Professional Documents
Culture Documents
Chapter 5
Chapter 5
Chapter Five:
Degradation model
Noise model
2
Image restoration
• Image restoration means tries to reconstruct or recover an image
which was degraded using a priori knowledge of degradation.
• Here we have model degradation and apply the inverse process to
recover the original image.
• The main aim of restoration is to improve an image in some
predefined way.
• Image restoration attempts to restore images that have been degraded:
o Identify the degradation process and attempt to reverse it
o Similar to image enhancement, but more objective
3
Degradation: Example
4
Model of Image Degradation
5
Model of Image D egradation
6
Model of Image Degradation
G (u , v ) H (u , v ) F (u , v ) N (u , v )
7
Noise Sources
• The principal sources of noise in digital images arise during image
acquisition and/or transmission.
o Image acquisition
e.g., light levels, sensor temperature, etc.
o Transmission
e.g., lightning or other atmospheric disturbance in
wireless network
8
Noise type
Noise in images can be defined as unwanted variation of brightness or color
information.
Noise type describes the nature or characteristics of the noise present in the
image, such as its distribution, spatial or temporal properties, and how it
affects pixel values
◦ Impulse Noise (Salt and Pepper Noise): Impulse noise also referred as data drop-
out-noise, intensity spikes or salt and pepper noise.
o Impulse noise affects individual pixels by randomly changing their intensity
values to either the maximum (1) or minimum (0) intensity levels.
oThis type of noise basically occurs during transmission.
10
Noise Model
A noise model is a mathematical representation or approximation of a specific
type of noise.
Noise models allow researchers and engineers to analyze the properties of
noise, develop algorithms for noise reduction or removal in digital images.
◦ Gaussian/additive white Gaussian noise (AWGN) model: Electronic circuit
noise, sensor noise due to poor illumination.
◦ Rayleigh noise model: Often seen in range imaging applications, such as radar
or sonar.
◦ Gamma(Erlang) and Exponential Noise Models: Found in laser (highly focused)
imaging systems, like laser scanning microscopy.
◦ Impulse(salt-and-pepper) noise model: can occur due to various factors,
including faulty switches, sensor glitches, or electrical interference.
◦ Uniform noise model: Least commonly used, lacks a specific application.
11
Noise Probability Density Functions
Noise probability density functions (PDFs) describe the statistical
distribution of noise values in digital images.
Different types of noise exhibit unique PDFs, which help characterize
their properties and behavior.
Noise cannot be predicted but can be approximately described in
statistical -- using the probability density function (PDF)
◦ Gaussian noise
Electronic circuit noise and sensor noise
Used frequently in practice
Tractable or handle both in the spatial and frequency domains
1 ( z ) 2 / 2 2
p( z ) e
2
12
Noise Probability Density Functions
Rayleigh noise
b(4 )
a b / 4 and 2
14
Noise Probability Density Functions
Exponential noise
ae az for z 0 Where a > 0
p( z )
0 for z 0
– The mean and variance of this density are given by:
1
1 / a and 2
2
15
Noise Probability Density Functions
Uniform Noise
ab
2
(b a ) 2
2
12
16
Noise Probability Density Functions
Impulse (salt-and-pepper) noise
- Quick transients, such as faulty switching during imaging.
Pa for z a
p ( z ) Pb for z b
0 otherwise
- Where,
• a: Represents one of the intensity levels corresponding to "salt" noise, where
pixels are typically set to a high intensity value.
• b: Represents the other intensity level corresponding to "pepper" noise, where
pixels are typically set to a low intensity value.
• Pa: Represents the probability density associated with intensity level a, which is
the probability of occurrence of "salt" noise.
• Pb: Represents the probability density associated with intensity level b, which is
the probability of occurrence of "pepper" noise.
17
Noise Probability Density Functions
18
Noise Probability Density Functions
g ( x, y ) f ( x, y ) ( x, y )
Degraded images
Original image
Histogram
20
Noise Probability Density Functions example
Degraded images
Histogram
21
Periodic Noise
Arises typically from electrical or electromechanical interference
during image acquisition.
It can be observed by visual inspection both in the spatial domain and
frequency domain
Can be reduced significantly via frequency domain filtering
◦ Order-Statistics: Median filter, Max and min filters, Mid-point filter, Alpha-
trimmed filters.
23
Mean Filter
Let Sxy represent the set of coordinates in a rectangular subimage
window of size m*n, centered at point (x,y).
The arithmetic mean filtering process computes the average value of
the corrupted image g(x,y) in the area defined by Sxy.
The value of the restored image f at any point (x,y) is simply the
arithmetic mean computed using the pixels in the region defined by Sxy.
In other words:
ˆf ( x, y ) 1
g ( s, t )
mn ( s ,t )S xy
Is the simplest of the mean filters mn = size of moving window
Smooths local variations in an image
Noise is reduced as a result of blurring
24
Mean Filter
Degradation model: g ( x, y ) f ( x, y ) h ( x, y ) ( x, y )
Where,
• The symbol ∏ like ∑, but instead of summing up terms, it multiplies them together.
• 𝑓(𝑥+𝑖,𝑦+𝑗) represents the intensity value of the pixel at position (𝑥+𝑖,𝑦+𝑗) in the
original image, where 𝑖 and j range over the dimensions of the filter kernel
(typically M×N).
• MN is the total number of pixels within the filter kernel
Image
Original
corrupted
image
by AWGN
Image Image
obtained obtained
using a 3x3 using a 3x3
arithmetic geometric
mean filter mean filter
26
Mean Filter
Harmonic mean filter
mn
fˆ ( x , y ) Works well for salt noise
1
( s ,t )S xy g ( s, t )
but fails for pepper noise
mn = size of moving window
Contraharmonic mean filter
g ( s , t ) Q 1
Positive Q is suitable for
eliminating pepper noise.
fˆ ( x , y )
( s ,t )S xy
g ( s, t )
( s ,t )S xy
Q Negative Q is suitable for
eliminating salt noise.
27
Mean Filter
Image Image
corrupted corrupted
by pepper by salt
noise with noise with
prob. = 0.1 prob. = 0.1
Image Image
obtained obtained
using a 3x3 using a 3x3
contra- contra-
harmonic harmonic
mean filter mean filter
With Q = With Q=-
1.5 1.5
28
Order-Statistic Filters
subimage
Statistic parameters
Original image Mean, Median, Mode,
Min, Max, Etc.
Moving
window
Output image
29
Order-Statistic Filters
Median filter
fˆ ( x , y ) median g ( s, t )
( s ,t )S xy
Max filter
Min filter
1
Midpoint filter fˆ ( x, y ) max g ( s, t ) min g ( s, t )
2 ( s ,t )S xy ( s ,t )S xy
30
Median Filter
A median filter is good for removing impulse, isolated noise
Pepper noise
Median
Sorted
Degraded image Moving array
window
Salt noise Filter output
31
Median Filter Example
1 2
Image
corrupted
by salt-and-
pepper
noise with
pa=pb= 0.1
3 4
Image Image
obtained obtained
using a 3x3 using a 3x3
max filter min filter
33
Adaptive Filter
• An adaptive filter is a type of digital filter that adjusts its parameters
based on the input signal and the desired output signal.
• Unlike traditional fixed filters with predefined coefficients, adaptive
filters dynamically modify their coefficients or parameters to adapt to
changes in the input signal or the environment.
• Thus far the filtration techniques are applied to an image without
regard for how image characteristics vary from one point to another.
34
Adaptive, Local Noise Reduction Filter
36
Adaptive, Local Noise Reduction Filter
Fig: (a) Image corrupted by additive Gaussian noise of zero mean and variance 1000.
(b) Result of arithmetic mean filtering.
(c) Result of geometric mean filtering.
(d) Result of adaptive noise reduction filtering. All filters were of size 7X7. 37
Adaptive Median Filter
Median filter performs well if the spatial density of the impulse noise is
not large, i.e., (Pa and Pb less than 0.2)
38
Adaptive Median Filter
Remove salt-and-pepper noise
Providing smoothing of other noise.
Remove distortion (excessive thickening and thinning of object
boundary
40
Bandreject
41
Cont…
42
Bandreject Example
44
Notch Filter
A notch filter rejects ( or passes) frequencies in predefined
neighborhoods about a center frequency.
45
Notch Filter
46
Notch Filtering
Fig: (a) Satellite image of Florida and the Gulf of Mexico (note horizontat sensor scan lines).
(b) Spectrum of (a). (c) Notch pass filter shown superimposed on (b).
(d) inverse Fourier transform of filtered image , showing noise pattern in spatial domain.
(e) result of notch reject filtering. 47
Estimation of Image Degradation
Degradation model:
g ( x, y ) f ( x, y ) h ( x, y ) ( x, y )
or
G ( u, v ) F ( u, v ) H ( u, v ) N ( u, v )