UNIT-III
[Link]
[Link]
[Link]
Not True object
Change in color
Geometric distortion
Fluctuation in the pixel (rotation, translation,
of image scale change)
Dust, movement of object
Motion
Gaussian
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
An Impulse of light Degraded impulse
[Link]
[Link]
Noise Models
• The main sources of noise in digital images arise during image acquisition (digitization)
and / or transmission
• Spatial property of refers to the correlation of noise with the image
• Frequency property refers to the frequency content in the noise and not the
electromagnetic portion
• When the Fourier spectrum noise is constant, it is called White noise
• Except the spatially periodic noise, we assume that noise is independent of spatial co-
ordinates and that it is uncorrelated with the image
• i.e. there is no correlation between pixel values of the image and that of the noise
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
Restoration with only noise
spatial parameters
[Link]
[Link]
• Spatial filter is the only choice
of noise removal
• Convolution, inverse filtering
helps , If artifacts and image
is blur
[Link]
Noise is
removed due
to blurring
mn=size of
moving
window [Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
[Link]
Adaptive Filters
[Link]
[Link]
[Link]
[Link]
Local mean =
sum of all in
window/9 [Link]
[Link]
[Link]
[Link]
[Link]
Global noise variance = (sum of [Link]
variance2) /16
0.9361-0.1709/0.2394[0.9361-0.4373]
=0.9361-0.7138[0.4988]
=0.9361 - 0.3560
=0.5802
[Link]
[Link]
Periodic Noise
Periodic noise means repeat itself at regular interval
Periodic noise arises from electrical or electromagnetic interference
The periodic noise can be reduced significantly through frequency
domain filtering
Need to identify the type of periodic noise then need to apply the
appropriate filters
Band pass : Passing certain specific bands of frequencies and
rejecting other bands
Band reject: Rejecting certain specific bands of frequencies and
passing other bands
[Link]
Continuous repetition
of noise denoted as dot
line
Rejecting dot line
[Link]
[Link]
Periodic noise
Band pass filter Band reject filter Notch filter
D1 D2
Reject frequency Pass/Reject frequency
Passing frequency
Dh Dl Dh Dl For specified frequency Dh
Dl
Reject frequency
As band reject
Reject/pass frequency
[Link] For specified zone/patches to remove more periodic noise
rather than a board spectrum
Frequency domain
[Link]
passing
particular bands
[Link]
[Link]
Frequency domain
[Link]
Rejecting
particular bands
W is width of rejection, n is order of filter, [Link]
D(u,v) distance, D0 is cutoff frequency
[Link]
[Link]
Notch filter
o Notch filter is the spatial form of band reject filter.
o Instead of removing the entire range of frequencies, it removes only
selective components
o It is useful in removing a periodic signal of a clearly define frequency
like the interference pattern caused by electrical distribution.
o A notch filter rejects (or passes) frequencies in a predefined
neighborhood of the frequency rectangle.
Purpose
Image restoration : Image having added with noise and degraded is to
be restored with its original shape
Background concept: Identification of Notch filter in frequency domain
Types
Pulses are very sharp. Rejection or
passes of certain frequency bands
are limited/specific
Equation
Ideal filter of radios D0 with center at (u0,v0 ) , must have a corresponding notch at symmetric location (−u0 ,−v0 )
rejects the frequency given as
Ideal Predefine frequency output=0
D1 and D2 are taken because of selected
frequency
These centers are specified with respect to the center of the frequency rectangle, (M /2,N /2), where, as usual, M and N are
the number of rows and columns in the input image
Butterworth n represent the
order of
Butterworth filter
Gaussian
Condition applied and transformation
o These three notch reject filters become high pass filter if u0= v0= 0
o If notch pass, they become low pass filter provided u0= v0= 0
o Due to symmetric nature of Fourier transform, notch must appear as symmetric pairs about
the origin
o Exception to this rule: if notch located at the origin so it represent by itself
np= notch pass
nr=notch reject
Optimum notch filter
o The interference patterns have been simple to identify and characterize in the frequency
domain, leading to the specification of notch filter transfer functions that also are simple
to define heuristically.
o When several interference components are present, heuristic specifications of filter
transfer functions are not always acceptable because they may remove too much image
information in the filtering process (a highly undesirable feature when images are unique
and/or expensive to acquire).
o In addition, the interference components generally are not single-frequency bursts and
It tend to have broad skirts that carry information about the interference pattern. which
are not always easily detectable from the normal transform background.
o Alternative filtering methods that reduce the effect of these degradations are quite useful
in practice.
o Hence the requirement of is optimum notch filter, in the sense that it minimizes local
variances of the restored estimate.
Procedure
Procedure
Procedure
Procedure
Output
Inverse filtering
Reverse filtering – to nullify particular portion
Image is blur due to motion blurring/smoothing means it has low
frequency signal – require De-blurring
Image corrupted with noise – so have high frequency – to remove
high frequency need to use low pass filter - need to apply denoising
Blurring and noise both are present means required different
approach
[Link]
PSF - Point spread function
Type of blur is known
Partially blur + addition with noise only known
Known about noise only
[Link]
Inverse filtering because noise is not present full knowledge
of blur
If noise is not zero, and h value is less , tend to be infinite and
there will be lot of noise, so inverse filter is not the good choice
[Link]
undegraded image =Inverse filter x degraded image
Multiplication converted to convolution
[Link]
[Link]
[Link]
Noise & blur present - It compromises both low and high pass filter resulting a band pass filter
Rejecting low pass blurring & high [Link]
noise
Statistical value of noise such as mean value
of noise, power spectrum of noise, estimation
of undegraded image
H(u,v) is
multiplying
Random
noise is
unknown
Assumption
Noise and image are uncorrelated
Either the blurring H(u,v) or noise should have 0 mean
Gray level in the estimate are the linear function of the
degraded image
[Link]
Need to change multiple
factor to estimate
[Link]
[Link]
[Link]
Wiener Filtering vs. Inverse Filtering
Full inverse filtering Limited inverse filtering Weiner filtering
[Link]
Corrupted by Image Inverse Filtering WinerFiltering
Filtering - comparison
Noise variance magnitude less by 1
Noise variance magnitude less by 5
[Link]