B. Govinda Lakshmi Int. Journal of Engineering Research and Applications www.ijera.

com
ISSN : 2248-9622, Vol. 4, Issue 9( Version 1), September 2014, pp.165-167

RESEARCH ARTICLE OPEN ACCESS

Edge Detection Using Fuzzy Logic

B. Govinda Lakshmi#1, B. Hemalatha#2
#
Sri Sivani College of Engineering, Department of Computer Science Srikakulam, Andhrapradesh,, India

Abstract
Edge detection is still difficult task in the image processing field. In this paper we implemented fuzzy
techniques for detecting edges in the image. This algorithm also works for medical images. In this paper we also
explained about Fuzzy inference system, which is more robust to contrast and lighting variations.
Keywords- Fuzzy, FIS.

I. INTRODUCTION of the image to find edges. The original image can be

A. Criteria for Edge Detection easily restored from its edges.
Several algorithms are existed for edge detection.
Each algorithm is having its own drawbacks. A good Two main methods:-
edge detection algorithm should satisfy all of the a) Gradient-based method: Gradient-based methods
below conditions. detect edges by checking for maxima and minima in
a) Good detection: There should be maximum the first derivative of the image.
number of good edges.
b) Noise sensitivity: Edge detection algorithms b) Laplacian (zero-crossing) based method: The
should be able to detect edges with acceptable Laplacian based methods search for zero crossings in
noise. the second derivative of the image in order to find
c) Good localization: The edge location must be edges usually the zero-crossings of the Laplacian or
reported as close as possible to the correct the zero-crossings of a non-linear differential
position, i.e. edge localization accuracy (ELA). expression. A number of edge detection techniques
d) Orientation sensitivity: The operator not only are available but there is no single detection method
detects edge magnitude, but italso detects edge that performs well in every possible image context.
orientation correctly. Various edge detection techniques are used for edge
e) Speed and efficiency: The algorithm should be detection like Canny edge detection, Krisch edge
fast enough to be usable in animage processing detection which are applied on various images.
system. An algorithm that allows recursive Choice of edge detector to be used depends upon the
implementationor separately processing can image properties like noise sensitivity, orientation
greatly improve efficiency. sensitivity, speed and efficiency.

B. Various Techniques for Edge Detection C. Brief review of Fuzzy Logic

Edge detection algorithms preserves most Fuzzy logic is a perfect problem-solving
important information in an image. Hence the output. methodology with a myriad of applications in
The original image can be easily restored from its embedded control and information processing [2].
edge map. Various edge detection algorithms have Fuzzy provides a remarkably simple way to draw
been developed in the process of finding the perfect definite conclusions from vague, ambiguous or
edge detector. imprecise information. It means that fuzzy logic
Most of techniques may be grouped into two resembles human decision making with its ability to
types such as gradient based edge detection and work from approximate data and find precise
Laplacian-based edge detection. In the gradient based solutions. Fuzzy logic and probability theory are the
edge detection we find an estimate of the gradient most powerful tools to overcome the imperfection.
magnitude using the smoothing filter and use the Fuzzy logic may appear similar to probability and
found estimate to determine the position of the edges. statistics as well. Even though fuzzy logic is different
It means that the gradient method detects the edges from probability the result appears to be same. The
by looking for the maximum and the minimum in the probability statement, there is a 70% chance that
first derivative of the image. In the Laplacian method Bill is tall supposes that Bill is either tall or he is not.
we find the second derivative of the signal and the There is 70% chance that we know which set Bill
derivative magnitude is maximum when second belongs. The fuzzy logic statement, Bills degree of
derivative is zero. In short, Laplacian method membership in the set of tall people is .80 supposes
searches for zero crossings in the second derivative that Bill is rather tall. The fuzzy logic answer
www.ijera.com 165 | P a g e
B. Govinda Lakshmi Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 4, Issue 9( Version 1), September 2014, pp.165-167

determines he is a member. There are no probability image filtering technique is presented which is based
statements that pertain to fuzzy logic. Fuzzy logic on fuzzy inference systems (FIS).
deals with the degree of membership. First, an input image is processed in different
Fuzzy image processing has three main phases: non-successive linear filtering stages, which means
image fuzzification, modification of membership that the input to each filter is always the original
values and if necessary, image defuzzification. The image. The gray level in each pixel of the resulting
fuzzification and defuzzification steps are due to the image is then obtained by applying the FIS system to
fact that we do not possess fuzzy hardware. Thus the the corresponding values in the output images of the
coding of image data (fuzzification) and decoding of linear operators, in the same pixel. And evaluate the
the results (defuzzification) are steps that make efficiency of a FIS system applied to the edge
possible to process the images with fuzzy techniques. detection problem.
The main power of fuzzy image processing is in the During input image pre-processing, three kinds
middle step (modification of membership values). of linear filters are applied to it: Sobel operators, used
After the image data is transformed from gray-level to estimate its derivatives in horizontal and vertical
plane to the membership plane (fuzzification), directions (hDH and hDV filters), a low-pass (mean)
appropriate fuzzy techniques modify the membership filter and a high-pass filter. The developed fuzzy
values. This can be a fuzzy clustering, a fuzzy rule- systems purpose is to determine if pixel evaluated is
based approach, and fuzzy integration approach. or is not present in one of images edges.
Fuzzy image processing is important to represent An easy way to comply with the conference
uncertainty in data. Some of the main benefits of paper formatting requirements is to use this document
fuzzy image processing are given below: as a template and simply type your text into it.
Fuzzy techniques are powerful tools for
knowledge representation and processing A. Methodology
Fuzzy techniques can manage the vagueness and Step1: Form 16 edge-detected templates with values
ambiguity efficiently. a, b.
Fuzzy logic is tolerant of imprecise data. Step2: Apply the edge templates over the image by
Fuzzy logic is conceptually easy to understand. placing the centre of each template at each point (i, j)
The mathematical concepts behind fuzzy over the normalized image.
reasoning are very simple. What makes fuzzy nice is Step3: Calculate the intuitionistic fuzzy divergence
the naturalness of its approach and notits far- (IFD) between each elements of each template and
reaching complexity. In many image processing the image window (same size as that of template) and
applications, expert knowledge is used to overcome choose the minimum IFD value.
the difficulties (e.g. object recognition, scene Step4: Choose the maximum of all the 16 (total no. of
analysis). Fuzzy set theory and fuzzy logic offer templates) minimum intuitionistic fuzzy divergence
powerful tools to represent and process human values.
knowledge in form of fuzzy if-then rules. Step5: Position the maximum value at the point
On the other side, many difficulties in image where the template was centered over the image.
processing arise because the data/tasks/results are Step6: For all the pixel positions (considering the
uncertain. This uncertainty, however, is not always border pixels by taking the mirror values of the
due to the randomness but to the ambiguity and image), the maxmin value has been selected and
vagueness. Beside randomness which can be positioned.
managed by probability theory, imperfection in the Step7: A new intuitionistic divergence matrix has
image processing can be distinguished into three been formed.
types as follows: Step8: Threshold the intuitionistic divergence matrix
Grayness ambiguity and thin.
Geometrical fuzziness Step9: An edge-detected image is obtained.
Vague (complex/ill-defined) knowledge these
problems are fuzzy in the nature. III. RESULTS
Lung image of size 189 _ 189 shown in (a).
II. PROPOSED WORK The hesitation constant for image is ci = 0.05. Edge-
Edge detection using fuzzy logic provides an detected results with varying values of ct are shown.
alternative approach to detect edges. First-order With c= 0.3and above, in (b)(d) are giving a better
linear filters are mostly applied for edge detection in result where the outer edges are clearly detected and
digital images. Nevertheless they dont allow good no false edges are shown. Fig. (f) shows, the result
results to be obtained from images where the contrast with different values ofa = 0.25, b = 0.7 where the
varies a lot, due to non-uniform lighting. But in this edges are not properly detected.
research fuzzy inference system is applied for edge
detection to improve the performance. A non-linear

www.ijera.com 166 | P a g e
B. Govinda Lakshmi Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 4, Issue 9( Version 1), September 2014, pp.165-167

[2] Hamid R. Tizhoosh, Fuzzy Image

Processing. www.pami.uwaterloo.ca/tizho
osh/edge.html
[3] Md. Shoiab Bhuiyan, Yuiji Iwahori, and
Akira Iwata. Optimal edge detection under
difficult imaging conditions. Technical
report, Educational Center for Information
Processing and Dept. of Electrical and
Computer Engineering,Nagoya Institute of
Technology, Showa, Nagoya, 466-8555,
JAPAN, URL:-http://www.center.nitech.ac.
jp/people/bhuiyan /pub.html.
[4] Ursula Kretschmer, Maria S. Orozco, Oscar E.
Ruiz and Uwe Jasnoch, Edge and corner
identification for tracking the line of sight.
[5] Renyan Zhang, Guoling Zhao, Li Su. New
edge detection method in imageprocessing.
College of Autom., Harbin Engineering
(a) Lung Image University, China.
(b) c = 0.2 th= 0.9 [6] Zhengquan He, M.Y.Siyal. An image
(c) c = 0.3 th=0.20 detection techniquebased onmorphological
(d) c=0.35th=0.20 edge detection and background differencing
(e) c=0.4 th= 0.27 for real-timetraffic analysis. Elsevier
(f) c=0.45 th= 0.29 Science Inc. New York, NY, USA
[7] Stamatia Giannarou, Tania Stathaki. Edge
IV. CONCLUSION Detection Using QuantitativeCombination of
This algorithm is adaptable to various Multiple Operators. Communications and
environments. The weights associated with each SignalProcessing Group, Imperial College
fuzzy rule were tuned to allow good results to be LondonExhibition Road, SW7 2AZLondon,
obtained while extracting edges of the image, where UK.
contrast varies a lot from one region to another. [8] M. Hanmandlu, Rohan Raj Kalra, Vamsi
During the performance tests, however, all Krishna Madasu, Shantaram Vasikarla.Area
parameters were kept constant. based Novel Approach for Fuzzy Edge
The results allow us to conclude that, the Detection. Department ofElectrical
implemented FIS system presents greater robustness Engineering, Indian Institute of Technology,
to contrast and lighting variations, besides avoiding Delhi.
obtaining double edges. [9] Hamid R. Ezhoosh, Fast Fuzzy Edge
Detection. Pattern Recognition andMachine
Intelligence Lab Systems Design Engineering,
V. FUTURE WORK
University of WaterlooWaterloo, Ontario,
In fact, the proposed technique is to find more
N2L 3G1, Canada.
fine edges using fuzzy logic technique. In future,
[10] Madasu Hanmandlu, John See, Shantaram
modification of fuzzy rules can produce better result.
Vasikarla. Fast fuzzy edgedetector using
Further tuning of the weights associated to the fuzzy
entropy optimization. Dept. of Electrical,
inference rules is still necessary to reduce even more
Engineering I.I.T.Delhi, Faculty of
inclusion in the output image of pixels not belonging
Engineering, Multimedia University,
to edges. Our proposed technique is restricted only to
Cyberjaya Selangor D.E.63100, Malaysia, IT
gray scale images, this can be extended to colour
Department, Intercontinental University Los
images in that case, and the detection would become
Angeles, CA90066, USA.
significantly more
[11] Jinbo Wu, Zhouping Yin, and Youlun Xiong,
complex.
The Fast Multilevel FuzzyEdge Detection of
Blurry Images.
REFERENCES [12] Cristiano Jacques Miosso, Adolfo
[1] Ng Geok See and Chan Khue hiang. Edge
Bauchspiess. Fuzzy Inference System
Detection using supervised Learning and
Applied to Edge Detection in Digital Images.
voting scheme. Nanyang Technological
Departamento de Engenharia Electrical
University, National university of Singapore,
Faculdade de Tecnologia Universidade de
Singapore.
Braslia UnB.

www.ijera.com 167 | P a g e