.5
●
i.e. down 50% from its maximum value of 1.
18 / 35
Butterworth Filter (Lowpass)

19 / 35
20 / 35
21 / 35
 Gaussian Filter (Lowpass)

 D 2 (u,v )/ 2 2
H(u,v) e
●
D(u,v): distance from the origin of FT
●
Parameter:  = D0 (cutoff frequency)
●  inverse FT of the Gaussian filter is also a Gaussian
The

22 / 35
Gaussian Filter (Lowpass)

23 / 35
24 / 35
Gaussian Filter (Lowpass)

25 / 35
Gaussian Filter (Lowpass)

26 / 35
Gaussian Filter (Lowpass)

27 / 35
 Sharpening (Highpass) Filtering
●
Image sharpening can be achieved by a highpass
filtering process, which attenuates the low-frequency
components without disturbing high-frequency
information.

●
Zero-phase-shift filters: radially symmetric and
completely specified by a cross section.
H hp (u,v) 1 H lp (u,v)

28 / 35
Sharpening (Highpass) Filtering

29 / 35
 Ideal Filter (Highpass)

●
This filter is the opposite of the ideal lowpass filter.

0 if D(u , v) D0
H (u , v) 
1 if D(u , v)  D0

30 / 35
 Butterworth Filter (Highpass)

●
High-frequency emphasis: Adding a constant to a
highpass filter to preserve the low-frequency components.
1
H (u , v) 
1  [ D0 / D(u, v)]2 n

31 / 35
 Gaussian Highpass Filter

 D 2 (u,v )/ 2 2
H(u,v) 1 e

●
D(u,v): distance from the origin of FT

 ●
Parameter:  = D0 (cutoff frequency)

32 / 35
 Laplacian (recap)
2 2
2  f  f
 f  2  2
x y
2 f
2 2
 f (x 1, y)  f (x  1, y)  2 f (x, y)
 x
2 f
 2 2
 f (x, y 1)  f (x, y  1)  2 f (x, y)
 y

2 f [ f (x 1, y)  f (x  1, y)  f (x, y 1)  f (x, y  1)]  4 f (x, y)

●
FT
pair:

2 f (x,
y)   [(u  M /2) 2  (v  N /2) 2 ]F(u,v)

34 / 35
35 / 35