195
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
MODIFIED SOBEL MASK TO LOCATE KNEE JOINT
BOUNDARIES
S. Sheik Abdullah
Research Scholar, Department of Electronics and Communication Engineering,
Kalasalingam Academy of Research and Education, Virudhunagar, (India).
E-mail: sabdullah787@gmail.com ORCID: https://orcid.org/0000-0001-6765-8374
M. Pallikonda Rajasekaran
Professor, Department of Electronics and Communication Engineering,
Kalasalingam Academy of Research and Education. Virudhunagar, (India).
E-mail: mpraja80@gmail.com ORCID: https://orcid.org/0000-0001-6942-4512
Recepción:
05/12/2019
Aceptación:
20/12/2019
Publicación:
23/03/2020
Citación sugerida:
Abdullah, S. S., y Rajasekaran, M. P. (2020). Modied sobel mask to locate knee joint boundaries.
3C Tecnología. Glosas de innovación aplicadas a la pyme. Edición Especial, Marzo 2020, 195-205. http://doi.
org/10.17993/3ctecno.2020.specialissue4.195-205
Suggested citation:
Abdullah, S. S., & Rajasekaran, M. P. (2020). Modied sobel mask to locate knee joint boundaries.
3C Tecnología. Glosas de innovación aplicadas a la pyme. Edición Especial, Marzo 2020, 195-205. http://doi.
org/10.17993/3ctecno.2020.specialissue4.195-205
196 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
ABSTRACT
Sobel masking algorithm is a very important technique to detect edges in an image.
Comparing the Sobel gradient operator with other edge/boundary detection operators
used repeatedly; Making an additional study on the traditional Sobel gradient operator,
the benets of Sobel mask are its quick speed of detecting edges. Meanwhile, it has also
an impact on suppressing and smoothing noise. In addition, Sobel operator has a standard
eect on detecting the edges. Although Sobel gradient operator has some advantages in
dierent aspects, it exists some problems. The Existing Sobel masking technique is a type
of edge detection in vertical and horizontal directions only and it ignores the boundary
points in other directions. It cannot attain a true location of edge points in an image. In this
paper, the existing sobel technique is improved by adding an increase of 315 degrees and
360 degrees in horizontal and vertical directions. This will have an eect of detecting the
knee joint space of osteoarthritis. According to simulation results, they show this method is
very simple and feasible, and the outcomes are more abundant than traditional Sobel edge
detection. In this paper edge detection and noise interference problems are improved.
KEYWORDS
Osteoarthritis, Sobel mask, Image Processing Techniques.
197 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
1. INTRODUCTION
In digital image processing, edge/boundary feature is one of the very important
characteristics of the image, and it is a signicant part of image processing, analyzing,
pattern recognition and computer vision. Edge detection outcomes aect further image
analyzing and pattern/texture recognition directly (Amer & Abushaala, 2015). In recent
days, Image edge detection has become the main research theme in image processing
technology. With the advance of science and technology, researchers have analyzed and
proposed some techniques for the detection of edges in an image and assessment of edge
detection. At the same time, these edge/boundary recognition methods are applied to
the area of digital vision and pattern recognition, which make the use of edge detection
technology more broadly. Over the years, segmentation of an image has been creating more
and more attention. Lots of image segmentation techniques have been put forward. They
can be divided into dierent methods like bit threshold method, edge detection method and
regional growth method (Argyle, 1971; Canny, 1989). Edge detection method comprises
of: edge detection operator which contains mask like Roberts operator, Prewitt operator,
LOG operator and Sobel operator (Abbasi & Abbasi, 2007). Sobel mask is slightly better
than others. The classical Sobel technique also has some problems such as it is sensitive to
the vertical and horizontal direction only (Lakshmi & Sankaranarayanan, 2010). However,
the information in the image is not restricted to the horizontal and vertical directions; it can
make an element of the image information lose. In this paper, a new improved operator
is proposed to detect more image information. In the modied Sobel operator, 2 direction
patterns (315 degrees and 360 degrees) are added to get multi-directional image acquisition.
Then calculate the threshold by using the Otsu method and rene the detected rough edges
by using the method to achieve the results of image edge detection. Edge detection eect
can be achieved better by using the Matlab simulation method.
2. LITERATURE SURVEY-COMPARISION OF TRADITIONAL EDGE
DETECTION OPERATORS
Roberts operator: It did not pass smooth analysis, so it is very sensitive to the noise.
Prewitt operator and Sobel operator: extraction of edge/boundaries eect is almost
the same (Lakshmi & Sankaranarayanan, 2010; Abbasi & Abbasi, 2007). Sobel operator is
a weighted average lter, Prewitt operator is an average lter; Sobel operator have better
198 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
detection eect on images which have low level noise, but the detection of the edge eect
is not clear.
LOG operator: detecting edges by using second order derivatives zero crossing edge
method (Yu-quian, Wei-hua, Zhen-cheng, Jing-tian, & Ling-yun, 2005). Smoothing eect
is more important, noise removal is improved, but the loss of information in an image is
higher, the edge accuracy is lower. So there is a challenge between placing edge accuracy
and removing noise level.
3. TRADITIONAL SOBEL OPERATOR
Sobel operator, because of its task in the pattern is small, the computation is also very small,
and the image information of the shape can be attained. Operator template size is even,
the pending pixel cannot be placed in the center position of the template. Sobel dierential
mask is a dierential mask of 3 x 3 size template (Argyle, 1971). The expressions of formula
as follow:
Gx(u,v)=f[u-1,v+1]+2
*
f[u,v+1]+f[u+1,v+1]-
f[u-1,v-1]-2
*
f[u,v-1]-f[u+1,v-1]
f[u-1,v-1]-2
*
f[u,v-1]-f[u+1,v-1]
(1)
Gx(u,v)=f[u-1,v-1]+2
*
f[u+1,v]+f[u+1,v+1]-
f[u-1,v-1]-2
*
f[u-1,v]-f[u-1,v+1]
f[u-1,v-1]-2
*
f[u-1,v]-f[u-1,v+1]
(2)
The convolution template of the Sobel operator is expressed as the formula
Gx=
-1 -2 -1
Gy=
-1 0 1
0 0 0 -2 0 2
1 2 1 -1 0 1
The calculating steps of Sobel operator: rst, the edge detection image is divided into
matrix form
199 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
f(x,y) f(x+1,y) f(x+2,y)
f(x,y+1) f(x+1,y+1) f(x+2,y+1)
f(x,y+2) f(x+1,y+2) f(x+2,y+2)
Multiply the vertical direction by horizontal direction of the template,
Fx =Gx.*A
(3)
Fx =Gy.*A
(4)
gradient size calculation, as shown in the formula
Gx
2
+Gy
2
G=
(5)
The formula for calculating the gradient direction is shown in the formula
θ=tan
-1
(Gy/Gx)
(6)
The Sobel mask set up the weighted local average, The operator not only inuences the
edge detection of an image but also hold back the noise further, but the edge is wider. The
basic idea of Sobel operator algorithm: The edge of the image is situated at the place
in which the brightness varies signicantly (Kalpana & Padmaa, 2014), the gray value of
pixels exceeds a set threshold depending on the specic steps for the edge (Xing, 2005). The
specic steps of the Sobel operator algorithm are as follows:
• Moving the horizontal and vertical direction templates from right to left, from top to
bottom, and moving from one pixel to another.
• Multiplying the pixel values in the image with operator coecient.
• Calculated gradient value is the new gray value by using 2 convolution values.
4. IMPROVED SOBEL OPERATOR
Adding 315 degrees and 360 degrees with respect to the template in a basis of the traditional
Sobel operator, the direction template is changed into two directions (Gx=315 degree, Gy=360
degree): the horizontal and vertical direction with respect to 315 degrees and 360 degrees. It
200 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
improved the weights of the new template in the direction of boundaries. Specic details
are as follows:
According to the calculation of the two template directions and calculating an image point
by point, the maximum value is observed as the pixel gray values. According to the threshold
setting, the edge point is determined.
Sx(o,p)=f[o+2,p+1]+2
*
f[o+2,p+2]+f[o+1,p+2]-
f[o+1,p]-2
*
f[o,p]-f[o,p+1]
f[o+1,p]-2
*
f[o,p]-f[o,p+1]
(7)
Sx(o,p)=f[o,p+2]+2
*
f[o+1,p+2]+f[o+2,p+2]-
f[o,p]-2
*
f[o+1,p+2]-f[o+2,p]
f[o,p]-2
*
f[o+1,p+2]-f[o+2,p]
(8)
Sx=
-3 -1 0
Sy=
-1 -2 -1
-1 0 0 0 0 0
0 1 2 1 2 1
Start
Convert RGB image to grayscale
Sobel Operator
Sobel Operator
Edge Detection
Finish
Figure 1. Flowchart of proposed system.
201 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
5. RESULTS
Edge detection process followed below:
Step 1: Set Threshold value T=255.
Step 2: If Gradient value(S) <255 is less than the Thresh, considered as 1, other than are
0 (value below 255 set to be 0).
Figure 2. Original Image. Figure 3. Proposed System.
Figure 4. Sobel Operator. Figure 5. Roberts Operator.
Figure 6. Prewitt Operator. Figure 7. Homogeneity Operator.
202 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Table 1. PSNR comparison of different operators.
Test
images
PSNR value (dB)
Sobel Zero Cross Prewitt Roberts Homogeneity
Modied
Sobel
01 +19.39 +13.48 +18.51 +14.25 +10.23 +20.95
02 +19.58 +13.92 +18.76 +14.73 +10.18 +21.34
03 +19.63 +13.79 +13.71 +14.63 +10.27 +21.15
04 +19.68 +13.86 +18.80 +15.21 +10.20 +21.22
05 +20.03 +13.76 +18.96 +14.32 +10.31 +21.49
06 +19.99 +13.84 +19.22 +16.23 +10.22 +21.30
07 +19.88 +13.86 +19.00 +16.10 +10.19 +21.45
08 +19.43 +13.87 +18.63 +15.34 +10.21 +21.03
09 +19.63 +13.67 +18.72 +15.64 +10.41 +21.21
10 +19.13 +13.82 +18.16 +13.62 +10.23 +20.86
11 +19.20 +13.67 +18.12 +13.65 +10.08 +20.98
6. CONCLUSION
This paper analyzes the classic sobel edge detection algorithm and improves the algorithm
from the gradient calculation. The improved algorithm is realized that result outcomes
prove that the modied algorithm is better and clearer on the edge detection of the image.
From experiment, it proved that this proposed system is better than the traditional Sobel
operator in image edge detection and achieves the specic accurate detection and reduces
the loss of edge. The experiments show that the method provided in this paper is feasible.
Improve the masking performance by increasing PSNR value and detect nite boundaries/
edges of intra articular space in future.
203 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
REFERENCES
Abbasi, T. A., & Abbasi, M. U. (2007). A novel FPGA-based architecture for Sobel edge
detection operator. International Journal of Electronics, 13(9), 889-896. https://doi.
org/10.1080/00207210701685253
Amer, G. M. H., & Abushaala, A. M. (2015). Edge Detection Methods. In G. Deng, Z.
Liu. Comparison and Analysis for Edge Detection Algorithms based on SQI Image Enhancement.
IEEE, 3615-3617.
Argyle, E. (1971). Techniques for edge detection. IEEE proceedings, 59, 285-286.
Canny, J. (1989). A computational approach to edge detection. IEEE Transactions on Pattern
Analysis and Machine Intelligence, PAMI-8(6), 679-698. https://doi.org/10.1109/
TPAMI.1986.4767851
Kalpana, Y. B., & Padmaa, M. V. (2014). An ecient edge detection algorithm for ame
and re image processing. In 2014 International Conference on Communication and Signal
Processing, 696-700. https://doi.org/10.1109/ICCSP.2014.6949932
Lakshmi, S., & Sankaranarayanan, V. (2010). A study of Edge Detection Techniques
for Segmentation Computing Approaches. IJCA Special Issue on Computer Aided Soft
Computing Techniques for Imaging and Biomedical Applications. https://pdfs.semanticscholar.
org/3c0e/8119096edd337002d8430c6c6ad69f126520.pdf
Rama Bai, M. (2010). A new approach for border extraction using morphological methods.
International Journal of Engineering Science and Technology, 2, 3832-3837.
Xing, J. (2005). Digital image edge detection based on Sobel operator. Journal of microcomputer
development, 18-19.
Yu-quian, Z., Wei-hua, G., Zhen-cheng, C., Jing-tian, T., & Ling-yun, L. (2005).
Medical Images Edge detection Based on mathematical Morphology. In 2005 IEEE
Engineering in Medicine and Biology 27th Annual Conference, Shanghai, China. https://doi.
org/10.1109/IEMBS.2005.1615986
204 http://doi.org/10.17993/3ctecno.2020.specialissue4.195-205
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Zhang, R., Zhao, G., & Su, L. (2005). A New Edge Detection Method in Image Processing.
IEEE International Symposium on Communications and Information Technology. ISCIT, 1, 445-
448. https://doi.org/10.1109/ISCIT.2005.1566889
