Image Processing

Image Filtering in the

Frequency Domain

• Low Pass Filter

• High Pass Filter
• Band pass Filter
• Blurring
• Sharpening
 Frequency Bands

Image Fourier Spectrum

Percentage of image power enclosed in circles

(small to large) :

90, 95, 98, 99, 99.5, 99.9

Blurring - Ideal Low pass Filter

90% 95%

98% 99%

99.5% 99.9%
 The Power Law of Natural Images

• The power in a disk of radii r=sqrt(u2+v2 )

follows: P(r)=Ar- where 2

Images from: Millane, Alzaidi & Hsiao - 2003

 Image Filtering

• Low pass
• High pass
• Band pass
• Local pass
• Usages
 Recall:
The Convolution Theorem

g=f*h g = fh

implies implies

G = FH G=F*H
 Low pass Filter
spatial domain frequency domain
f(x,y) F(u,v) filter

G(u,v) = F(u,v) • H(u,v)

g(x,y) G(u,v)

f(x,y) F(u,v)

g(x,y) H(u,v)
 H(u,v) - Ideal Low Pass Filter

1 D(u,v)  D0
H(u,v) =
0 D(u,v) > D0

D(u,v) =  u2 + v2
D0 = cut off frequency

H(u,v)
H(u,v)

1
v

u 0 D0 D(u,v)
Blurring - Ideal Low pass Filter

99.7%

99.37%

98.65%
Blurring - Ideal Low pass Filter

96.6% 98.0%

99.0% 99.4%

99.6% 99.7%
 The Ringing Problem

G(u,v) = F(u,v) • H(u,v)

Convolution Theorem

g(x,y) = f(x,y) * h(x,y)

IFFT sinc(x)

H(u,v) h(x,y)

 D0  Ringing radius +  blur

 The Ringing Problem

Freq. domain

Spatial domain

250

200

150

100

50

0
0 50 100 150 200 250
 H(u,v) - Gaussian Filter

H(u,v) H(u,v)
1

1/ e

v
u 0 D0 D(u,v)

-D2(u,v)/(2D20)
H(u,v) = e

D(u,v) =  u2 + v2

Softer Blurring + no Ringing

Blurring - Gaussain Lowpass Filter

99.11%

98.74%

96.44%
 The Gaussian Lowpass Filter

Freq. domain

Spatial domain

300

250

200

150

100

50

0
0 50 100 150 200 250 300
 Blurring in the Spatial Domain:

Averaging = convolution with 1 1

1 1
= point multiplication of the transform with sinc:

1 2 1
Gaussian Averaging = convolution with 2 4 2
1 2 1

= point multiplication of the transform with a gaussian.

0.15

1
0.1 0.8

0.6
0.05 0.4

0.2
0 0
0 50 100 -50 0 50
Image Domain Frequency Domain
Image Sharpening - High Pass Filter

H(u,v) - Ideal Filter

0 D(u,v)  D0
H(u,v) =
1 D(u,v) > D0

D(u,v) =  u2 + v2
D0 = cut off frequency

H(u,v)
H(u,v)

v 1

u 0 D0 D(u,v)
 High Pass Gaussian Filter

H(u,v)
H(u,v)
1

1 1 / e
v

u 0 D0 D(u,v)

-D2(u,v)/(2D20)
H(u,v) = 1 - e

D(u,v) =  u2 + v2
 High Pass Filtering

Original High Pass Filtered

 High Frequency Emphasis

Original High Pass Filtered

+
 High Frequency Emphasis

Emphasize High Frequency.

Maintain Low frequencies and Mean.

H'(u,v) = K0 + H(u,v)

(Typically K0 =1)

H'(u,v)

0 D0 D(u,v)
High Frequency Emphasis - Example

Original High Frequency Emphasis

Original High Frequency Emphasis

 High Pass Filtering - Examples

Original High pass Emphasis

High Frequency Emphasis

+
Histogram Equalization
 Band Pass Filtering

0 D(u,v)  D0 - w
2

H(u,v) = 1 D 0- w
2
 D(u,v)  D0 + w
2
w
0 D(u,v) > D0 + 2

D(u,v) =  u2 + v2
D0 = cut off frequency
w = band width

H(u,v) H(u,v)

1
v
D(u,v)
0 D0- w D0 D0+w
2 2
u
 Local Frequency Filtering

H(u,v)

u
H(u,v)

0 D0 D(u,v)
-u0,-v0 u0,v0

1 D1(u,v)  D0 or D2(u,v)  D0
H(u,v) =
0 otherwise

D1(u,v) =  (u-u0)2 + (v-v0)2

D2(u,v) =  (u+u0)2 + (v+v0)2

D0 = local frequency radius

u0,v0 = local frequency coordinates
 Band Rejection Filtering

H(u,v)

u
H(u,v)

0 D0 D(u,v)
-u0,-v0 u0,v0

0 D1(u,v)  D0 or D2(u,v)  D0
H(u,v) =
1 otherwise

D1(u,v) =  (u-u0)2 + (v-v0)2

D2(u,v) =  (u+u0)2 + (v+v0)2

D0 = local frequency radius

u0,v0 = local frequency coordinates
Demo