Volume 4, Issue 11, November – 2019 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Edge Detection Techniques on Digital Images -

A Review
J. S. Owotogbe*1,T. S. Ibiyemi2, and B. A. Adu3
2
Vice chancellor office,
1, 3
Dept. of Mathematical Sciences,
Achievers University, Owo, Nigeria

Abstract:- Edge is known to be an important changes of

intensity in a digital image. It is a sudden changes of
discontinuous noticed in an image. The three types of
edges are: Horizontal, Vertical and digital images. In this
paper, different edge detection methods are reviewed and
comparing of different edge detection method, with their
advantages and disadvantages. Implementation was done
using MATLAB. Fig 1a :- Step edge

Keywords:- Edge Detection, Prewitt Operator, Sobel

Operator, Log Operator, Canny Operator.

I. INTRODUCTION

Edges are very important in computer images analysis as

it is known as change of intensity in a digital image. It is also a
sudden change of discontinuities noticed in an image. Edges Fig 1b :- Line edge
show the boundaries that exist between regions in a digital
image. They are known to also be signs of lack of continuity
in a digital image[13]. This helps with segmenting a part of an
image and recognition of an object[10]. It shows important
variations. Edges are significant information, since it is noted
that they correspond to geometric or physical variations in
scene object. The physical variations such as illumination and
surface reflections notable changes in physical areas show
themselves in several ways such as texture, color, and Fig 1c :- Ramp edge
intensity. Edge detection is known as rudimentary of low –
level image processing and it is noted that higher level
processing requires good edges[7]. It is one of the techniques
that are frequently used in digital image processing[3,4].

II. EDGE DETECTION OVERVIEW

Edge detection is known as powerful tool frequently used Fig 1d :- Roof edge
in image processing. Its applications are seen in many areas
such as pattern recognition, motion analysis and objet To segment the non – trivial images which seems to be
recognition. Edge detection make use of edged which are sometimes complicated [5]. Some factors needed to be taken
present in an image[17]. Based on difference of grey level or into consideration are edge-like features, Noise, image
intensity of an image, the following are the types of edges as brightness and corners. Image segmentation is based on two
shown in Fig 1a – 1d. major characteristics features of intensity values which are (i)
Discontinuity which involves partitioning of an image based
on sudden changes in intensity and (ii) Similarity which
involves partitioning of an image area that look alike or
similar based on some predefined criteria. Digital image has

IJISRT19NOV223 www.ijisrt.com 329

Volume 4, Issue 11, November – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
three types of discontinuities which are edges, lines and two masks or kernels as shown in Fig. 3a and 3b to
points. Since it is said that edges encapsulate changes in the digitally calculate the first derivative Gx and Gy
intensity of an image[6][15]. The causes of this intensity
changes are physical events, such as -1 0 1
Gx
(a) Illumination discontinuity e.g shadow
-1 0 1
(b) Discontinuity of surface orientation
(c) Depth discontinuity which occurs when surface orientation -1 0 1
is perpendicular to one’s line of sight
Fig. 3a:- Gx(Vertical mask of sobel operator)
(d) Surface reflectance discontinuity which is when there is a
change in the degree of light that is incident on the surface
that is reflected to the viewer -1 0 1
Gy -2 0 2
The following are the descriptors of edge
(a) Edge strength: This works by tracing the local image -1 0 1
contrast along the normal Fig. 3b :- Gy(Horizontal mask of sobel operator)
(b) Edge normal: This is when unit vectors is in the direction
of maximum intensity change (2) Prewitt operator: This operator is very much similar to the
(c) Edge position: This is the image position in which we can sobel operator which also detect vertical and horizontal edges
use to locate the edge of an image[9-12]. This detector is one of the best ways to
(d) Edge direction: This is when unit vector is perpendicular to detect the orientation and magnitiude of an image. It uses the
edge normal as shown in Fig. 2 mask in Fig 4a and Fig 4b respectively.

Gx -1 -1 -1
0 0 0
1 1 1
Fig. 4a :- Gx(Vertical mask of prewitt operator)

Gy -1 0 1
-1 0 1
-1 0 1
Fig. 4b:- Gy(Horizontal mask of prewitt operator)

Fig. 2:- Edge direction and edge normal (3) Robert operator: This gradient operator calculates the
addition of squares of the differences between adjacent pixels
III. VARIOUS EDGE DETECTION at the diagonal in an image via discrete differentiation. Then
the gradient approximation is made. It uses 2 x 2 kernels as
Edge detection techniques are divided into two main shown in Fig. 5a and Fig 5b. respectively.
categories which are: (i) Gradient which helps to compute first
order derivations in an image and (ii) Gaussian – based which Gx 1 0
helps to compute second order derivations in an 0 -1
image.[8][14][16], and also shown in Table 1.
Fig. 5a :- Gx(Vertical mask of robert operator)
Gradient - based Sobel operator, Prewitt operator, Robert
operator Gy 0 1
Gaussian - based Canny edge detector, Laplacian of -1 0
Gaussian
Fig. 5b :- Gy(Horizontal mask of robert operator)
Table 1:- Edge detection techniques categories
(4) Canny operator:This is a robust operation which helps in
(1) Sobel operator: This is known as a discrete differentiation detecting edges unlike sobel and Robert operators, this
operator which calculate gradient approximation of image operator is not susceptible to noise. It can work in different
intensity function for image edge detection[1]. Among the environments. It is one of the best methods because it extracts
pixels of the image, Sobel operator produces either the an image features without affecting or altering the feature. It
normal to a vector or gradient vector. It makes use of 3 x 3 works based on three criteria which are: (i)Low error rate (ii)

IJISRT19NOV223 www.ijisrt.com 330

Volume 4, Issue 11, November – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Edge points must be accurately localized (iii) There should be
just one single edge response[4]

(5) Laplacian of Gaussian: Laplacian of Gaussian which was

proposed by [2] is also referred to as LoG. This really works
well when transition of the grey level seems to be abrupt. In
order to have a smooth transition, it is advisable to calculate
the second- order derivative. It is important to note when the
second order derivative crosses zero, which means that the Fig 6c :- Prewitt Operator
location corresponds to a maximum level. This location is
called edge location. The LoG of a digital image f(x,y) which
is a second order derivative is as shown in equation 3

IV. IMPLEMENTATION

Edge detection was implemented on the image as

displayed in Fig. (6a -6f) using various techniques. This was
performed using Matlab 8.0 (2013a)
Fig 6d :- Roberts Operator

Fig 6e :- Canny Operator

Fig 6a :- Original image

Fig 6f :- Laplacian
Fig 6b :- Sobel Operator
Table 3 shows the advantages and disadvantages as
observed in fig 6a – 6f and as reviewed by various researchers
[2][18][19]

S/No Techniques Advantages Disadvantages

1 Sobel operator Very easy at searching for smooth edges Inaccurate
2 Prewitt Operator There is a good edge detection performance on Not too accurate
vertical and horizontal edges The magnitude of coefficient is fixed and
cannot be changed
3 Robert operator Detection of edges and orientation are very easy Inaccurate
Very sensitive to noise
4 Canny edge Operator It has good localization There is false zero crossing
It extract image features without altering the
features

IJISRT19NOV223 www.ijisrt.com 331

Volume 4, Issue 11, November – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
5 Laplacian of Gaussian Easy to detect edges and their various Very sensitive to noise
orientations The operation gets diffracted by some of the
There is fixed characteristics in all directions existing edges in a noisy image
Difficulty in detecting few edges
Table 3 :- Advantages and Disadvantages of Edge detection techniques

V. CONCLUSION approaches, International Journal of Recent trends in

Engineering, vol. 1 no 2, 2015.
This paper shows studies on different edge detections [12]. R. C. Gonzalez and R.E.Woods, Digital Image
and are categorized as gradient and Gaussian based edge Processing 2nd ed. Prentice Hall, 2002.
detection. [13]. R.Maini, H.Aggarwal. Study and comparison of various image edge
detection techniques, International Journal of Image
REFERENCES processing (IJIP), volume (3), issue (1) 2016.
[14]. Rashmi 1, Mukesh Kumar2, And Rohini Saxena
[1]. A.. Rosenfel, Computer vision, a source of models for biological Algorithm And Technique On Various Edge Detection:
visual process, IEEE Transaction on Biomedical 36(1), A Survey, Signal & Image Processing: An International
pp. 83-94, 2013. Journal (SIPIJ) Vol.4, No.3, June 2013
[2]. D. Marr, E.C.Hildreth. Theory of edge [15]. S. J. Lakshmi, Prasanna,“Image segmentation using
detection, proceeding of the Royal Society, 201b, various edge detection techniques”, International Journal
pp187-217, 2014. of Emerging Technologies in Engineering Research,
[3]. G.T. Shrivakshan, Dr.C. Chandrasekar, “A 2016; 129–32.
Comparison of various Edge Detection Techniques [16]. S. Kumar, I. Singh “Comparison between edge detection
used in Image Processing” IJCSI International Journal Techniques”, International Journal of Computer
of Computer Science Issues, Vol. 9, Issue 5, No 1, Applications, 2016; 15–8.
September 2012 [17]. Sunanda Gupta, Charu Gupta, and S.K. Chakarvarti
[4]. J. Canny. A computational approach to edge detection, “Image Edge Detection A Review”, International Journal
IEEE Transactions in pattern analysis and machine of Advanced Research in Computer Engineering &
intelligence vol. 8 pp. 679-698, 2013. Technology (IJARCET) Volume 2, Issue 7, July 2013
[5]. K Bala Krishnan, Shiva Prakash Ranga and Nageswara [18]. T. Peli and D.malah. A study of edge detection algorithm,
Guptha, “A Survey on Different Edge Detection computer graphics and image processing vol20 pp. 1-21,
Techniques for Image Segmentation”, Indian Journal of 2011
Science and Technology, Vol10(4), DOI: [19]. Y. Yakimovsky. Boundary and object detection in real
10.17485/ijst/2017/v10i4/108963 January 2017. world images. Journal of ACM, vol. 23, no 4 pp. 598-
[6]. M. Ibrahiem, M.EL Emery, On the application of Artificial 619, 2014.
Neural in analysing and classifying human chromosome,
Journal of Computer science vol. 2(1) pp. 72-75 2015.
[7]. M.B. Ahmad and T.S. Choi , Local Threshold and
Boolean Function Based Edge Detection, IEEE
Transactions on Consumer Electronics, Vol. 45, and No
3.August 1999
[8]. Muthukrishnan.R and M.Radha,”Edge Detection
Techniques For Image Segmentation”, International
Journal of Computer Science & Information Technology
(IJCSIT) Vol 3, No 6, Dec 2011 DOI :
10.5121/ijcsit.2011.3620 259
[9]. N. Senthilkumaran and R.Rajesh, A study on split and
merge for region based image segmentation, proceedings of
UGC sponsored national conference network security
(NCNS-08) pp57-61, 2014.
[10]. N. Senthilkumaran and R.Rajesh, Edge Detection
Techniques for image segmentation-A survey, Proceedings of the
international conference on managing next generation
software applications (MNGSA-08) pp. 749-760, 2013.
[11]. N. Senthilkumaran and R.Rajesh, Edge detection
Techniques for image segmentation-A survey of soft computing

IJISRT19NOV223 www.ijisrt.com 332