Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

Simplified Novel Method for Edge Detection in

Digital Images
Mrityunjay Kumar Ray Deboleena Mitra Sankhadip Saha
Department of Electrical Engineering Department of Electrical Engineering Department of Electrical Engineering
Netaji Subhash Engineering College Netaji Subhash Engineering College Netaji Subhash Engineering College
Kolkata, India Kolkata, India Kolkata, India
[email protected] [email protected] [email protected]

Abstract— Edge detection techniques are widely used in image multiple segments. Image segmentation is typically used to
processing for object detection based on segmentation of images locate objects and boundaries (lines, curves, etc.) in images
on local changes in intensity. Here a novel method is presented [4]. Segmentation accuracy determines the eventual success or
that detects edges in an image with more accuracy and precision. failure of computerized analysis procedures. Most of the
The proposed technique locates the edge by comparing the pixels
segmentation algorithms are based on one or two basic
in the entire image using a 3x3 binary matrix. A comparative
study of different edge detection techniques like Sobel, Canny, properties of intensity values: discontinuity and similarity. In
Prewitt, Roberts with the proposed technique is presented. The the first category, the approach is to partition an image based
new technique gives better results in terms of smoothness, on abrupt change in intensity such as edges. The second
sharpness with the execution time much faster than the technique is based on the threshold and the region growing.
conventional edge detection techniques. Our technique is based on the first approach.

III. EDGE DETECTION TECHNIQUES

Keywords— Edge detection, Digital Image Processing, Image Edge detection is by far the most common approach for
Segmentation, Thresholding, Execution Time. detecting meaningful discontinuities in intensity values. Edge
models are classified according to the intensity profiles. The
I. INTRODUCTION
edge is the boundary between two different regions with
Edge detection technique is one of the most commonly used relatively distinct grey level properties. It is assumed that the
object detection method in image analysis. An edge is defined transition between these two regions can be determined on the
by a discontinuity in gray level values. In other words, an basis of grey level discontinuity.
edge is the boundary between an object and the background
[15]. The shape of edges in an image depends on many The choice for finding edge strength and direction at
parameters: The geometrical and optical properties of the location (x, y) of an image, f, is the gradient denoted by ,
object, the illumination conditions, and the noise level in the and defined as the vector:
images [14]. Many edge detection algorithms have been
developed based on computation of the intensity gradient
vector. Classical edge operators like Sobel, Prewitt, Roberts (1)
and Canny are based on the gradient of image calculated using
first order derivation [1, 15]. In many cases these operators are
not suitable for sharp detection of different geometrical shapes The magnitude of vector , denoted as , , is
in images. Moreover, the visible clarity of hidden alphabets in
images is not well detected. Some advanced techniques based , (2)
on soft computing like fuzzy logic [11, 12] give better results
than the conventional techniques but takes longer execution This gives the value of the rate of change in the direction of
time. Our method gives better results as compared to the gradient vector. The direction of the gradient vector is
conventional techniques with less computational time. given by the angle,

, tan (3)
II. IMAGE SEGMENTATION
Segmentation subdivides an image into its constituent
regions or objects. In computer vision, segmentation refers to
the process of partitioning a digital image into

978-1-61284-653-8/11/$26.00 ©2011 IEEE 197

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

A. Roberts Edge Detector IV. SIMPLIFIED NOVEL METHOD FOR EDGE DETECTION
The Roberts operator performs a simple, quick to compute, This method is based on the first derivative of the gradient
2-D spatial gradient measurement on an image [5]. It thus vector. The image is at first converted to a binary image
highlights regions of high spatial gradient which often using global thresholding. Thresholding is the simplest
correspond to edges. In its most common usage, the input to method of image segmentation. Global thresholding method
the operator is a grayscale image, as is the output. Pixel values maximizes the between-class variance. Otsu’s method is a
at each point in the output represent the estimated absolute special case of global thresholding which is based entirely
magnitude of the spatial gradient of the input image at that on computations performed on the histogram of an image,
point. In theory the operator consists of a pair of 2×2 an easily obtainable 1-D array. Then a 3X3 binary matrix
convolution kernels as shown below. with query point at the centre is taken. The method is used
to decide whether the query point is a part of edge or not.
1 0 0 1 The matrix spans across the image from one pixel to another
And (4)
0 1 1 0 and so on. It compares the conditions given in the flowchart,
if any one of the conditions is satisfied then that pixel is
B. Sobel Edge Detector decided as an edge. Decision of edge does not depend on the
method of scanning so no other intelligent method is
In case of Sobel operator computation of the partial required to decide the scanning technique for better edge
derivation in gradient may be approximated in digital images detection. The technique uses first order derivative of
[6] .It is based on convolving the image with a small, gradient. The magnitude and the direction of the gradient are
separable, and integer valued filter in horizontal and vertical given as:
direction and is therefore relatively inexpensive in terms of
computations. On the other hand, the gradient approximation
which it produces is relatively crude. Mathematically, the ,
operator uses two 3×3 kernels which are convolved with the
original image to calculate approximations of the derivatives.
, tan (7)
1 1 1 1 0 1
0 0 0 And 2 0 2 (5)
1 2 3 1 0 3 An empirical design procedure of the method can be
adopted. The flowchart is given in Fig. 1.
C. Prewitt Edge Detector
Prewitt edge detection obtains the orientation directly from
the kernel with the maximum response [7]. Unlike the Sobel
operator this operator does not place any emphasis on pixels
that are closer to the centre of the matrix and . The
Prewitt operator measures two components. The vertical edge
component is calculated with kernel and the horizontal
edge component is calculated with kernel .

1 0 1 1 1 1
1 0 1 And 0 0 0 (6)
1 0 1 1 1 1
D. Canny Edge Detector
The Canny edge detection operator uses a multi-stage
algorithm to detect a wide range of edges particularly for the
noisy images [8]. In Canny Edge Operator calculus of
variations is used. It is a technique which finds the function
which optimizes a given functional. The optimal function in
Canny's detector can be approximated by the first derivative
of a Gaussian. The effect of the canny operator is determined
by three parameters i.e. the width of the Gaussian kernel used
in the smoothing phase, the lower and upper thresholds used
by the tracker. The upper tracking canny edge detector is the
most powerful edge detector. Here the image is smoothed
using a Gaussian filter with a specified standard deviation to
reduce noise. The local gradient and edge direction are
computed at each point.

Fig. 1 Flowchart of our proposed method

978-1-61284-653-8/11/$26.00 ©2011 IEEE 198

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

V. EXPERIMENTAL VERIFICATIONS

Fig. 2 (a) Original image of chess board

b Fig.2 (b) The edge identified by Robberts Edge Detector

Fig.2 (c) The edge identified by Sobel Edge

E Detector Fig. 2 (d) The edges identified by Prrewitt Edge Detector

Fig. 2 (e) The edge identified by Canny Edge Detector Fig.2 (f) The edge identified by ouur proposed method

978-1-61284-653-8/11/$26.00 ©2011 IEEE 199

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

Fig.3 (a) Original image of gear tooth edges Fig.3 (b) The edge identified by Roberts Edge Detector

Fig.3 (c) The edge identified by Sobel Edge Detector Fig.3. (d) The edge identified by Prewitt Edge Detector

Fig. 3 (e) The edge identified by Canny Edge Detector Fig. 3 (f) The edge identified by our proposed method

978-1-61284-653-8/11/$26.00 ©2011 IEEE 200

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

Fig.4 (a) Original image of differrent polygons Fig.4 (b) The edge identified by Roberts
R Edge Detector

Fig.4 (c) The edge identified by Sobel Edge Detector Fig.4 (d) The edge identified by Prewitt
P Edge Detector

Fig.4 (e) The edge identified by Cannny Edge Detector Fig.4 (f) The edge identified byy our proposed method

978-1-61284-653-8/11/$26.00 ©2011 IEEE 201

Proceedings of 2011 International Conference on Signal Processing, Communication, Computing and Networking Technologies (ICSCCN 2011)

The proposed technique is tested with 3 different sets of digital image processing plays an important role as in robotics
images, and its performance is compared with the classical where fast processing of image is necessary. The proposed
operators namely Sobel, Canny, Roberts and Prewitt. technique is tested and compared in MATLAB environment.
An automatic thresholding method can be devised in future
Fig. 2 shows that our proposed technique gives continuous which would minimize the noise during edge detection for
outer boundary which the conventional edge operators fail to online image processing.
detect. The inner edges are also clearly visible with precise
continuity. While the classical operators fail to detect the ACKNOWLEDGMENT
numbers in the image but the numbers are clearly visible with The authors acknowledge the support from the ‘Control and
the propose method. Automation Centre’ of Netaji Subhash Engineering College.
In Fig. 3 it is seen that the image of gear tooth with the REFERENCES
proposed technique yields smoother boundary and less noisy [1] M. Sharifi, M. Fathy, M. T. Mahmoudi, “A Classified and Comparative
flat areas. It also gives sharper result in the edgy regions than Study of Edge Detection Algorithm”, Proceedings of the International
the conventional operators. Conference on Information Technology: Coding and Computing
(ITCC’02), IEEE, 2002.
Fig. 4 shows that that the corners in the image are [2] M. Roushdy, “Comparative Study of Edge Detection Algorithms
Applying on the Grayscale Noisy Image Morphological Filter”, GVIP
discontinuous and blunt with the classical operators. But with Journal, Volume 6, Issue 4, December, 2006.
the proposed method the corners are sharp, continuous. [3] C. Li, K. He, J. Zhou, “Edge Detection of Image on the Local Feature”,
Moreover the edges are smoother with the proposed novel Second International Symposium on Intelligent Information Technology
method. The hidden text is also clearly visible in the proposed Application, IEEE, 2008.
[4] S.S. Al-amri, N.V. Kalyankar, S.D Khamitkar, “Image Segmentation
novel method whereas with the classical edge operators the By Using Edge Detection”, International Journal on Computer Science
letters are dull and blurred. and Engineering, Vol. 02, No. 03, 2010, pp.804-807.
[5] Roberts, L. G. [1965], “Machine Perception of Three-Dimensional
Moreover the Execution Time of the proposed method is Solids”, in Optical and Electro-optical Information Processing, Tipped,
J.T.(ed.), MIT Press, Cambridge, Mass.
much faster when compared with the traditional operators. [6] Sobel, I.E. [1970]. “Camera Models and Machine Perception,” Ph.D.
The simulation results are given in the TABLE I. dissertation, Stanford University, Palo Alto, Calif.
[7] Prewitt, J. M. S. [1970]. “Object Enhancement and Extraction,” in
TABLE I. EXECUTION TIME Picture Processing and Psychopictorics, Lipkin, B. S., and Rosenfeld, A.
(eds.), Academic Press, New York.
[8] Canny, J., “A Computational Approach to Edge Detection”, IEEE
Figure Execution Time (in second) Trans. Pattern Analysis and Machine Intelligence, 8:679-714, November
No. Robert Sobel Prewitt Canny Proposed
1986.
Method
[9] Feng, X. Wang, Y. Liu, “An Edge Detection Method for Infrared Image
2 0.3218 0.3438 0.8281 1.1094 0.1406
Based on Grey Relational Analysis”, 2008.
3 0.1563 0.1406 0.1563 0.1875 0.0781 [10] H. G. Kaganami, Z. Beiji, “Region-Based Segmentation versus Edge
4 0.0781 0.1094 0.0938 0.2031 0.0469 Detection”, Fifth International Conference on Intelligent Information
Hiding and Multimedia Signal Processing, IEEE, 2009.
[11] S. Mathur, A. Ahlawat, “Application of Fuzzy Logic on Image Edge
VI. CONCLUSION AND FUTURE WORK Detection”, Intelligent Information and Engineering Systems”, INFOS
2008,
The classical edge operators suffer from various [12] A.A.Alshennawy, A.A.Aly,”Edge Detection in Digital Images Using
shortcomings which are minimized using our proposed Fuzzy Logic Technique”, World Academy of Science, Engineering and
method that offers greater accuracy with less computational Technology 51 2009.
[13] I.M. Elewa, H.H Soliman and A.A. Alshennawy.”Computer vision
complexity. Hence edge detection is simplified to a large Methodology for measurement and Inspection: Metrology in Production
extent. Reduction of noise with sharpened and well defined area“. Mansoura Eng. First conf. Faculty of Eng.Mansoura Univ., March
corners, smoother edge boundary and better feature detection 28-30, 1995, pp. 473-444.
makes this method versatile and robust. Though simple the [14] H. Chidiac, D. Ziou, “Classification of Image Edges”, Vision Interface’
99, Troise-Riviers, Canada, 1999, pp. 17-24.
method offers best result among the classical operators. The [15] Gonzalez R., Woods R. “Digital Image Processing” 2/E, Prentice Hall
method can be useful in a wide variety of applications where Publisher, 2002.

978-1-61284-653-8/11/$26.00 ©2011 IEEE 202