IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31,2012 55

Edge Detection of Angiogram Images Using the

Classical Image Processing Techniques
S.Sathish Kumar
l
, Research Scholar,
Sri Chandrasekharendra Saraswathi Viswa Mahavidyalaya
University, Kanchipuram, India
Senior Lecturer, School of Engineering, Asia Pacifc
University College of Technology and Innovation, Kuala
Lumpur, Malaysia
msskm 1 O@gmail.com
Abstract- The Blood vessels of the human body can be
visualized using many medical imaging methods such as X-ray,
Computed Tomography (CT), and Magnetic Resonance (MR). In
medical image processing, blood vessels need to be extracted
clearly and properly from a noisy background, drift image
intensity, and low contrast pose.
Angiography is a procedure widely used for the
observation of the blood vessels in medical research, where the
angiogram area covered by vessels and/or the vessel length is
required. For this purpose we need vessel enhancement and
segmentation. Segmentation is a process of partitioning a given
image into several non-overlapping regions. Edge detection is an
important task and in the literature, complex algorithms have
been modeled for the detection of the edges of the blood vessels.In
this paper, the edges of the vessels in the angiogram image are
detected using the proposed algorithm which is done using the
classical image processing techniques. This involves the Pre
processing step, where the noise is removed using a simple flter
and Histogram equalization technique, instead of the Canny edge
Detector. The proposed algorithm is not complicated but
accurate and involves very simple steps.
Keywords- Angiogram image, Segmentation, Vessel extraction,
Canny edge detector, fltering, Image enhancement, Histogram
equalization
I. INTRODUCTION
Segmentation plays a vital role in the detection of blood
vessels in an angiogram image. It is a process of partitioning
an angiogram into several non-overlapping regions. Thus it is
used to extract the vascular and background regions. Based on
the partitioning results, surfaces of vasculatures can be
extracted, modelled, manipulated, measured and visualized.
Hence it is used to detect the various vascular diseases.
Therefore, developing reliable and robust image segmentation
methods for angiography has been the priority and by the
other research groups [4, 10, 11].
Generally segmentation subdivides the image into its
constituent parts or objects. Autonomous segmentation
is one of the most dificult tasks in image processing.
Segmentation algorithms for monochrome images
R.Amutha
2
, Professor,
Department of Electronics and Communication Engineering
SSN College of Engineering
Chennai, India
amuthar@ssn.edu.in
generally are based on two basic properties of gray level
values i) Discontinuity and
ii) Similarity. It is challenging to perform image
segmentation in angiography [12, 13, 15].
Angiograms can be analyzed using computers to detect the
blood vessel boundary as a frst step. In the literature, this is
performed using the magnitude of the image gradient but this
method does not provide suffcient information for locating
the boundary of the blood vessel and hence the performance of
segmentation becomes complicated [1, 2, 6, 8, 11]. Thus, the
quality of the image segmentation can be improved using the
proposed histogram equalization technique, rather than the
gradient magnitude.
Edge detection is done to segment the blood vessels from
the angiogram images. Edge detection algorithms are followed
by linking and boundary detection procedures. Edge detection
is used for detecting discontinuities in gray level. First and
second order digital derivatives are implemented to detect the
edges in an image. Edge is defned as the boundary between
two regions with relatively distinct gray-level properties. An
edge is a set of connected pixels that lie on the boundary
between two regions [3, 5, 16].
The complexity of the image is reduced by detecting the
edges. They can be used to measure parameters related to
blood fow or to locate some patterns in relation to vessels in
angiographic images. Thus edge detection is done using frst
order derivative (Gradient operator), Second-order derivative
(Laplacian operator) and also using the sobel and prewitt
algorithms [7, 9].
Canny proposed the hysteresis thresholding method in
which two threshold values have to be given. However, its
performance was not good enough with respect to detection,
localisation, and resolution and noise rejection. The
percentage of true edges detected is also less as compared to
the other algorithms so proposed in the literature [14].
In this paper, the edges of the vessel in the angiogram
image is detected using the proposed algorithm which
involves the Pre-processing step, where the noise is removed
using the Median flter and Histogram equalization technique,
instead of the Canny edge Detector. Median fltering is useful
in eliminating the intensity spikes while preserves the edges in
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 56
a better way. Histogram equalization stretches or compresses
an image, which is used to detect the edges of a blood vessel.
II. EXISITNG METHOD
CANNY EDGE DETECTOR
The Canny edge detection operator was developed in the
year 1986 by John.F.Canny. He used a multi-stage algorithm
to detect a wide range of edges in the images. The following
are the various stages of the Canny Edge Detection algorithm.
A. Noise Reduction
The Canny edge detector uses the frst derivative of a
Gaussian as a flter. It flters the noise by convolving the raw
image with the Gaussian flter. By doing so, we obtain a
slightly blurred version of the original image but it is not
affected by a single noisy pixel to any signifcant degree.
Thus, it is used to remove the noise present on raw
unprocessed data.
B. Intensit Gradient
This algorithm uses four flters to detect the various edges in
the blurred image. Since an edge in the image may point in a
variety of directions, horizontal, vertical and diagonal edges
are to be detected. This is done by calculating the gradient of
the pixel relative to its neighborhood. A good approximation
of the frst derivate is given by the two Sobel operators .Since
the derivatives enhance noise; the smoothing effect is a
particularly attractive feature of the Sobel operators. First
derivatives are implemented using the magnitude of the
gradient.
For a function f (x, y), the gradient f at co-ordinate (x, y) is
defned as the 2-dimesional column vector
4
=
[ J
=
[;: 1
ff = mag(ff) =

,
= ,

.,

If (x,y) represent the direction angle of the vector Vf at

(x,y), then, (x,y) = tan-
I
(Gy/Gx)
Thus, the direction of an edge at (x,y) is perpendicular to the
direction of the gradient vector at that point
Tho d;",olio" of th' yod;e"' ;b, a = tan -{ J
Hence the edge direction angle is rounded to one of the four
angles representing vertical, horizontal and the two diagonals
(0, 45, 90 and 135 degrees).
C Non - maimum suppression
The edges are still colored to indicate direction. When the
estimates of the image gradient are given, a search is done to
determine if the gradient magnitude assumes a local maximum
in the gradient direction. Thus for example,

If the rounded angle is zero degrees the point will be

considered to be on the edge if its intensity is greater
than the intensities in the north and south directions,

If the rounded angle is 90 degrees the point will be

considered to be on the edge if its intensity is greater
than the intensities in the west and east directions,

If the rounded angle is 135 degrees the point will be

considered to be on the edge if its intensity is greater
than the intensities in the north east and south west
directions,

If the rounded angle is 45 degrees the point will be

considered to be on the edge if its intensity is greater
than the intensities in the north west and south east
directions
This is done by passing a 3 x 3 grid over the intensity map,
where a set of edge points, in the form of binary images is
obtained. This is referred to as "non - maximum suppression"
and is also called as "thin edges". In most cases it is
impossible to specify the threshold at which the intensity
gradient switches from being an edge, when it is traced
through the image and hysteresis. Hence, Canny used
threshold and hysteresis. This required two thresholds - high
and low. With this assumption, the edges are traced along the
continuous curves in the image. This allows us to follow a
faint section of a given line and to discard the noisy pixels that
do not constitute a line but are with a high intensity gradient.
Thus, this is done by applying high threshold. Finally the
edges are traced starting from here, with the help of the
directional information derived earlier. While tracing an edge,
the lower threshold is applied, which allows tracing faint
sections of edges as long as the starting point is found. Thus,
fnally a binary image is obtained.
From this binary image, each pixel is marked as either an
edge pixel or a non- edge pixel. Thus, the binary edge map
obtained in this way can also be treated as a set of edge
curves, which afer further processing can be represented as
polygons in the image domain.
D. Dif erential Geometric Formulation
A more refned approach to obtain edges with sub-pixel
accuracy is done by using the differential edge detection,
where the requirement of non-maximum suppression is
formulated in terms of second- and third- order derivatives
computed from a scale-space representation, proposed by
Lindeberg in the year 1998.
E Parameters
The Canny algorithm contains a number of adjustable
parameters, which can affect the computation time and
effectiveness of the algorithm.

The size of the Gaussian flter used in the frst stage

directly affects the results of the canny algorithm.
Smaller flter cause less blurring, and allow detection
of small, sharp lines. A larger flter causes more
blurring, smearing out the value of a given pixel over
a larger area of the image. Larger blurring radii are
more useful for detecting larger, smoother edges, for
instance , the edge of a rainbow
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31,2012 57

Thresholds: the use of two thresholds with hysteresis

allows more fexibility than in a single threshold
approach, but the general problems of thresholding
approaches still apply. A threshold set too high can
miss important information. On the other hand, a
threshold set too low will falsely identify irrelevant
information like noise as important. It is difcult to
give a generic threshold that works well on all
images. No tried and tested approach to this problem
yet exists.
III. PROPOSED METHOD
The new algorithm is proposed to overcome the above
said drawbacks using the classical image processing
techniques as shown in the fow diagram.
Proposed Algorithm:
Step 1: Read the given angiogram image, and convert it into a
matrix form where each pixel value is in the range from 0-255.
Step 2: Apply median fltering to remove noise.
Step 3: Take Histogram of the given input image.
Step4: Obtain a uniform histogram using histogram
equal ization or linearization technique.
Step 5: Repeat the above process again.
Step 6: 2D FIR flter is used to detect the edges of the
angiogram image.
In Fig.l.l, frst the input image is preprocessed using
median flter to remove noise and then the histogram of the
input angiogram image is obtained. Then by a technique called
histogram equalization, uniform histogram is obtained. Again,
the histogram of the histogram equalized image is obtained.
Finally, the edges of the vessel from the given angiogram
image is obtained.
Input Angiogram Image
Histogram Equalization
Figure I. I Flow Diagram
Thus, the above mentioned algorithm is used to detect the
edges of the vessel from the given angiogram image.
A. image Enhancement
Image Enhancement process consists of a collection of
technique that seek to improve the visual appearance of an
image. Thus the basic aim is to make the image look better.
The objective of enhancement technique is to process an
image so that the result is more suitable than the original
image for a particular application.
Image enhancement refers to accentuation or sharpening of
image features such as edges, boundaries or contrast to make a
graphic display more useful for display and analysis. Image
enhancement includes gray level and contrast manipulation,
noise reduction, edge crispening & sharpening, fltering,
interpolation & magnifcation, pseudo coloring and so on.
E. Histogram Equalization
A Technique which is used to obtain uniform histogram is
known as Histogram Equalization or Histogram Linearization.
Let r represent the grey levels in the image to be enhanced.
Assume r to be normalized in the interval [0, 1], with r =
representing black and r = 1 representing white.
For any value r in the interval [0,1], the image transformation
is given as,
S = T(r), OS r S 1
This transformation produces a level s for every pixel value r
in the original image. The transformation function T(r)
satisfes the following conditions,

T(r) is single -valued and monotonically increases in

the interval S r S 1 (ie., preserves the order from
black to white in the output image) and
OS T(r S 1 for S r S 1 (ie., guarantees that the
output gray levels will be in the same range as the
input levels)

The Inverse transformation is given as,

If the gray levels in an image can be viewed as

random variables in the interval [0,1], then Plr) and
Ps(s) denote the probability density functions of the
random variables r and s.

If Plr) and T(r) are known and r

l
(s) satisfes the
frst condition, then the probability density function
of the transformed variables is determined by the
gray level probability density function of the input
image and by the chosen transformation function.
C Edge Detection
The magnitude of the first derivative can be used to
detect an edge. The sign (zero crossing) of the second
derivative can be used to detect an edge. The same idea
can be extended into 2-D. 2-D derivatives should be
used. The magnitude of the gradient and sign of the
Laplacian are used.
Thus, a 2 D FIR filter is used to compute the result
using a two-dimensional correlation. The filter is rotated
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 58
180 degrees in order to perform the two-dimensional
correlation which basically involves the two-dimensional
convolution to detect the edges. Usually one dimension
process is carried in the literature but here a 20 FIR filter
is used to improve the results of the detection process.
IV. RSULTS
20 Coronary angiogram images were taken for testing out of
which four input MRI images are shown in Figure 4.1 for edge
detection. Each image had a size of 395x395 pixels and 72 dpi
for both the vertical and the horizontal resolution. The
algorithm was tested in MATLAB 7.12.0 on a Pentium IV PC.
Figure 4.1 Input Angiogram Image
10(
)0'
Figure 4.2 Canny edge detected output of the Angiogram Imagel

ll
Figure 4.3 Canny edge detected output of the Angiogram lmage2
.00
Figure 4.4 Canny edge detected output of the Angiogram lmage3
Figure 4.2-4.5 shows the output of the various edge
detected angiogram images using the canny edge detection
algorithm. The following shows the different steps to obtain
the edges of the blood vessel from the given angiogram image.
Figure 4.6 shows the histogram of the given input angiogram
images. Figure 4.7 and 4.8 shows the histogram equalized
image and its histogram. Finally Figure 4.9 shows the edges
of the blood vessel from the given angiogram image being
detected using the proposed algorithm.
The results show that the edges of the angiogram blood
vessel so detected using the proposed algorithm is more
effcient as compared to the canny edge detection algorithm.
This is evident from fgure 4.2 to 4.5 which uses the canny
edge detection algorithm as compared to fgure 4.9 which uses
the proposed algorithm. From the fgures one can easily
predict that the proposed algorithm can detect more complete
edges of the coronary angiogram blood vessels, with good
contrast.
z
i^^
'OC ::J nu
.. 11" " hI1'J

m
m
^ J^ i ::
Figure 4.5 Canny edge detected output of the Angiogram Image4
Figure 4.6 Histogram of the input image
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 59
,
Figure 4.7 Histogram equalized image
" t" - -, _; .[
,- -
III III
,
Figure 4.8 Histogram of Histogram equalized image
Figure 4.9 Edge detected Angiogram Image
V. CONCLUSION
In this paper, the proposed algorithm detects the edges of
the blood vessel from the given angiogram image using the
classical image processing techniques. The edges segmented
are accurate and clear as compared to the canny edge detection
and the steps involved to obtain the edges of the blood vessel
are simple and easy to implement. The results provide that the
proposed algorithm is effective and effcient in detecting the
edges. The future work will focus on developing an algorithm
for detecting the blocks and types of disease in the angiogram
image in a simpler way, as of the existing methods. Thus a
more generalized algorithm will be addressed in the near
future.
REFERENCES
[1] A. K. Klein, and A. Amin, "Quantities coronary
angiography with deformable spline models", IEEE
Trans. Med. Imag., vol. 16, 468-482, 1997.
[2] C. Kirbas and F. Quek, "A review of vessel
extraction techniques and algorithms, ACM
Computing Surveys", Vol. 36 (2): pp.81-121, 2004.
[3] D. Marr and E. Hildreth, "Theory of edge detection,"
Proc.R. Soc. Lond., vo1.l2, pp. 187-217, 1980.
[4] 1.Brieva, P.Ponce, " Evaluation of Segmentation
Algorithms for Coronary Angiography", proceedings
of the 29
th
Annual International Conference of the
IEEE, France, August 23-26, 2007.
[5] 1.Canny, "A computational approach to edge
detection," IEEE Trans. Pattern Anal. Machine
Intell., vol. 8, pp. 679-698,lune 1986.
[6] 1. S. Suri, K. Liu, L. Reden, and S. Laxminarayan, "A
Reivew on MR Vascular Image Processing
[7]
[8]
[9]
[10]
Algorithms: Acquisition and refltering: Part I,"
IEEE Trans. Irormation Technolog in Biomedicine,
voI. 6, no. 4, pp. 324-337, 2002.
Huertas, A. and Medioni, G., "Detection of intensity
changes with sub pixel accuracy using Laplacian-
Gaussian masks," IEEE Trans. On Pattern Analysis
and Machine Intelligence, PAMI,vol. 8, pp. 651-664,
1986.
Lee 1.S.1., Haralick R.M., and Shapiro L.G.,
"Morphology edge Detection," IEEE 1.Robot.
Automat, vol. 3, pp. 142-156,Feb 1987.
L. G. Brown, "A survey of image registration
techniques," ACM Com put. Surv., vol. 24, on. 4, pp.
352-376, 1992.
Lo Kwee-Seong, "Image Segmentation Methods for
Detecting Blood vessels in Angiography," Conf.
Control, Automation, Robotics and vision Singapore,
December 2006
[11] L. Zhou, M. S. Rzeszotarski, L. J. Singerman, and J.
M. Chokreff, "The detection and quantifcation of
retinopathy using digital angiograms,"IEEE Trans.
Med Imag., vol. 13, no. 4, pp. 619-626, Dec. 1994.
[12] Wenwei Kang, Ke Wang, Wanzhong Chen, Yong Li,
" Segmentation of Coronary Arteries Based on
Transition Region Extraction", 2n
d
International Asia
Conference on Informatics in Control, Automation
and Robotics, IEEE 2010.
[13] Y.H.Yu, et aI, "Study of Coronary artery
Segmentation from Coronary digital angiography",
Shandong Journal of Biomedical Engineering, vol.
21, pp. 5 -10, 2002.
[14] Zohaib Hameed and Chunyan Wang, "Edge detection
using Histogram Equalization and Multi-fltering
process",IEEE International Symposium on Circuits
and Systems,pp. 1077-1080,May 201l.
[15] Zulong Yu, Kaiqiong Sun, " Vessel Segmentation on
Angiogram using Morphology driven deformable
Model", 3r
d
International Conference on Biomedical
Engineering and Informatics, IEEE 2010.
[16] Z. W .Tang, H.Zhang, and G.S.Hu, "Morphological
multiscale vessel enhancement for coronary
angiograms", J Tsinghua Univ (Sci & Tech), vol. 46,
no. 3, pp. 418-420, 2006.
Mr.S.Sathish Kumar is pursuing his Ph.D
program at Sri Chandrasekhardendra
Saraswathi Viswa Mahavidyalaya University,
Kanchipuram. He completed his B.E degree
in Electronics and Communication
Engineering in the year 2001 and M.E
Applied Electronics in the year 2006, at
Arulmigu Meenakshi Amman College of Engineering,
Kanchipuram, Chennai, India.
He has 10 years of teaching experience and he is
currently working as a Senior Lecturer in the School of
Engineering, Asia Pacifc University College of Technology
and Innovation, Kuala Lumpur, Malaysia. He has published
and presented many papers in National and International
ISBN: 978-81-909042-2-3 2012 IEEE
IEEE-International Conference On Advances In Engineering, Science And Management (ICAESM -2012) March 30, 31, 2012 60
Conference in the area of Image processing. He has also
published papers in National journals. His research interests
are Image Segmentation in Angiogram images, Brain and
Liver images, Image Enhancement, Image Compression,
Image Retrieval Watermarking and Speech Detection and
Speech processing.
ISBN: 978-81-909042-2-3 2012 IEEE