Iw3515281533

Navjot Kaur et al Int. Journal of Engineering Research and Application
ISSN : 2248-9622, Vol. 3, Issue 5, Sep-Oct 2013, pp.1528-1533

RESEARCH ARTICLE

www.ijera.com

OPEN ACCESS

An Efficient Edge Detection Approach Based On Pollination
Based Optimization
Navjot Kaur1, Parminder Singh2
1

Research Fellow, 2Asstt. Professor
Chandigarh Engineering College, Landran, Mohali(PB)

1,2

Abstract
Edge detection of pictures is a vital task in computer vision and image processing. Edge detection is always
study focus in the field of medical image processing and analysis. It is necessary step in medical image
processing. Edge detection of noise free pictures is comparatively less complicated, however in most sensible
cases the photographs area unit degraded by noise. Edges in photos provide low-level cues, which could be
utilized in higher level processes, like object detection, recognition, and classification, furthermore as motion
detection, image matching, and trailing. Edges and textures in image are typical samples of high-frequency
information. High-pass filters deduct low-frequency image information and therefore enhance high-frequency
information like edges. Many approaches to image interpretation measure supported edges. This paper proposed
an enhanced edge detection using Pollination based optimization (PBO) algorithm. In this, The samples of
medical images (MRI) with resolution 128×128 is given as input and output as edges of image is produced. All
images are gray scaled and we converted all samples to same size (128×128). In this firstly add speckle noise
then filter this image by using bilateral filter to make image noise free. A bilateral filter preserves sharp edges
by systematically looping through each pixel and adjusting weights to the adjacent pixels accordingly. It extends
the concept of Gaussian smoothing by weighting the filter coefficients with their corresponding relative pixel
intensities. Then we use PBO for edge detection. PBO based edge detection is a new technique and it perform
as well in medical field also and we used MRI images in our work.
Keywords— Edge Detection, Medical field, MRI images, PBO, Bilateral Filter.
I. INTRODUCTION
In many computer vision systems,
orientation and intensity information about edges in
images are used as inputs for further processing to
detect objects. Precise information about edges is
vital to the success of such systems [1]. Information
about edges is widely used in image segmentation,
image registration, image classification and pattern
recognition. Hence, detection of exact edges is a very
important part of image processing algorithms [2].
From an application-level view, an edge
detection algorithm is one which could be able to
provide continuous contours of the object boundaries.
However, the computations required to establish these
continuous contours would be very time consuming
and complex. From a pixel level view, the edges are
the areas of an image where the pixel intensities
undergo a sharp change. These areas shape the
contours which represent the boundary of objects.
Although many edge detection algorithms have been
proposed in the literature over the past three decades
to improve precision of recognized edges, they still
suffer from producing broken edges [3]. A noise
phenomenon is the most important obstacle to the
detection of continuous edges [4].
It causes some variation of pixel intensities
and accordingly reduces the performance of an edge
detection algorithm in noisy images. Another
www.ijera.com

important barrier which complicates the operation of
edge detection is illumination phenomenon which
causes the magnitude of the edges in the illuminated
areas to become weak [5]. Since most edge detection
algorithms utilise a thresholding technique to classify
a pixel as an edge or non-edge based on its
magnitude, a pixel with a weak magnitude may be
recognised as non-edge and accordingly the edges
become broken.
Traditional edge detection algorithms are
very fast but they cannot perform well on noisy
images and usually produce broken edges or noise
spots. Advanced edge detection algorithms, which
usually utilise soft computing techniques such as
neural networks and support vector machines for edge
detection, are highly problem-dependent and domain
specific [12].
II. EDGE DETECTION
Edge detection may be a vital space within
the field of pc Vision. Edges outline the boundaries
between regions in a picture that helps with
segmentation and beholding. They’ll show wherever
shadows fall in a picture or the other distinct
modification within the intensity of a picture. Edge
detection may be a basic of low level image process
and sensible edges are necessary for higher level
process. Edge detection refers to the method of
1528 | P a g e

characteristic and locating sharp discontinuities in a
picture. The discontinuities or abrupt changes in
element intensity characterize boundaries of objects
during a scene.
Edge Detection Techniques are classified as
follows: the primary order by-product of selection in
image process is that the gradient. The second order
derivatives of selection in image process are typically
computed exploitation Laplacian. For Sobel, a Prewitt
& Roberts technique performs finding edges by
thresholding the gradient for the log. By default edge
perform mechanically computes the edge to use. For
Sobel& Prewitt strategies, we are able to opt to
discover horizontal edges, vertical edges or each.
Laplacian of a Gaussian (LOG) finds edges by
searching for zero crossing once filtering with a
Gaussian filter. Zero crossing finds edges by
searching for Zero crossing once filtering with a userspecified filter. Clever finds by searching for native
maxima of the gradient. The gradient is calculated
exploitation the by-product of a Gaussian filter. The
strategy used 2 thresholds to discover sturdy & weak
edges, and includes the weak edges within the output
provided that they're connected to sturdy edges.
Therefore; this technique is a lot of doubtless to
discover true weak edges. Sobel edge detector
technique is somewhat tough than Prewitt edge
detector. Prewitt edge detector technique is slightly
easier to implement computationally than the Sobel
detector. However it tends to supply somewhat
noisier results. Parliamentarian edge detector is one
amongst the oldest & simplest edge detectors in
digital image process. It’s still used oftentimes in
hardware implementations wherever simplicity &
speed are dominant factors. This detector is employed
significantly but the others. Attributable to partly to
its restricted practicality. Log smoothes the image
(thus reducing noise) and it computes the Laplacian
that yields a double edge image. Zero crossing edge
detector supported same thought because the LOG
technique however the convolution, is disbursed
employing a nominal filter. Clever edge detector is
that the most powerful edge detector provided by
performs edge [2]. The disadvantages of clever edge
detector ar advanced computation, false zerocrossing and time overwhelming [12].
In the system represented in [5], they need
planned
a
unique
methodology
supported
mathematical logic reasoning for edge detection in
digital pictures while not determinant the edge worth.
The planned approach begins by segmenting the
photographs into regions exploitation floating 3x3
binary matrix. an instantaneous fuzzy illation system
mapped a spread of values distinct from one another
within the floating matrix to notice the sting by
exploitation eight planned rules.
In the system represented in [8], they need
planned a brand new edge detection technique
supported the BP neural network. They classified the
sting patterns of binary pictures into sixteen potential
www.ijera.com

www.ijera.com

varieties of visual patterns. Then once coaching the
pre-defined edge patterns, the BP neural network is
applied to corresponds any form of edges with its
connected visual pattern.
Edge detection refers to the method of
characteristic and locating sharp discontinuities in a
picture. The discontinuities area unit abrupt changes
in constituent intensity that characterize boundaries of
objects in a very scene. Classical ways of edge
detection involve convolving the image with associate
degree operator (a 2-D filter), that is made to be
sensitive to massive gradients within the image
whereas returning values of zero in uniform regions.
This can be a particularly sizable amount of edge
detection operators offered, every designed to be
sensitive to bound kinds of edges. Variables
concerned within the choice of a footing detection
operator include:
 Edge orientation: The pure mathematics of the
operator determines a characteristic direction
during which it's most sensitive to edges.
Operators are optimized to seem for horizontal,
vertical, or diagonal edges.
 Noise environment: Edge detection is
troublesome in screeching pictures, since each
the noise and therefore the edges contain highfrequency content. tries to cut back the noise lead
to blurred and distorted edges. Operators used on
screeching pictures area unit usually larger in
scope, in order that they will average enough
information to discount localized screeching
pixels. This leads to less correct localization of
the detected edges.
 Edge structure: Not all edges involve a step
amendment in intensity. Effects like refraction or
poor focus may end up in objects with boundaries
outlined by a gradual amendment in intensity.
The operator has to be chosen to be alert to such
a gradual amendment in those cases. Newer
wavelet-based techniques really characterize the
character of the transition for every draw near
order to tell apart, for instance, edges related to
hair from edges related to a face.
There is a unit many ways to perform edge
detection. However, the bulk of various ways is also
classified into 2 categories:
(i) Gradient: The gradient methodology detects the
sides by craving for the utmost and minimum within
the derivative of the image.
(ii) Laplacian: The Laplacian methodology searches
for zero crossings within the second by-product of the
image to search out edges. a footing has the onedimensional form of a ramp and hard the by-product
of the image will highlight its location.
Clearly, the by-product shows a most placed
at the centre of the sting within the original signal.
This methodology of locating a footing is
characteristic of the “gradient filter” family of edge
detection filters and includes the Sobel methodology.
A constituent location is said a footing location if the
1529 | P a g e

worth of the gradient exceeds some threshold. As
mentioned before, edges can have higher constituent
intensity values than those encompassing it. Therefore
once a threshold is about, you'll compare the gradient
price to the edge price and sight a footing whenever
the edge is exceeded. What is more, once the primary
by-product is at a most, the second by-product is zero.
As a result, another various to finding the placement
of a footing is to find the zeros within the second byproduct. This methodology is understood because the
Laplacian and therefore the second by-product of the
signal.
The purpose of selecting sharp changes in
image brightness is to capture vital events and
changes in properties of the planet. It is shown that
below rather general assumptions for a picture
formation model, discontinuities in image brightness
area unit seemingly to correspond to:
• Discontinuities thorough,
• Discontinuities in surface orientation,
• Changes in material properties and
• Variations in scene illumination.
III.



Comparison of PBO Results with the previous.

3.2 Basic Block Design
Edges in photos provide low-level cues,
which could be utilized in higher level processes, like
object detection, recognition, and classification,
furthermore as motion detection, image matching, and
trailing. Edges and textures in image are typical
samples of high-frequency information. High-pass
filters deduct low-frequency image information and
therefore enhance high-frequency information like
edges. Many approaches to image interpretation
measure supported edges. Since analysis supported
edge detection is insensitive to vary among the
illumination level.

Start

Upload Image

PROPOSED APPROACH USING PBO

It has been already explained that the Edges
are signiﬁcant local changes of intensity in an image.
Edge is the boundary between an object and the
background, and identifies the boundary between
overlapping and non-over lapping objects. This means
that if the edges in an image can be identified
accurately, all of the objects can be located and basic
properties such as area, perimeter, and shape can be
measured. Our first problem is to study the edge
detection by using different techniques on different
type of images and need to compare the results in
terms of PSNR, MSE, SSIM & EPI with the
implementation of PBO. In this proposed work we
use Gaussian filter, bilateral filter & trilateral filter to
remove noise. Pollination based optimization (PBO)
based Edge detection is a new technique and we
expect the results of compression to be far better in
comparison to the previous techniques.
3.1 Proposed Model
The proposed model focuses on the above
four objectives which are helpful in improving the
edge detection parameters and are practically
implemented using MATLAB 7.11.0 environment. In
this proposed work, we used Pollination based
optimization algorithm to optimize the results of edge
detection and to form a new technique for edge
detection using PBO. The objectives of our proposed
work are:
 Study of Edge Detection.
 Study of previous techniques for Edge detection.
 Pointing out the pros and cons of the previous
Algorithm.
 Study of PBO based Implementation for the same
purpose.
 Implementation of PBO based Edge Detection.
www.ijera.com

www.ijera.com

Add noise to Image

Noise Level
40

20

80

60

90

Apply Bilateral Filter

Pollination based
Optimization

End
Fig 1 Basic Block Design of Proposed Work
1530 | P a g e

Fig 1 shows the basic block design of the proposed
model. The uploaded image is provided with the noise
called speckle noise. We have also tested the code
with the increased and decreased level of noise. Like
we have increased the noise level by twenty, forty and
sixty, eighty and ninety percent of the noise level. By
default the taken noise level is ten percent. The added
noisy image is provided to the Bilateral Filter. Then
filtered image goes to PBO method for edge
detection.

www.ijera.com

Image Acquisition (MRI Images)

Add Noise (Speckle)

Remove Noise (Bilateral Filter)

3.3 Algorithm level Design
The algorithm design is shown in Fig 2,
which involves:
Step 1: Image Acquisition (MRI Images).

Initialize PBO Parameters

Step 2: Add Speckle Noise to all images.
Step 3: Remove Noise by using Bilateral Filter.
Step 4: Initialize PBO Parameters.
 a=1.2, A=0.9, D=1.2, N41.9, P=2,
 Number of Plants = 8,
 Number of weeks = 14,
 Number of seasons = 8, (number of iterations)
 Pollination weekly goal= [0.10 0.25 0.50 0.75
0.90 1.00]
Step 5: Randomly generate vectors.
 For season = 1 : number of seasons (iterations)
 For week = 1: number of weeks
 For k = 1: number of plants

Randomly generate Vectors

Evaluate Reproduction Vector
(R)

Based on R Evaluate Error

Update N, D, A

Step 6: Evaluate Reproduction Vector:

Step 7: Based on R, update number of seasons.


Error Acceptable?

Evaluate Error = Goal - R

Step 8: Based upon error update N, D, A
Step 9: Exit, if Error acceptable.

Edge Detected
Fig 2: Proposed Algorithm level Design using PBO
IV. RESULTS
The samples of medical images (MRI) with
resolution 128×128 is given as input and output as
edges of image is produced. All images are gray
scaled and we converted all samples to same size
(128×128). Results are shown as below:

www.ijera.com

1531 | P a g e


www.ijera.com

done as per the natural phenomenon of PBO. In this
we use Bilateral filter to remove noise. Then we use
PBO that is pollination based optimization algorithm.
This technique achieves better results.

REFERENCES
[1]

[2]

Fig 3: Sample Images
[3]

[4]

[5]
Fig 4: Images with speckle noise

[6]

Fig 5: Testing Images
[7]

[8]
Fig 6: Edge Detection using PBO
V. CONCLUSION
The detection of the edge is one of the
important part in the field of Image Processing. In this
paper we proposed an efficient PBO based algorithm
for images edge detection. The detection is basically
www.ijera.com

[9]

Gupta, A., Ganguly, A. ; Bhateja, V., “A
noise robust edge detector for color images
using
Hilbert
Transform”
Advance
Computing Conference (IACC), 2013 IEEE
3rd International, pp. 1207 – 1212, 2013.
Verma, Rohit ; Dabbagh, Mohamed-Yahia.,
“Binary pattern based edge detection for
motion
estimation
in
H.264/AVC”
Electrical and Computer Engineering
(CCECE), 2013 26th Annual IEEE Canadian
Conference , pp. 1-4, 2013.
Zhu, T., Xiao, X. ; He, Q. ; Diao, D.,
“Enhancement of SNR and Spatial
Resolution in -OTDR System by Using
Two-Dimensional Edge Detection Method”
Lightwave
Technology,
Journal
of
IEEE ,Volume: 31 , Issue: 17 , pp. 2851 –
2856, 2013.
Ricardo
Contreras,
M.
Angélica
Pinninghoff, Jaime Ortega., “Using Ant
Colony Optimization for Edge Detection in
Gray Scale Images” 5th International WorkConference on the Interplay Between
Natural
and
Artificial
Computation,
IWINAC 2013, Mallorca, Spain, June 10-14,
2013. Proceedings, Part I, pp. 323-331.
Junna Shang ; Hangzhou Dianzi Univ.,
Hangzhou, China ; Feng Jiang, “An
algorithm of edge detection based on soft
morphology” Signal Processing (ICSP),
2012 IEEE 11th International Conference
on signal processing, Volume:1, pp. 166169, 2012.
Rajeswari, R., and Rajesh, R., “A modified
ant colony optimization based approach for
image edge detection”, Image Information
Processing (ICIIP), 2011 International
Conference
on
image
information
processing, pp. 1-6, 2011.
Akansha Mehrotra, Krishna Kant Singh, M.
J. Nigam, “A Novel Algorithm for Impulse
Noise Removal and Edge Detection,”
International
Journal
of
Computer
Applications (0975 – 8887) Volume 38–
No.7, January 2012.
wagatam Das, Arijit Biswas, Sambarta
Dasgupta, and Ajith Abraham, “Bacterial
Foraging
Optimization
Algorithm:
Theoretical Foundations, Analysis, and
Applications”
M. Y. Jiang and D. F. Yuan “A multi-grade
mean morphologic edge detection” 6th
International
Conference
on
Signal

1532 | P a g e


[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

www.ijera.com

Processing Beijing, China, pp.1079-1082,
2002.
Hossein Nezamabadi-pour · Saeid Saryazdi
Esmat Rashedi, Edge detection using ant
algorithm”, in proc. of Springer-Verlag,
pp.623- 628, 2005.
Raymond H. Chan, Chung-Wa Ho, and Mila
Nikolova,“Salt-and-pepper noise removal by
median-type noise detectors and detailpreserving
regularization”,IEEE
Transactions on Image Processing, vol. 14,
no. 10, pp. 1479–1485, Oct. 2005.
X. Zhuang, “Edge Feature Extraction in
Digital Images with the Ant Colony System”
in proc. of the IEEE international
Conference an computational intelligence for
Measurement Systems and Applications, pp.
133-136,2004.
Hossein Nezamabadi-pour,Saeid Saryazdi
Esmat Rashedi, “Edge detection using ant
algorithms”, in proc. of Springer-Verlag,
pp.623-628, 2005.
Feng-ying Cui ,Li-jun Zou and Bei Song ,
“Edge Feature Extraction Based on digital
Image processing techniques,” Proc. IEEE
International conference Automation and
logistics, Qingdao,China September 2008.
M.Y.Jiang,and D.F.Yuan “A multi-grade
mean morphologic edge detection” 6th
International
Conference
on
Signal
Processing Beijing, China, pp.1079-1082,
2002.
Hossein Nezamabadi-pour · Saeid Saryazdi
Esmat Rashedi, Edge detection using ant
algorithm”, in proc. of Springer-Verlag,
pp.623- 628, 2005.
Raymond H. Chan, Chung-Wa Ho, and Mila
Nikolova,“Salt-and-pepper noise removal by
median-type noise detectors and detailpreserving
regularization”,IEEE
Transactions on Image Processing, vol. 14,
no. 10, pp. 1479–1485, Oct. 2005.
X. Zhuang, “Edge Feature Extraction in
Digital Images with the Ant Colony System”
in proc. of the IEEE international
Conference an computational intelligence for
Measurement Systems and Applications, pp.
133-136,2004.

www.ijera.com

1533 | P a g e

Iw3515281533

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Iw3515281533 (20)

Recently uploaded (20)

Iw3515281533