Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
Procedia
Engineering
Procedia Engineering
Procedia 00 (2011)
Engineering 000–000
15 (2011) 1313 – 1318
www.elsevier.com/locate/procedia

Advanced in Control Engineering and Information Science

Research and analysis of Image edge detection

algorithm Based on the MATLAB
Guowei Yang, Fengchang Xu *

(School of Information Engineering, Nanchang Hangkong University, Nanchang 330063, China)

Abstract

Image edge detection is one of the important contents of image processing. In this paper, we show a new
edge detection operator, which is Log_Sobel. This operator processes image with logarithm of luminosity
(luminous density) not with luminosity. It processes fast, and suits online detection. What’s more, the
detect result is in dependent of luminosity. At the same time, it is applied to the online detection. The
experiment results show that the effect for image processed by Log_Sobel operator is better than those
operators, including Roberts algorithm, Prewitt algorithm and Sobel algorithm, proposed by predecessor.
© 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and/or peer-review under responsibility of [CEIS 2011]
Key words: edge detection; Log_Sobel operator; simulation

1. Introduction

Edge detection is the important basis of image analysis, including image segmentation, texture feature
extraction and shape feature extraction. Therefore, edge detection becomes particularly important. In
recent years, it becomes the most active field of the image processing. Predecessors have developed
many edge detection operations, such as Roberts operator, Sobel operator, Prewitt operator, Kirsch
operator and so on, which are directly on the image intensity of the light. In the Fig.1, we describe the
relationship of imaging among incident light intensity f ( x, y ) , reflected light intensity g ( x, y ) and
object reflectivity e( x, y ) . The relationship formula is as follows:

*
Corresponding author: Guowei Yang. Tel.: +86-791-3953432; +86-791-3953432.
E-mail address: ygw_ustb@163.com

1877-7058 © 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
doi:10.1016/j.proeng.2011.08.243
1314
2 GuoweiYang,
Guowei Yang Fengchang
and Fengchang
Xu / Xu / Procedia
Procedia Engineering
Engineering 15 (2011)
00 (2011) 1313 – 1318
000–000

g ( x , y ) = e( x , y ) f ( x , y ) (1)

Fig.1. Reverberating of optics

If every part of the object is uneven illumination intensity during the processing of the imaging, or images
may image under the different light intensity. Uneven illumination intensity has the serious influence on
the edge detection. In photograph technology, logarithm of light intensity is called light intensity, which
is proportional to the logarithm of the objects reflectivity. So it is usually the logarithm function of the
exposure. During the both sides of formula (1)’s make the logarithmic transformation, then we will gain
light intensity formulation (2):
log( g ( x, y )) = log(e( x, y )) + log( f ( x, y )) (2)
The incident light f ( x, y ) is slowly changed in a small local area within weight evenly, so the
adjacent points of light intensity f ( x, y ) can be regarded as phase equivalence. Adopting logarithmic
subtraction can eliminate the light intensity factor, and then only retains the change of objects reflectivity.
This paper puts forward the operator of Log_Sobel based on this one principle for Sobel operator
improvement, which not only processes optical density directly to eliminate the effect caused by the
uneven illumination intensity, but also inherits smooth function of Sobel operator and improves anti-
jamming of the detection. In addition, the two boundary threshold values which can be preset have only
contacted with the contrast sensitivity. Therefore, two values of detection area on the borders can finish
simultaneously during the testing processing.
When the Log_Sobel algorithm is used to process the image, the texture is clear and the effect is
obvious. This algorithm is compared with those algorithms proposed by the predecessor, including
Roberts algorithm, Prewitt algorithm and Sobel algorithm. The effect of the former is much better, and
the processing speed is much quicker. Meanwhile, it has certain realistic application, and also suits in
online detection.

2. Log_Sobel operator

2.1. Sobel operator

Sobel operator is formed by two convolution kernels which shows in the table 1 and table 2. In image
processing, these two kernels are used to convolute each point of the image. Then takes the biggest one as
output:

Table 1. Sobel horizon operator

-1 -2 -1
0 0 0
1 2 1
 Guowei YangYang,
Guowei and Fengchang
FengchangXu
Xu/ /Procedia
ProcediaEngineering
Engineering1500(2011)
(2011)1313 – 1318
000–000 1315
3

Table 2. Sobel vertical operator

-1 0 1
-2 0 2
-1 0 1

2.2. Log_Sobel operator

Log_Sobel operator adopts the convolution kernels of Sobel operator, which processes the object that
is not image gray-scale, but the logarithm of image gray-scale. This can retain the insensitivity for the
original method of optical strength changes, simultaneously through the smooth to overcome the
shortcomings of anti-jamming weakness. It can enhance the signal-to-noise ratio. The computation
formula is as follows:
A = (log(g ( x + 1, y − 1)) + log(2 * g (x + 1, y )) + log(g (x + 1, y + 1))) −
(log(g (x − 1, y − 1)) + log(2 * g (x − 1, y )) + log(g ( x − 1, y + 1)))
(3)
B = (log( g ( x − 1, y + 1)) + log(2 * g ( x, y + 1)) + log( g (x + 1, y + 1))) −
(log( g ( x − 1, y − 1)) + log(2 * g (x, y − 1)) + log(g ( x − 1, y + 1)))
(4)
G ( x, y ) =| A | + | B |
(5)

2.3. experiment result

Fig. 2a is the original image of the illumination uneven, which is taken by the digital camera. Fig. 2b is
the Roberts operator test result. Fig. 2c is the Sobel operator test result. Fig.2d is the Pr ewitt operator
test result. Fig. 2e is the Log_Sobel operator test result. In order to enhance the speed reduction floating
point calculation, we process the light density of each pixel. Which can be obtained by using the table
look-up law.
1316
4 GuoweiYang,
Guowei Yang Fengchang
and Fengchang
Xu / Xu / Procedia
Procedia Engineering
Engineering 15 (2011)
00 (2011) 1313 – 1318
000–000

(a) Original figure (b) Result of Roberts operator

(c) Result of Sobel operator (d) Result obtained by Prewitt operator

 Guowei YangYang,
Guowei and Fengchang
FengchangXu
Xu/ /Procedia
ProcediaEngineering
Engineering1500(2011)
(2011)1313 – 1318
000–000 1317
5

(e) Result obtained by Log_Sobel operator

Fig.2. Image of edge detection

3. Conclusions

The Log_Sobel algorithm is proposed in this article. When it is used to process the image, the texture is
clear and the effect is obvious. This algorithm is compared with those algorithms proposed by the
predecessor, including Roberts algorithm, Prewitt algorithm and Sobel algorithm. The effect of the former
is much better, the processing speed is much quicker. Meanwhile, it has certain realistic application, and
also suits in online detection.

Acknowledgements

This program is sponsored by the National Natural Science Foundation of China under Grant
(No.60973048), the Natural Science Foundation of Jiangxi Province under Grant (No.2009GZS0084), the
Research Fundation of Nanchang Hangkong University under Grant (No.EA200906012), and the
Postgraduate Innovates Fundation of Nanchang Hangkong University under Grant (No.YC2010030).
1318
6 GuoweiYang,
Guowei Yang Fengchang
and Fengchang
Xu / Xu / Procedia
Procedia Engineering
Engineering 15 (2011)
00 (2011) 1313 – 1318
000–000

References

[1] Wang Lu. The license plate identification system based on MATLAB [D]. Shanghai Jiao tong university, 2009.
[2] Xue Hongquan. The research of the license plate identification system Based on the wavelet analysis [D]. University of
electronic science and technology, 2009.
[3] Yun Tao, Cui Qianhuang. Automatic license extraction from moving vehicles. IEEE Transaction on Pattern Analysis and
Machine Intelligence, 1999, 7 (4); 126-130
[4] Wang HuiYan. The image edge detection and the image matching research and application (PH.D, Thesis [D]). Hangzhou,
Zhejiang university, 2003.
[5] Xu Guobao, Wang Ji, Zhao Guiyan, Yin Yixin, Xie Shiyi. A new operator of adaptive edge detection based on mathematical
morphology [J]. Computer application, 2009, (04).
[6] Gan Jinlai. Comparative study of the image edge detection algorithm [D]. University of electronic science and technology,
2005.
[7] L. Dlagnekov. Video-based car surveillance: License plate make and model recognition. Masters Thesis, University of
California at San Diego, 2005.
[8] Chen Fen. Edge detection based on the multi-scale wavelet transformation [J]. Computer engineering and application, 2009,
(27).
[9] Kang Mu, Wang Baoshu. Self-adaptive Kirsch edge detection algorithm [J]. Journal of Huazhong technology university
(natural science edition), 2009, (04).
[10] Ming Zhang, Bahadir Gunturk. A New Image Denoising Method Based on the Bilateral Filter. IEEE, 2008, 929-923.
[11] Y. Gao and M. K. H. Leung. Face recognition using line edge map. In IEEE Trans. on Pattern Analysis and Machine
Intelligence Vol. 24, 764-779, 2002.
[12] R. O. Duda P. E. Hart and D. G. Stork. Pattern Classification. John Wiley & Sons, Inc. New York, second edition, 2004.
[13] M. K. H. Leung and Y. H. Yang. Dynamic two-strip algorithm in curve fitting. In Pattern Recognition, vol. 23, 69-79, 1990.
[14] K. Reumann and A. P. Witkam. Optimizing curve segmentation in computer graphics. Int. Comput. Symp. 244–256, 1974.