Dip

Digital Image Processing
Presented By,
T. Janani,
II-M.sc(CS&IT)
Nadar Saraswathi College
Of Arts & Science, Theni.

Degradation Model :
H – Degradation operator
f(x , y) - Input Image
g(x , y) - Degraded Output

Degradation models : noise only
),(),(),(
),(),(),(
vuNvuFvuG
yxyxfyxg

 
Noise models
1. Spatial characteristics (independent or dependent)
2. Intensity ( distribution, spectrum)
* Uniform, Gaussian, Rayleigh, Gamma (Erlang),
Exponential, impulse
3. Correlation with the image (additive, multiplicative)
De-noising
1. Spatial filtering
2. Frequency domain filtering

Noise models – periodic noise

De-noising
Estimation of noise parameters:
* By spectrum inspection: for periodic noise
* By test image: mean, variance and histogram shape,
if imaging system is available
* By small patches, if only image is available
De-noising :
1. Spatial filtering ( for additive noise)
* Mean filters
* Order-statistics filters
* Adaptive filters
2. Frequency domain filtering (for periodic noise)



xySts
tsg
mn
yxf
),(
),(
1
),(ˆ
mn
Sts xy
tsgyxf
1
),(
),(),(ˆ








 
Arithmetic mean filter
Geometric mean filter
De-noising – Gaussian noise example

De-noising – evaluation
1. PSNR
2. Visual perception
   

M
i
N
j
ijij SSMN
MSE
1 1
2
'1







MSE
PSNR
2
10
255
log10

De-noising – evaluation example

Degradation models: linear vs non-linear
1. Many types of degradation can be approximated
by linear, space invariant processes
* Can take advantages of the mature
techniques developed for linear systems
2. Non-linear and space variant models are more
accurate
* Difficult to solve
* Unsolvable

Linear, space-invariant degradations
 
 
 
 
),(),(),(
),(),,,(),(
),(),(),(
),(),(),(
),(),(),(
),(),(),(
),(),(),(
yxddyxhf
yxddyxhf
yxddyxHf
yxddyxfH
yxddyxfH
yxyxfHyxg
ddyxfyxf



















Sampling theorem -->
Linearity, additivity -->
Linearity, homogeneity -->
Space-invariant -->
(convolution integral)

Diagonalization of circulant and block- circulant
matrices
Circulant Matrices:
For MXM circulant matrix :
Define a scalar :

Define a vector :
for k=0,1,2…..M-1

Effects of Diagonalization on the Degradation Model :

In other words, 1-D discrete Formulation of
degradation model is 1D Fourier transform
G(k)=MH(k)F(k)
for k=0.1,2,………….,M-1

Dip

More Related Content

What's hot (20)

Similar to Dip (20)

More from janani thirupathi (17)

Recently uploaded (20)

Dip