Content uploaded by Christo Ananth
Author content
All content in this area was uploaded by Christo Ananth on Apr 29, 2016
Content may be subject to copyright.
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
35
A
ll Rights Reserved © 2015 ICRAMATE16
BLUR REDUCTION AT HIGH DENSITY IMPULSE NOISE
Kavya.C.1, Chitra.V2, Blessie Beulah.K.3, Christo Ananth4
P.G. Scholars, Department of M.E. Communication Systems,
Francis Xavier Engineering College, Tirunelveli 1,2,3
Associate Professor, Department of ECE,
Francis Xavier Engineering College, Tirunelveli4
Abstract--This improved method is a
simple, and efficient way to remove impulse
noise from highly corrupted digital images. This
method has two stages. The first stage is to
detect the impulse noise in the image. In this
stage, the pixels are divided into two classes
(noise free pixels/ noise free pixels) based on
only the intensity values .Then, the second stage
is to eliminate the impulse noise from the image.
In this stage, only the “noise-pixels” are
processed. But the “noise-free pixels” are not
modified and are copied directly to the output
image. The method used gradient based adptive
median filter,so that this method adaptively
changes the size of the median filter based on
the number of the “noise-free pixels” in the
neighborhood. For the filtering, the gradian
value of every pixel location at (x,y) is
calculated. Then the median value can be find
out under the consideration of only “noise-free
pixels” .In this algorithm for effective noise
detection is proposed. The Proposed algorithm
produces better edge and fine details
preservations and reduces blurring at the high
density impulse noise. Because of its simplicity,
this proposed method is suitable to be
implemented in consumer electronics products
such as digital television, or digital camera.
Index Terms — Noise reduction, salt-and-pepper
noise,impulse noise, median filter, adaptive
filter.
I.INTRODUCTION
Impulse noise removal in digital images
is an important pre-processing step as images are
often corrupted by impulse noise. Impulse noise is
caused by malfunctioning pixels in camera sensors,
faulty memory locations in hardware, or
transmission in a noisy channel. Two common
types of impulse noise are the salt-and-pepper noise
and the random valued noise. For images corrupted
by salt- and-pepper noise (respectively random-
valued noise), the noisy pixels can take only the
maximum and the minimum values (respectively
any random value) in the dynamic range. Two
types of filters used for remove the impulse noise is
linear filter and non-linear filter.
A large number of linear and non linear
filtering algorithms have been proposed to remove
impulse noise from corrupted image to enhance
image quality. Most of the linear filters mechanism
generally used to remove impulse noise and tend to
destroy all high frequency details like edges, lines
and other fine image details. This led to the
development of nonlinear median-type filters.
Some of them are fixed and some others are
adaptive. Median filter is one of the order-statistic
filters, which falls in the group of nonlinear filter.
In this project, I present a new median
filter based technique, which is a hybrid of adaptive
median filter and switching median filter. This
proposed method is fast, simple, and adaptable to
the local noise level. The method can remove the
impulse noise effectively from the image, and at
the same time can preserve the details inside the
image, even when the input image is very highly
corrupted by the noise. The method is also does
not require any parameter to be tuned, thus suitable
for an automated system. The method does not
need previous training.
I use switching median filter framework in
order to speed up the process, because only the
noise pixels are filtered. In addition to this,
switching median filter also allows local details in
the image to be preserved. We divide this method
into two stages, which are the noise detection, and
the noise cancellation. In order to implement this
method, we need three 2D arrays of the same size
to hold the pixel values of the input image f, the
output image g, and the mask to mark the noise
pixels α. The dimensions of these arrays are equal
to the dimensions of f (i.e. M×N). The two stages
of my proposed method is noise detection ,noise
cancellation .
First part of my project is to detect the
pixels from highly corrupted image, and classify
that particular image pixels into two categories ,
such as “noise free pixels’’ and “noise pixels”. then
calculate the total number of noise pixels, from that
result estimate the local impulse noise level. And
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
36
A
ll Rights Reserved © 2015 ICRAMATE16
also I compare various non-linear filter
performance over on impulse noise with my
proposed filter.
A novel class of nonlinear filter for image
processing known as order statistics (OS) filter.
Median based filters are the popular methods to be
employed for reducing the impulse noise level from
corrupted images. This is because of their
simplicity and capability to preserve edges . This
median filter method, processes all pixels in the
image equally, including the noise free pixels. This
will result the elimination of fine details such as
thin lines and corner, blurring, or distortion in the
images.
Another type of the median based methods
is the switching method, which is constructed from
two stages. The first stage in these methods is
normally to detect the noise pixel. The second
stage is to remove the noise. In this stage, only
noise pixels are filtered. Other pixels, which are
considered as noise-free pixels, are kept
unchanged.
Luo has proposed a simple switching
median filter based on fuzzy impulse detection
technique. First, the method finds the two peaks
from the histogram to approximate two intensity
values that present the impulse noise. Pixels with
these intensity values become the candidates for the
impulse noise. For each candidate, the minimum
absolute intensity difference between a pixel with
its eight-neighbors in a 3×3 window is determined.
This minimum value is assigned to the value of r, a
parameter that indicate how likely the pixel to be
an impulse noise. The minimum value is selected in
order to preserve the fine details in the image.
Then, based on the value of r and two predefined
threshold values, T1 and T2 (i.e. T1 < T2), the
pixels are grouped into three groups, which are;
“noise-free pixels”, “noise-pixels”, and “possibly
noise pixels”. The filtering is only carried out to
the “noise-pixels” and “possibly noise-pixels”. For
the “noise-pixels”, filtering in (1) is carried out. For
the “possible noise-pixels”, the median value based
on (1) is first determined, and then this value is
recombined with the intensity of the image, using
weighting value that is calculated based on a fuzzy
membership function.
Another filter known as signal adaptive median
filter, that performs better than other nonlinear
adaptive filters for different kinds of noise.
Adaptive filters adapt themselves to the noise type
and noise power level and, thus, they perform
better than the fixed filters. Two types of adaptive
images filters are proposed in this thesis. One is
based on Order statistics adaptation scheme. It
performs much better than the linear filter. It may
not need an on-line training always. An off-line
training is good enough. The size of the median
filter applied to the individual pixel is determined
based on the approximation of the local noise level.
Bigger filter is applied to the areas with high level
of noise, and smaller filter is applied to the areas
with low level of noise. That is approximate the
noise pixels based on the minimum, maximum and
median intensity values contained inside a local
window. Christo Ananth et al. [1] proposed a
method in which the minimization is per-formed in
a sequential manner by the fusion move algorithm
that uses the QPBO min-cut algorithm. Multi-shape
GCs are proven to be more beneficial than single-
shape GCs. Hence, the segmentation methods are
validated by calculating statistical measures. The
false positive (FP) is reduced and sensitivity and
specificity improved by multiple MTANN. Christo
Ananth et al. [2] proposed a system, this system has
concentrated on finding a fast and interactive
segmentation method for liver and tumor
segmentation. In the pre-processing stage, Mean
shift filter is applied to CT image process and
statistical thresholding method is applied for
reducing processing area with improving detections
rate. In the Second stage, the liver region has been
segmented using the algorithm of the proposed
method. Next, the tumor region has been
segmented using Geodesic Graph cut method.
Results show that the proposed method is less
prone to shortcutting than typical graph cut
methods while being less sensitive to seed
placement and better at edge localization than
geodesic methods.
This leads to increased segmentation
accuracy and reduced effort on the part of the user.
Finally Segmented Liver and Tumor Regions were
shown from the abdominal Computed
Tomographic image. Christo Ananth et al. [3]
proposed a system, in which a predicate is defined
for measuring the evidence for a boundary between
two regions using Geodesic Graph-based
representation of the image. The algorithm is
applied to image segmentation using two different
kinds of local neighborhoods in constructing the
graph. Liver and hepatic tumor segmentation can
be automatically processed by the Geodesic graph-
cut based method. This system has concentrated on
finding a fast and interactive segmentation method
for liver and tumor segmentation. In the
preprocessing stage, the CT image process is
carried over with mean shift filter and statistical
thresholding method for reducing processing area
with improving detections rate. Second stage is
liver segmentation; the liver region has been
segmented using the algorithm of the proposed
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
37
A
ll Rights Reserved © 2015 ICRAMATE16
method. The next stage tumor segmentation also
followed the same steps.
Finally the liver and tumor regions are
separately segmented from the computer
tomography image. Christo Ananth et al. [4]
proposed a system in which the cross-diamond
search algorithm employs two diamond search
patterns (a large and small) and a halfway-stop
technique. It finds small motion vectors with fewer
search points than the DS algorithm while
maintaining similar or even better search quality.
The efficient Three Step Search (E3SS) algorithm
requires less computation and performs better in
terms of PSNR. Modified objected block-base
vector search algorithm (MOBS) fully utilizes the
correlations existing in motion vectors to reduce
the computations. Fast Objected - Base Efficient
(FOBE) Three Step Search algorithm combines
E3SS and MOBS. By combining these two existing
algorithms CDS and MOBS, a new algorithm is
proposed with reduced computational complexity
without degradation in quality. Christo Ananth et
al. [5] proposed a system in which this study
presented the implementation of two fully
automatic liver and tumors segmentation
techniques and their comparative assessment. The
described adaptive initialization method enabled
fully automatic liver surface segmentation with
both GVF active contour and graph-cut techniques,
demonstrating the feasibility of two different
approaches. The comparative assessment showed
that the graph-cut method provided superior results
in terms of accuracy and did not present the
described main limitations related to the GVF
method. The proposed image processing method
will improve computerized CT-based 3-D
visualizations enabling noninvasive diagnosis of
hepatic tumors. The described imaging approach
might be valuable also for monitoring of
postoperative outcomes through CT-volumetric
assessments.
Processing time is an important feature for
any computer-aided diagnosis system, especially in
the intra-operative phase. Christo Ananth et al. [6]
proposed a system in which an automatic anatomy
segmentation method is proposed which effectively
combines the Active Appearance Model, Live Wire
and Graph Cut (ALG) ideas to exploit their
complementary strengths. It consists of three main
parts: model building, initialization, and
delineation. For the initialization (recognition) part,
a pseudo strategy is employed and the organs are
segmented slice by slice via the OAAM (Oriented
Active Appearance method). The purpose of
initialization is to provide rough object localization
and shape constraints for a latter GC method,
which will produce refined delineation. It is better
to have a fast and robust method than a slow and
more accurate technique for initialization. Christo
Ananth et al. [7] proposed a system which uses
intermediate features of maximum overlap wavelet
transform (IMOWT) as a pre-processing step. The
coefficients derived from IMOWT are subjected to
2D histogram Grouping. This method is simple,
fast and unsupervised. 2D histograms are used to
obtain Grouping of color image. This Grouping
output gives three segmentation maps which are
fused together to get the final segmented output.
This method produces good segmentation results
when compared to the direct application of 2D
Histogram Grouping.
IMOWT is the efficient transform in
which a set of wavelet features of the same size of
various levels of resolutions and different local
window sizes for different levels are used. IMOWT
is efficient because of its time effectiveness,
flexibility and translation invariance which are
useful for good segmentation results. Christo
Ananth et al. [8] proposed a system in which OWT
extracts wavelet features which give a good
separation of different patterns. Moreover the
proposed algorithm uses morphological operators
for effective segmentation. From the qualitative
and quantitative results, it is concluded that our
proposed method has improved segmentation
quality and it is reliable, fast and can be used with
reduced computational complexity than direct
applications of Histogram Clustering. The main
advantage of this method is the use of single
parameter and also very faster. While comparing
with five color spaces, segmentation scheme
produces results noticeably better in RGB color
space compared to all other color spaces. Christo
Ananth et al. [9] presented an automatic
segmentation method which effectively combines
Active Contour Model, Live Wire method and
Graph Cut approach (CLG). The aim of Live wire
method is to provide control to the user on
segmentation process during execution. Active
Contour Model provides a statistical model of
object shape and appearance to a new image which
are built during a training phase. In the graph cut
technique, each pixel is represented as a node and
the distance between those nodes is represented as
edges. In graph theory, a cut is a partition of the
nodes that divides the graph into two disjoint
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
38
A
ll Rights Reserved © 2015 ICRAMATE16
subsets. For initialization, a pseudo strategy is
employed and the organs are segmented slice by
slice through the OACAM (Oriented Active
Contour Appearance Model). Initialization
provides rough object localization and shape
constraints which produce refined delineation.
This method is tested with different set of
images including CT and MR images especially 3D
images and produced perfect segmentation results.
Christo Ananth et al. [10] proposed a work, in this
work, a framework of feature distribution scheme
is proposed for object matching. In this approach,
information is distributed in such a way that each
individual node maintains only a small amount of
information about the objects seen by the network.
Nevertheless, this amount is sufficient to efficiently
route queries through the network without any
degradation of the matching performance. Digital
image processing approaches have been
investigated to reconstruct a high resolution image
from aliased low resolution images. The accurate
registrations between low resolution images are
very important to the reconstruction of a high
resolution image. The proposed feature distribution
scheme results in far lower network traffic load.
To achieve the maximum performance as
with the full distribution of feature vectors, a set of
requirements regarding abstraction, storage space,
similarity metric and convergence has been
proposed to implement this work in C++ and QT.
Christo Ananth et al. [11] discussed about an
important work which presents a metal detecting
robot using RF communication with wireless audio
and video transmission and it is designed and
implemented with Atmel 89C51 MCU in
embedded system domain. The robot is moved in
particular direction using switches and the images
are captured along with the audio and images are
watched on the television .Experimental work has
been carried out carefully. The result shows that
higher efficiency is indeed achieved using the
embedded system. The proposed method is verified
to be highly beneficial for the security purpose and
industrial purpose. The mine sensor worked at a
constant speed without any problem despite its
extension, meeting the specification required for
the mine detection sensor. It contributed to the
improvement of detection rate, while enhancing the
operability as evidenced by completion of all the
detection work as scheduled. The tests
demonstrated that the robot would not pose any
performance problem for installation of the mine
detection sensor. On the other hand, however, the
tests also clearly indicated areas where
improvement, modification, specification change
and additional features to the robot are required to
serve better for the intended purpose. Valuable data
and hints were obtained in connection with such
issues as control method with the mine detection
robot tilted, merits and drawbacks of mounting the
sensor, cost, handling the cable between the robot
and support vehicle, maintainability, serviceability
and easiness of adjustments.
These issues became identified as a result
of our engineers conducting both the domestic tests
and the overseas tests by themselves, and in this
respect the findings were all the more practical.
Christo Ananth et al. [12] discussed about Vision
based Path Planning and Tracking control using
Mobile Robot. This paper proposes a novel
methodology for autonomous mobile robot
navigation utilizing the concept of tracking control.
Vision-based path planning and subsequent
tracking are performed by utilizing proposed stable
adaptive state feedback fuzzy tracking controllers
designed using the Lyapunov theory and particle-
swarm-optimization (PSO)-based hybrid
approaches. The objective is to design two self-
adaptive fuzzy controllers, for x-direction and y-
direction movements, optimizing both its structures
and free parameters, such that the designed
controllers can guarantee desired stability and,
simultaneously, can provide satisfactory tracking
performance for the vision-based navigation of
mobile robot.
The design methodology for the
controllers simultaneously utilizes the global search
capability of PSO and Lyapunovtheory-based local
search method, thus providing a high degree of
automation. Two different variants of hybrid
approaches have been employed in this work. The
proposed schemes have been implemented in both
simulation and experimentations with a real robot,
and the results demonstrate the usefulness of the
proposed concept. Christo Ananth et al. [13]
discussed about a model, a new model is designed
for boundary detection and applied it to object
segmentation problem in medical images. Our edge
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
39
A
ll Rights Reserved © 2015 ICRAMATE16
following technique incorporates a vector image
model and the edge map information. The proposed
technique was applied to detect the object
boundaries in several types of noisy images where
the ill-defined edges were encountered. The
proposed techniques performances on object
segmentation and computation time were evaluated
by comparing with the popular methods, i.e., the
ACM, GVF snake models. Several synthetic noisy
images were created and tested.
The method is successfully tested in
different types of medical images including aortas
in cardiovascular MR images, and heart in CT
images. Christo Ananth et al. [14] discussed about
the issue of intuitive frontal area/foundation
division in still pictures is of awesome down to
earth significance in picture altering. They maintain
a strategic distance from the limit length
predisposition of chart cut strategies and results in
expanded affectability to seed situation. Another
proposed technique for completely programmed
handling structures is given taking into account
Graph-cut and Geodesic Graph cut calculations.
This paper addresses the issue of dividing liver and
tumor locales from the stomach CT pictures. The
absence of edge displaying in geodesic or
comparable methodologies confines their capacity
to exactly restrict object limits, something at which
chart cut strategies by and large exceed
expectations. A predicate is characterized for
measuring the confirmation for a limit between two
locales utilizing Geodesic Graph-based
representation of the picture. The calculation is
connected to picture division utilizing two various
types of nearby neighborhoods in building the
chart. Liver and hepatic tumor division can be
naturally prepared by the Geodesic chart cut based
strategy. This framework has focused on finding a
quick and intuitive division strategy for liver and
tumor division.
In the pre-handling stage, Mean
movement channel is connected to CT picture
process and factual thresholding technique is
connected for diminishing preparing zone with
enhancing discoveries rate. In the Second stage, the
liver area has been divided utilizing the calculation
of the proposed strategy. Next, the tumor district
has been portioned utilizing Geodesic Graph cut
strategy. Results demonstrate that the proposed
strategy is less inclined to shortcutting than run of
the mill diagram cut techniques while being less
delicate to seed position and preferable at edge
restriction over geodesic strategies. This prompts
expanded division exactness and decreased
exertion with respect to the client. At long last
Segmented Liver and Tumor Regions were
appeared from the stomach Computed
Tomographic picture. Christo Ananth et al. [15]
discussed about efficient content-based medical
image retrieval, dignified according to the Patterns
for Next generation Database systems (PANDA)
framework for pattern representation and
management. The proposed scheme use 2-D
Wavelet Transform that involves block-based low-
level feature extraction from images. An
expectation–maximization algorithm is used to
cluster the feature space to form higher level,
semantically meaningful patterns. Then, the 2-
component property of PANDA is exploited: the
similarity between two clusters is estimated as a
function of the similarity of both their structures
and the measure components. Experiments were
performed on a large set of reference radiographic
images, using different kinds of features to encode
the low-level image content. Through this
experimentation, it is shown that the proposed
scheme can be efficiently and effectively applied
for medical image retrieval from large databases,
providing unsupervised semantic interpretation of
the results, which can be further extended by
knowledge representation methodologies.
II. BACKGROUND
Median filter is a low-pass filter that
attempts to remove noisy pixels while keeping the
edges intact. The value of the pixels in the window
are sorted and the median (the middle value in the
sorted list) is chosen. The MED filter is a
smoothing device for discrete signals. In particular,
he noted this filtering process to be quite effective
in suppressing impulse noise as well as preserving
the locally monotonic signal structures often
containing significant information. The MED filter
has been extensively used in image processing,
particularly for suppressing impulse noise in an
image. Many variants of the MED filter have also
been proposed. One modified version of it is the
center weighted median (CWM) filter. The ranked-
ordered mean (ROM) is another variant.
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
40
A
ll Rights Reserved © 2015 ICRAMATE16
For example, if a 3×3 window is used for
spatial sampling, then 9 pixel data are available at a
time. First of all, the 2-D data is converted to a 1-D
data, i.e. a vector. Let this vector of 9 data be
sorted. Then, if the mid value (5
th
position pixel
value in the sorted vector of length 9) is taken, it
becomes median filtering with the filter weight
vector [0 0 0 0 1 0 0 0 0]. If all the order statistics
are given equal weightage, then it becomes a mean
or moving average (MAV) filter. Strictly speaking,
the MAV filter is a simple linear filter.
An input image the filtered image
is defined by
where R takes any positive integer value.
III.PROPOSED METHOD
The filters used in my proposed method
are given below.
A. Adaptive median filter
The size of the median filter applied to the
individual pixels is determine based on the
approximation of the local noise level. Big filter is
applied to the areas with high level of noise, and
the smaller filter is applied to the areas with low
noise level. That is approximate the noise level
based on the noise pixels contained inside the local
window. Here they start with smaller window size
first, and increase its size until conditions are met.
The Adaptive Median Filter is designed to
eliminate the problems faced with the standard
median filter. The basic difference between the two
filters is that, in the Adaptive Median Filter, the
size of the window surrounding each pixel is
variable. This variation depends on the median of
the pixels in the present window. If the median
value is an impulse, then the size of the window is
expanded. Otherwise, further processing is done
on the part of the image within the current window
specifications. ‘Processing’ the image basically
entails the following: The center pixel of the
window is evaluated to verify whether it is an
impulse or not. If it is an impulse, then the new
value of that pixel in the filtered image will be the
median value of the pixels in that window.
This method have a filter known as gradient
based adaptive median filter, the purpose of that
filter is if a chance a noise free pixel value in an
image has the same value of noise pixels that time
that paticular pixels is not filtered, intead it
replaced by its gradient value. Thus, the Adaptive
Median Filter solves the dual purpose of removing
the impulse noise from the image and reducing
distortion in the image.
1) Notations
The adaptive filter works on a rectangular
region
. The adaptive median filter changes the
size of
during the filtering operation depending
on certain criteria as listed below. The output of the
filter is a single value which the replaces the
current pixel value at (x, y), the point on which
is centered at the time. The following notation is
adapted from the book and is reintroduced here:
minimum gray-level value
in
,
maximum gray-level value in
,
median gray-level value in
,
gray-
level at (x,y),
maximum allowed size
of
2) Algorithm
1. The adaptive median filter
algorithm works in two levels: A
and B
2. Level A: A1=
A2=
3. If A1>0 and A2<0 go to level
B else increase the window size
4. If window size ≤
repeat
level A else output
5. Level B: B1=
B2=
6. If B1>0 and B2<0, output
else output
The algorithm has three main purposes,
1. To remove ‘Salt and Pepper’
noise
2. To smoothen any non impulsive
noise
3. To reduce excessive distortions
such as too much thinning or
thickening of object boundaries.
B. Switched median Filter
Switching median filter speed up the
process, because only the noise pixels are filtered.
In addition to this, switching median filter also
allows local details in the image to be preserved.
C. Noise detection
There are two main purposes of this
stage. The first one is to identify the “noise pixel”,
and the second one is to roughly approximate the
noise level of the image.
International Conference on Recent Advances in Management,
International Journal of Advanced Research in Management,
A
ll Rights Reserved © 2015 ICRAMATE16
The histogram of the input image to
find two intensity
values that presents the impulse
noises. The identification is based on the peak
values contained in the histogram. He assumes that
the bright and dark intensity values of the impulse
noise produce two peaks in the histogram.
However, in our method, we
do not use this
assumption because this statement is not always
true, especially when the input is only corrupted by
low level of noise.
Fig.1. schematic diagram of proposed
method
The steps needed for the noise
detection is,
1)
Identifying the pixels
free pixels and noise pixels).
2)
Calculate the total number of
noise pixels K.
3)
Roughly estimate the impulse
noise level .
1) Identifying The Pixels
we assume that the two intensities
that present the impulse noise are the maximum
and the minimum
values of the image’s dynamic
range (i.e. 0 and L-
1). Thus, in this stage, at each
pixel location (x,y), we
find out the gradient value
G by using the following equation,
………(2)
Then
mark the mask α by using the
following equation,
International Conference on Recent Advances in Management,
Architecture,
Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management,
Architecture,
Technology and Engineering
(IJARMATE)
21
st
March 2016
ll Rights Reserved © 2015 ICRAMATE16
The histogram of the input image to
values that presents the impulse
noises. The identification is based on the peak
values contained in the histogram. He assumes that
the bright and dark intensity values of the impulse
noise produce two peaks in the histogram.
do not use this
assumption because this statement is not always
true, especially when the input is only corrupted by
Fig.1. schematic diagram of proposed
The steps needed for the noise
Identifying the pixels
(i.e. noise
free pixels and noise pixels).
Calculate the total number of
Roughly estimate the impulse
we assume that the two intensities
that present the impulse noise are the maximum
values of the image’s dynamic
1). Thus, in this stage, at each
find out the gradient value
………(2)
α by using the
where the value 1 presents the “noise pixel” and
value 0 presents the “noise-
free pixel”.
Ananth et al. [18] gave a brief outline on Electronic
devices and circuits which is the basis for
formation of patterns.
TABLE I
comparison of RMSE values of various filters over noise
Noise
percenta
ge
RMSE
for
mean
filter
RMSE
for
media
n filter
RMSE
for
switch
ed
media
n filter
5
17.36
34
8.2822
7.0523
10
21.98
82
9.3530
7.7588
20
31.58
57
14.637
9
10.579
30
40.01
70
23.465
2
13.702
40
48.77
2
36.336
7
17.661
50
51.14
71
52.806
9
22.057
60
65.32
83
72.095
4
33.28
70
74.36
37
95.030
3
47.911
80
82.46
69
115.73
00
65.167
90
90.38
4
135.95
33
88.009
95
94.12
60
146.05
32
99.561
99
98.05
25
154.52
63
107.34
2)
Calculate The Total Number Of Noise Pixels K
We calculate the total number of the
“noise pixel”, K. This is given by,
Here we use adaptive median filter,
that means the filter mask size adaptively
local noise level. So finding the value of K is need
for the further operation.
Technology and Engineering
Technology and Engineering
41
where the value 1 presents the “noise pixel” and
the
free pixel”.
Christo
Ananth et al. [18] gave a brief outline on Electronic
devices and circuits which is the basis for
comparison of RMSE values of various filters over noise
RMSE
for
switch
ed
media
n filter
RMSE
for
adapti
ve
media
n filter
7.0523
6.0345
7.7588
6.4382
10.579
7.7832
13.702
10.503
17.661
14.532
1
22.057
18.923
8
33.28
24.02
47.911
36.256
3
65.167
52.164
7
88.009
87.923
99.561
92.952
3
107.34
101.26
5
Calculate The Total Number Of Noise Pixels K
We calculate the total number of the
Here we use adaptive median filter,
that means the filter mask size adaptively
changes
local noise level. So finding the value of K is need
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
42
A
ll Rights Reserved © 2015 ICRAMATE16
3) Estimate The Impulse Noise Level
Using the value of K, we can
roughly estimate the impulse noise level η that
corrupts the image. The value of η is the ratio of
the “noise pixels” to the total number of pixels
contained in the image, as defined in the following
equation,
The value of η is in between 0 and 1
(i.e. 0 ≤ η ≤ 1). This value and the noise mask α
will be used in the following stage, which is for the
noise removal.
D. Noise cancellation
In this stage, we filter the input image f,
and produce the filtered image g. Similar to many
switching median filter methods, the output is
defined as:
g(x, y) = [1−α (x, y)] f (x, y) +α (x, y)m(x, y)
………(6)
where α is the noise mask, defined by (2) in Stage
1, where m is the median value obtained from our
adaptive method. The determination of m will be
explained later.
As α(x,y) only can take value of either 0 or 1, as
defined by (5) the output value g(x,y) is either
equal to f(x,y) or m(x,y). Thus, the calculation of
m(x,y) is only done when f(x,y) is a “noise pixel”
(i.e. α(x,y) = 1). For the “noise-free pixel” (i.e.
α(x,y) = 0), the value of f(x,y) is copied directly as
the value of g(x,y). This significantly speeds up the
process, because not all pixels need to be filtered.
Thus, alternatively, g(x,y) can be re-written as:
We use the adaptive methodology to
determine m(x,y). This means that the size of the
filter used at every pixel location is changing
accordingly to the local information. In this work,
we only consider the square filters with odd
dimensions for the filtering process, as given by,
W =WM =WN = 2R +1………….(8)
where R takes any positive integer value.
Our method, uses only “noise-free pixels” that are
contained in the contextual region, defined by the
area of W×W (i.e. the filter size), as the samples for
the calculation of m(x,y). This procedure ensures
that the value of g(x,y) will not be affected by the
noise, but be more biased towards the real data
values. To determine the value of m(x,y), in our
proposed method, we set a rule that the minimum
number of samples of “noise-free pixels” needed
for this calculation must be greater or equal to eight
pixels. Christo Ananth et al. [16] discussed about
E-plane and H-plane waveguides (Microwave
Engineering) applicable to image processing.
Christo Ananth et al. [17] discussed about
principles of electronic devices to explain this
feature. Our novel adaptive method for finding
m(x,y) is described by the following algorithm.
For each pixel location (x,y) with α(x,y)=1 (i.e.
“noise pixel”), do the following
1. Initialize the size of the filter W = 2
+1,
where
is a small integer value.
2. Compute the number of “noise-free pixels”
contained in the contextual region defined by this
W×W filter.
3. If the number of “noise-free pixels” is less than
eight pixels, increase the size of the filter by two
(i.e. W = W+2) and return to step 2.
4. Calculate the value of m(x,y) based on the
“noise-free pixels” contained in W×W window.
5. Update the value of g(x,y) using the equation(6
or 7)
IV.EXPERIMENTAL RESULTS
In order to demonstrate the
performance of our method, we also implemented
three other median filtering methods. As our
method is a hybrid of an adaptive median filter and
a switching median filter, we implement the mean
filter, median filter, switched median filter and
adaptive median filter.
In this work, we use 100 images of size
1600×1200 as our test images. These images
present a wide variety of images, and with different
characteristics. Example of these images are shown
in Fig. 2(a), and . These images are free from
impulse noise, and we denote them as e. Then, we
contaminate these images with impulse noise to get
the corrupted images, f. If image e is corrupted by
Q% of noise, 0.5Q% will be the positive impulses
and another 0.5Q% will be the negative impulses.
We then process f using the four mentioned
methods in this section. The results are presented in
terms of the visual appearance, root mean square
error (RMSE) and processing time.
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
43
A
ll Rights Reserved © 2015 ICRAMATE16
(a)
(b)
( c)
(d)
(e)
(f)
Fig. 2. “Coin” (a) The original image. (b) Image corrupted by 50% of impulse noise. (c) Result obtained
by using mean filter (RMSE = 51.1471) (d) Result obtained
by using median filter (RMSE = 52.8069) (e) Result obtained
by using switching median filter (RMSE = 22.057) (f) Result obtained
by using adaptive median filter (RMSE = 18.9238)
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
44
A
ll Rights Reserved © 2015 ICRAMATE16
V.CONCLUSION
One very important merit of this adaptive
MED filter is its edge-preserving characteristic. If
an image is contaminated with low or medium
density SPN, then a adaptive MED filter can do
justice by rejecting the outliers very easily. But at
some locations, the density of impulses could be
very high. At those points, the simple adaptive
MED filter fails. Even if the impulse noise density
in an image is very low, the adaptive MED can
never guarantee the true pixel replacement. The
adaptive MED operation changes the pixel value to
the median of the neighborhood unnecessarily even
if the center pixel is noise-free. Thus, there is some
unwanted error in the output.
REFERENCES
[1] Christo Ananth, G.Gayathri, M.Majitha Barvin, N.Juki
Parsana, M.Parvin Banu, “Image Segmentation by Multi-shape
GC-OAAM”, American Journal of Sustainable Cities and
Society (AJSCS), Vol. 1, Issue 3, January 2014, pp 274-280
[2] Christo Ananth, D.L.Roshni Bai , K.Renuka, C.Savithra,
A.Vidhya, “Interactive Automatic Hepatic Tumor CT Image
Segmentation”, International Journal of Emerging Research in
Management &Technology (IJERMT), Volume-3, Issue-1,
January 2014,pp 16-20
[3]Christo Ananth, D.L.Roshni Bai, K.Renuka, A.Vidhya,
C.Savithra, “Liver and Hepatic Tumor Segmentation in 3D CT
Images”, International Journal of Advanced Research in
Computer Engineering & Technology (IJARCET), Volume
3,Issue-2, February 2014,pp 496-503
[4] Christo Ananth, A.Sujitha Nandhini, A.Subha Shree,
S.V.Ramyaa, J.Princess, “Fobe Algorithm for Video
Processing”, International Journal of Advanced Research in
Electrical, Electronics and Instrumentation Engineering
(IJAREEIE), Vol. 3, Issue 3,March 2014 , pp 7569-7574
[5] Christo Ananth, Karthika.S, Shivangi Singh, Jennifer
Christa.J, Gracelyn Ida.I, “Graph Cutting Tumor Images”,
International Journal of Advanced Research in Computer
Science and Software Engineering (IJARCSSE), Volume 4,
Issue 3, March 2014,pp 309-314
[6] Christo Ananth, G.Gayathri, I.Uma Sankari, A.Vidhya,
P.Karthiga, “Automatic Image Segmentation method based on
ALG”, International Journal of Innovative Research in
Computer and Communication Engineering (IJIRCCE), Vol. 2,
Issue 4, April 2014,pp- 3716-3721
[7] Christo Ananth, A.S.Senthilkani, S.Kamala Gomathy,
J.Arockia Renilda, G.Blesslin Jebitha, Sankari @Saranya.S.,
“Color Image Segmentation using IMOWT with 2D Histogram
Grouping”, International Journal of Computer Science and
Mobile Computing (IJCSMC), Vol. 3, Issue. 5, May 2014, pp-1
– 7
[8] Christo Ananth, A.S.Senthilkani, Praghash.K, Chakka
Raja.M., Jerrin John, I.Annadurai, “Overlap Wavelet Transform
for Image Segmentation”, International Journal of Electronics
Communication and Computer Technology (IJECCT), Volume
4, Issue 3 (May 2014), pp-656-658
[9] Christo Ananth, S.Santhana Priya, S.Manisha, T.Ezhil Jothi,
M.S.Ramasubhaeswari, “CLG for Automatic Image
Segmentation”, International Journal of Electrical and
Electronics Research (IJEER), Vol. 2, Issue 3, Month: July -
September 2014, pp: 51-57
[10] Christo Ananth, R.Nikitha, C.K.Sankavi, H.Mehnaz,
N.Rajalakshmi, “High Resolution Image Reconstruction with
Smart Camera Network”, International J ournal of Advanced
Research in Biology, Ecology, Science and Technology
(IJARBEST), Volume 1,Issue 4,July 2015, pp:1-5
[11] Christo Ananth, B.Prem Kumar, M.Sai Suman, D.Paul
Samuel, V.Pillai Vishal Vadivel, Praghash.K., “Autonomous
Mobile Robot Navigation System”, International Journal of
Advanced Research in Biology, Ecology, Science and
Technology (IJARBEST), Volume 1,Issue 4,July 2015,pp:15-19
[12] Christo Ananth , Mersi Jesintha.R., Jeba Roslin.R., Sahaya
Nithya.S., Niveda V.C.Mani, Praghash.K., “Vision based Path
Planning and Tracking control using Mobile Robot”,
International Journal of Advanced Research in Biology,
Ecology, Science and Technology (IJARBEST), Volume 1,Issue
4,July 2015, pp:20-25
[13] Christo Ananth, S.Suryakala, I.V.Sushmitha Dani,
I.Shibiya Sherlin, S.Sheba Monic, A.Sushma Thavakumari,
“Vector Image Model to Object Boundary Detection in Noisy
Images”, International Journal of Advanced Research in
Management, Architecture, Technology and Engineering
(IJARMATE), Volume 1,Issue 2,September 2015, pp:13-15
[14] Christo Ananth,” Geo-cutting Liver Tumor”, International
Journal of Advanced Research in Management, Architecture,
Technology and Engineering (IJARMATE), Volume 2,Issue 3,
March 2016,pp:122-128
0 20 40 60 80 100
0
20
40
60
80
100
120
140
160
Noise Percentage
R M S E
Mean Filter
Median Filter
Switched MED FIlter
Adaptive MED filter
Comparison Chart
International Conference on Recent Advances in Management, Architecture, Technology and Engineering
(ICRAMATE’16)
Organized by
International Journal of Advanced Research in Management, Architecture, Technology and Engineering
(IJARMATE)
21
st
March 2016
45
A
ll Rights Reserved © 2015 ICRAMATE16
[15] Christo Ananth, S.Shafiqa Shalaysha, M.Vaishnavi, J.Sasi
Rabiyathul Sabena, A.P.L.Sangeetha, M.Santhi, “Realtime
Monitoring Of Cardiac Patients At Distance Using Tarang
Communication”, International Journal of Innovative Research
in Engineering & Science (IJIRES), Volume 9, Issue
3,September 2014,pp-15-20
[16] Christo Ananth, S.Esakki Rajavel, S.Allwin Devaraj,
M.Suresh Chinnathampy. "RF and Microwave Engineering
(Microwave Engineering)." (2014): 300,ACES Publishers
[17] Christo Ananth, S.Esakki Rajavel, S.Allwin Devaraj,
P.Kannan. "Electronic Devices." (2014): 300, ACES Publishers.
[18] Christo Ananth,W.Stalin Jacob,P.Jenifer Darling Rosita. "A
Brief Outline On ELECTRONIC DEVICES & CIRCUITS."
(2016): 300.
[19] Zhou Wang, and David Zhang, “Restoration of impulse
noise corrupted images using long-range correlation”, IEEE
Signal Processing Letters, vol. 5, no. 1, pp. 4-7, 1998.
