An approach to improving edge

The International Journal of Multimedia & Its Applications (IJMA) Vol.7, No.1, February 2015
DOI : 10.5121/ijma.2015.7102 17
AN APPROACH TO IMPROVING EDGE
DETECTION FOR FACIAL AND REMOTELY
SENSED IMAGES USING VECTOR ORDER
STATISTICS
B O. Sadiq, S.M. Sani and S. Garba
Department of Electrical and Computer Engineering,
Ahmadu Bello University, Zaria
ABSTRACT
This paper presents an improved edge detection algorithm for facial and remotely sensed images using
vector order statistics. The developed algorithm processes coloured images directly without been converted
to grey scale. A number of the existing algorithms converts the coloured images into grey scale before
detection of edges. But this process leads to inaccurate precision of recognized edges, thus producing false
and broken edges in the output edge map. Facial and remotely sensed images consist of curved edge lines
which have to be detected continuously to prevent broken edges. In order to deal with this, a collection of
pixel approach is introduced with a view to minimizing the false and broken edges that exists in the
generated output edge map of facial and remotely sensed images.
KEYWORDS
Vector Order Statistics, Facial Images, Remotely Sensed Images and Coloured Images.
1. INTRODUCTION
One of the most important task in image processing is detection of edges [1]. Edge detection is a
low level feature in image processing that deals with the extraction of important features in
images. An edge in an image is caused by local discontinuity in pixel due to either light, shadows
or illumination [2]. The fundamental goal of edge detection is to produce a line drawing of a
scene from an image of that Scene. Thus, important features such as curves and corners can be
extracted from the edges of the images[3]. Face detection is a technique used to find faces at
different locations with different sizes in a given location. It is applicable in the field of image
processing in biometrics, multimedia applications, video surveillance amongst others [4]. The
fundamental aim of face and object detection is successful edge identification and extraction.
Edge maps are generated in face detection with a view to representing faces as a single unit.
These generated edge maps are one of the most popular way of representing facial images and it
features [5]. Remotely sensed images are data that contain important information which are
acquired about an object or phenomenon without making physical contact. This replaced
expensive and inefficient data collection on ground, assuring that areas of process are not

18
disturbed [6]. The Edge maps generated using edge detection algorithms concentrates on the
pertinent information of a remotely sensed image, the method used to extract these edge maps
effectively is extremely important for image processing applications [7]. Some of the pertinent
information generated by applying the edge detection algorithms on remotely sensed images are
road networks, geological features, and desert extraction amongst others [8].
Numerous researchers have developed edge detection algorithms for facial and remotely sensed
images such as the work of [4], [5], [6], [9] and [10]. The authors in [4] and [5] presented an
algorithm for extraction of edges in facial images using the Sobel and Canny edge detection
algorithms. But broken and false edges exist in the output edge map. The authors in [6] and [9]
also presented an algorithm for edge detection in satellite images. However the algorithm
produced displaced edges. The author in [10] presented a comprehensive edge detection
algorithm for satellite images using the laplacian mask. This method produced falsified edge
lines. In view of the imperfection associated with the existing works, there is need to develop an
improved edge detection algorithm that will produce thin and continuous edge lines in the
generated output edge maps.
2. VECTOR ORDER STATISTICS
A typical way to represent coloured images in a vector form. Coloured images are 3-D images
that assign three numerical values to each pixel in an image [11]. The ordering of these numerical
channels is defined as Vector Order Statistics. This ordering of component in coloured images are
of four different types namely: the Vector Range (VR), Minimum Vector Range (MVR), Vector
Dispersion (VD) and Mean Vector Dispersion (MVD) [12]. The easiest to implement and less
sensitive to noise is the minimum vector range. The minimum vector range calculates the
Euclidean distance between two pixels in an image after ordering of the sort using equation (2.1)
[13].
MVR = || Xn – X1|| (2.1)
Where; || || is the vector norm,
Xn is the nth
pixel in the image
X1 is the last pixel in the image
Figure 2.1 shows the 3x3 window indexing
Figure 2.1: Edge Pixel Indexing/ Arrangement

3. METHODOLOGY
i. From the input image, generate a 3x3 window
ii. Find the Euclidean distance between eac
iii. Apply the developed mask based on collection of pixel.
iv. Use non maximum suppression to reduce thick edges.
v. Determine which pixel is an edge or not using a threshold value.
3.1 Pixel Collection
With a view to reducing false and broken edges at curves, a collection of pixel scheme is
proposed based on the step and roof edge profile as depicted in figure 3.1
Figure 3.1 Step and Roof Edge Profile
The collection of pixels for each collection scheme are from integers 0
collection of pixels are that of an 8
for 8-Neighborhood Pixel in a 3x3 Window.
Figure 3.2 Integer Notation for 8
Based on the Step and Roof Profile, a Collection scheme is generated as in Figure 3.3
From the input image, generate a 3x3 window size pixel.
Find the Euclidean distance between each pixel in the window
Apply the developed mask based on collection of pixel.
Use non maximum suppression to reduce thick edges.
Determine which pixel is an edge or not using a threshold value.
proposed based on the step and roof edge profile as depicted in figure 3.1
Figure 3.1 Step and Roof Edge Profile
The collection of pixels for each collection scheme are from integers 0-8 which implies that the
collection of pixels are that of an 8- Neighbourhood pixel. Figure 3.2 shows the integer notation
Neighborhood Pixel in a 3x3 Window.
nteger Notation for 8-Neighborhood Pixel in a 3x3 Window.
19
8 which implies that the
Neighbourhood pixel. Figure 3.2 shows the integer notation

Figure 3.3 Collection Scheme based on
The developed collection scheme are represented as a mask and applied to the image with a view
to producing thin and continuous edge lines. The generated mask developed from the collection
scheme as shown in figure 3.4
Figure 3.4
3.2 Vector Range
Let R, G, B denote the unit vectors along the RGB axis of the RGB colour space. Given an image
I, the vector of the colour space in the image can be defined as
(m, n) = size (I) where, size (I) is the dimension of the image used.
The Euclidean distance of a pixel m is given by
Where, m is in the norm form
In a grid of pixels that constitutes the image, the Euclidean distance between two pixels is (m, n)
is
Collection Scheme for Step Edge
Collection Scheme for Roof Edge
Figure 3.3 Collection Scheme based on Step and Roof Edge
Figure 3.4 Developed Mask from the Collection Scheme
I, the vector of the colour space in the image can be defined as
b
P
B
g
P
G
r
P
R
m
∂
∂
+
∂
∂
+
∂
∂
=
b
Q
B
g
Q
G
r
Q
R
n
∂
∂
+
∂
∂
+
∂
∂
=
(m, n) = size (I) where, size (I) is the dimension of the image used.
The Euclidean distance of a pixel m is given by
mmm •=
mnnmD E −=),(
20
(3.1)
(3.2)
(3.4)
(3.5)

Since the input images is a colour image consisting of three components, this can be represented
as
Where DE is the Euclidean distance between two pixels
The Vector Range between grids of pixels is calculated by subtracting the last pixel for the
pixel
MVR = min {VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8}
Where, MVR = Minimum Vector Range
The Flow Chart for the Developed Algorithm is shown in Figure 3.5
input images is a colour image consisting of three components, this can be represented
is the Euclidean distance between two pixels
The Vector Range between grids of pixels is calculated by subtracting the last pixel for the
RGBPRGBPVR 1,12,31 −=
Where, MVR = Minimum Vector Range
The Flow Chart for the Developed Algorithm is shown in Figure 3.5
21
input images is a colour image consisting of three components, this can be represented
The Vector Range between grids of pixels is calculated by subtracting the last pixel for the first

Figure 3.5: Flow Chart for the Developed Algor
The sample of the facial and remotely sensed images collected are depicted in figures 3.6 and 3.7
Figure 3.5: Flow Chart for the Developed Algorithm
Figure 3.6: Sample of Faces Used
22

Figure 3.7: Sample of Remotely Sensed Image Used
4. RESULTS AND DISCUSSIONS
After applying the developed edge
output result are presented in Figures 4.1 and 4.2
Figure 4.1: Output Result of Applying the proposed Algorithm of Facial Images
Figure 4.1 shows the edge maps of applying the proposed
statistics to the sample of collected natural faces used. The visual performance of the result of
applying the algorithm showed that the edges that constitute the overall face was extracted
completely. The algorithm is not
resolution images the processing time increases, unless if used on a dedicated and specialized
systems. But different facial expression produces different generated edge map for the same
individual.
The output result of applying the developed algorithm to remotely sensed image is shown in
figure 4.2
Figure 3.7: Sample of Remotely Sensed Image Used
ISCUSSIONS
After applying the developed edge detection algorithm to the sample of images collected, the
output result are presented in Figures 4.1 and 4.2
Figure 4.1 shows the edge maps of applying the proposed algorithm based on vector order
completely. The algorithm is not dependent on the type of skin colour in the image. With high
23
detection algorithm to the sample of images collected, the
algorithm based on vector order
dependent on the type of skin colour in the image. With high

Figure 4.2 Output Result of Applying the proposed Algorithm of Remotely Sensed Images
Figure 4.2 shows the edge maps of applying the proposed
statistics to the sample of remotely sensed image used. The visual performance of the result of
applying the algorithm showed extracted regions and roads in the remotely sensed image.
5. CONCLUSION
This paper presents an approach to improving edge detection in facial and remotely sensed
images using a pixel collection scheme. This developed pixel collection scheme is developed
with a view to addressing the problem of false and broken edges that exist in these images due to
curves. The developed algorithm can be applied on more facial and remotely sensed images to
test its effectiveness and compare with other developed existing techniques.
REFERENCES
[1] Ayaz Akram and Asad Ismail, "Comparison of Edge Detectors,"
Science and Information Technology Research (IJCSITR), vol. 1, pp. pp.16
[2] Mehdi Ghasemi, Mahdi Koohi, and Abbas Shakery, "Edge detection in multispectral images based on
structural elements," The International
pp. pp.90-99, 2011.
[3] Samta Gupta and Susmita Ghosh Mazumdar, "Sobel edge detection algorithm," International journal
of computer science and management Research, vol. 2, pp. pp.1578
[4] Anila and Devarajan, "Simple and Fast Face Detection System Based on Edges," International
Journal of Universal Computer Sciences, vol. 1, pp. pp.54
[5] Vijayarani and Vinupriya, "Performance Analysis of Canny and Sobel Edge Detection
Image Mining," International Journal of Innovative Research in Computer and Communication
Engineering, vol. 1, pp. pp.1760
[6] Katiyar and Arun, "Comparative analysis of common edge detection techniques in context of
object extraction," IEEE Transactions of Geoscience and Remote Sensing, vol. 50, pp. pp.68
2014.
[7] Yu Xiong, "Research on an Edge Detection Algorithm of Remote Sensing Image Based on Wavelet
Enhancement and Morphology " JOURNAL OF COMPUTERS, vol. 9, pp.
[8] Beant Kaur, Anil Garg, and Amandeep Kaur, "Mathematical Morphological Edge Detection For
Remote Sensing Images," International Journal of Electronics & Communication Technology
(IJECT) vol. 1, pp. pp.29-33, 2010.
[9] Syed Jahanzeb and Ayesha Siddiqui, "Analysis of Edge Detection Algorithms for Feature Extraction
in Satellite Images " IEEE International Conference on Space Science and Communication
(IconSpace), vol. 3, pp. pp.238
[10] Abhishek Gudipalli and Ramashri Tiruma
Images," World Applied Sciences Journal, vol. 28, pp. pp.1042
Figure 4.2 shows the edge maps of applying the proposed algorithm based on vector order
pproach to improving edge detection in facial and remotely sensed
The developed algorithm can be applied on more facial and remotely sensed images to
test its effectiveness and compare with other developed existing techniques.
Ayaz Akram and Asad Ismail, "Comparison of Edge Detectors," International Journal of Computer
Science and Information Technology Research (IJCSITR), vol. 1, pp. pp.16-24, 2013.
Mehdi Ghasemi, Mahdi Koohi, and Abbas Shakery, "Edge detection in multispectral images based on
structural elements," The International Journal of Multimedia & Its Applications (IJMA), vol. Vol.3,
Samta Gupta and Susmita Ghosh Mazumdar, "Sobel edge detection algorithm," International journal
of computer science and management Research, vol. 2, pp. pp.1578-1583, 2013.
Anila and Devarajan, "Simple and Fast Face Detection System Based on Edges," International
Journal of Universal Computer Sciences, vol. 1, pp. pp.54-58, 2010.
Vijayarani and Vinupriya, "Performance Analysis of Canny and Sobel Edge Detection
Engineering, vol. 1, pp. pp.1760-1767, 2013.
Katiyar and Arun, "Comparative analysis of common edge detection techniques in context of
extraction," IEEE Transactions of Geoscience and Remote Sensing, vol. 50, pp. pp.68
Yu Xiong, "Research on an Edge Detection Algorithm of Remote Sensing Image Based on Wavelet
Enhancement and Morphology " JOURNAL OF COMPUTERS, vol. 9, pp. pp.1247-1252, 2014.
Beant Kaur, Anil Garg, and Amandeep Kaur, "Mathematical Morphological Edge Detection For
33, 2010.
nd Ayesha Siddiqui, "Analysis of Edge Detection Algorithms for Feature Extraction
(IconSpace), vol. 3, pp. pp.238-242, 2013.
Abhishek Gudipalli and Ramashri Tirumala, "Comprehensive Edge Detection Algorithm for Satellite
Images," World Applied Sciences Journal, vol. 28, pp. pp.1042-1047, 2013.
24
algorithm based on vector order
pproach to improving edge detection in facial and remotely sensed
The developed algorithm can be applied on more facial and remotely sensed images to
International Journal of Computer
Mehdi Ghasemi, Mahdi Koohi, and Abbas Shakery, "Edge detection in multispectral images based on
Journal of Multimedia & Its Applications (IJMA), vol. Vol.3,
Samta Gupta and Susmita Ghosh Mazumdar, "Sobel edge detection algorithm," International journal
Anila and Devarajan, "Simple and Fast Face Detection System Based on Edges," International
Vijayarani and Vinupriya, "Performance Analysis of Canny and Sobel Edge Detection Algorithms in
Katiyar and Arun, "Comparative analysis of common edge detection techniques in context of
extraction," IEEE Transactions of Geoscience and Remote Sensing, vol. 50, pp. pp.68-79,
Yu Xiong, "Research on an Edge Detection Algorithm of Remote Sensing Image Based on Wavelet
1252, 2014.
Beant Kaur, Anil Garg, and Amandeep Kaur, "Mathematical Morphological Edge Detection For
nd Ayesha Siddiqui, "Analysis of Edge Detection Algorithms for Feature Extraction
la, "Comprehensive Edge Detection Algorithm for Satellite

25
[11] Chris Solomon and Toby Breckon, "Fundermentals of Digital Image Processing A Practical
Approach With Examples in MATLAB," First ed: John Wiley & Sons Ltd, 2011, pp. 1-109.
[12] Panos E Trahanias and Anastasios N Venetsanopoulos, "Vector order statistics operators as color
edge detectors," IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, vol. 26,
pp. pp.135-143, 1996.
[13] Andreas Koschan and Mongi Abidi, "Detection and classification of edges in colour images," IEEE
Signal processing magazine, special issue on colour image processing, vol. 22, pp. pp.64-75, 2005.

An approach to improving edge

More Related Content

What's hot (20)

Similar to An approach to improving edge (20)

Recently uploaded (20)

An approach to improving edge