Glaucoma Detection and Image Processing Approaches: A Review
Human recognition system based on retina vascular network characteristics
1. Human Recognition System based on Retina Vascular Network Characteristics
Chandrashekhar B.N Honnaraju .B
Sr. Lecturer, Dept of ISE Lecturer, Dept of CSE
NMIT, Bangalore-64 BGS, Mandya
eye retina. These are the optic nerve, the
Abstract macula and the vascular network. The
This paper proposes an efficient method reason to the selection was given by the
for Human Recognition System based fact that these characteristics remain
on Retina Vascular Network unchanged through years and
Characteristics. Humans recognize each degradations are possible to occur only
other according to their various because of eye diseases, such as
characteristics for ages. We recognize glaucoma and retinopathy. Human
others by their face when we meet them interference on the retina vascular
and by their voice as we speak to them. network is not an issue at present.
Identity verification (authentication) in Part from the proposed algorithm can
computer systems has been traditionally also be used to extract information
based on something that one has (key, about blood vessels network in retinal
magnetic or chip card) or one knows images. This information can be used to
(PIN, password). Things like keys or grade disease severity or as a part of
cards, however, tend to get stolen or lost automated diagnosis of diseases
and passwords are often forgotten or (Biomedical Systems).
disclosed. Human Recognition System based on
Human Recognition System based on Retina Vascular Network
Retina Vascular Network Characteristics, This detection system
Characteristics, This detection system can be used effectively to carry out
can be used effectively to carry out accurate authentication of a person.
accurate authentication of a person. Also this detection system can be used
Retina Vascular Network Identification in or suited for environments
Algorithm for Human Recognition. requirements requiring maximum
This system authorizes a person based security such as government military
on his retinal vascular characteristics. and banking.
The system takes fundus image of the The iris is the coloured ring of textured
person as input, performs pre- tissue that surrounds the pupil of the
processing and produces edge detected eye. Even twins have different iris
image. This resultant image is patterns and everyone’s left and right
compared with the images stored in the iris is different, too. Research shows
database. If the image exists, then the that the matching accuracy of iris
person is authorized, else unauthorized. identification is greater than of the
Retina scan is based on the blood vessel
Keywords-image analysis, image pattern in the retina of the eye. Retina
recognition, neural network, Image scan technology is older than the iris
segmentation, Retina, Optic disc, scan technology that also uses a part of
Macula, Otsu method the eye. The first retinal scanning
1. Introduction systems were launched by Eye Dentify
The identification procedure is based on in 1985.
three structural elements of the human
2. The retinal scanning systems are said to has reputedly never falsely verified an
be very accurate. For example the unauthorized user so far. The false
EyeDentify’s retinal scanning system
rejection rate, on the other side, is Image segmentation stage clusters the
relatively high as it is not always easy image into two distinct classes and
to capture a perfect image of the detection of candidate bifurcation and
retina.DNA testing. cross over points is done during the
2. Literature review third stage using the MCN and neural
“Detection of Vascular Intersection in Retina network technique.
Fundus Image Using Modified Cross Point “Locating the Optic Nerve in a Retinal Image
Number and Neural Network Technique” By Using the Fuzzy Convergence of the Blood
M. I. Iqbal, A. M. Aibinu, M. Nilsson, I. B. Vessels” By Adam Hoover , Michael
Tijani, and M. J. E. Salami. Goldbaum
This paper talks about the application of In this paper an automated method to
the knowledge of digital image locate the optic nerve in images of the
processing, fuzzy logic and neural ocular fundus has been given . Their
network technique to detect bifurcation method uses a novel algorithm we call
and vein-artery cross-over points in fuzzy convergence to determine the
fundus images. The acquired images origination of the blood vessel network.
undergo pre-processing stage for They evaluate their method using 30
Illumination equalization and noise images of healthy retinas and 51 images
removal.[5] Segmentation stage clusters of diseased retinas, containing such
the image into two distinct classes by diverse symptoms as tortuous vessels,
the use of fuzzy c-means technique, choroidal neovascularization, and
neural network technique and modified haemorrhages that completely obscure
cross-point number (MCN) methods[4] the actual nerve. On this difficult data
were employed for the detection of set, this method achieved 89% correct
bifurcation and cross-over points. MCN detection.
uses a 5x5 window with 16 The optic nerve is one of the most
neighbouring pixels for efficient important organs in the human retina.
detection of bifurcation and cross over The central retinal artery and central
points in fundus images. Result retinal vein emanate through the optic
obtained from applying this hybrid nerve,[3]supplying the retina with
method on both real and simulated blood. The optic nerve also serves as
vascular points shows that this method the conduit for the flow of information
perform better than the existing simple from the eye to the brain. Most retinal
cross-point number (SCN) method, thus pathology is local in its early stages, not
an improvement to the vascular point affecting the entire retina, so that vision
detection and a good tool in the impairment is more gradual. In
monitoring and diagnosis of diabetic contrast, pathology on or near the nerve
retinopathy. can have a more severe effect in early
A three stage bifurcation and cross-over stages, due to the necessity of the nerve
points detection in FI(Fundus Image) is for vision.
hereby presented. These stages are: Fundus Image: An image which is
image pre processing, image obtained from a fundus camera is
segmentation and bifurcation and cross- referred to as the fundus image[2] A
over point’s detection. The acquired fundus camera or retinal camera is a
image undergoes pre processing stage specialized low power microscope with
(A); for colour space conversion, an attached camera designed to
illumination equalization and noise photograph the interior surface of the
filtering using a 5x5 median filter. eye, including the retina, optic disc,
3. macula, and posterior pole (i.e. the pixels (e.g. foreground and background)
fundus) then calculates the optimum threshold
A typical fundus image[2]consists of separating those two classes so that
three important parts the macula, the their combined spread (intra-class
optic nerve and the blood vessels .The variance) is minimal. The extension of
optic nerve is one from which the blood the original method to multi-level
vessels seems to originate from and is thresholding is referred to as the Multi
the brightest part of the retina. Otsu method .
The fundus image is taken as the input
for pre processing and edge detection[6]
and based on which comparison is done
and other operations are carried out on
it.
Fig:Before thresholding
Fig:Typical Fundus Image
2.1Image Pre-processing
Image Pre-processing essentially Fig: After thresholding
contains two phases. These are image 2.1.2 Histogram Equalization
enhancement and image restoration. This method usually increases the
The idea behind enhancement global contrast of many images,
techniques is to bring out detail that is especially when the usable data of the
obscured or simply to highlight certain image is represented by close contrast
features of interest in an image. Image values. Through this adjustment, the
restoration techniques tend to be based intensities can be better distributed on
on mathematical or probabilistic models the histogram. This allows for areas of
of image degradation. lower local contrast to gain a higher
2.1.1 Thresholding contrast. Histogram equalization
accomplishes this by effectively
Thresholding is defined as the process spreading out the most frequent
in which individual pixels in an image intensity values.
are marked as “object” pixels if their The method is useful in images with
value is greater than some threshold backgrounds and foregrounds that are
value (assuming an object to be brighter both bright or both dark. In particular,
than the background) and as the method can lead to better views of
“background” pixels otherwise. bone structure in x-ray images, and to
The thresholding technique that we better detail in photographs that are
have used is Otsu Thresholding which over or under-exposed. A key
is an automated thresholding technique. advantage of the method is that it is a
Otsu's method is used to automatically fairly straightforward technique and an
perform histogram shape-based image invertible operator. So in theory, if the
thresholding, or, the reduction of a histogram equalization function is
graylevel image to a binary image. The known, then the original histogram can
algorithm assumes that the image to be be recovered. The calculation is not
thresholded contains two classes of computationally intensive. A
4. disadvantage of the method is that it is for an image formation model,
indiscriminate. It may increase the discontinuities in image brightness are
contrast of background noise, while likely to correspond to:
decreasing the usable signal.
Histogram equalization is a specific
case of the more general class of • discontinuities in depth,
histogram remapping methods. These • discontinuities in surface
methods seek to adjust the image to orientation,
make it easier to analyze or improve • changes in material properties
visual quality and
• variations in scene illumination.
The edge detection techniques used here
are sobel edge filtering ,prewitt edge
filtering technique which are briefly
described below.
2.1.3.1 Sobel Filter
The Sobel operator is used in image
processing, particularly within edge
detection algorithms. Technically, it is a
discrete differentiation operator,
Fig:Input Image
computing an approximation of the
gradient of the image intensity function.
At each point in the image, the result of
the Sobel operator is either the
corresponding gradient vector or the
norm of this vector. The Sobel operator
is based on convolving the image with a
small, separable, and integer valued
filter in horizontal and vertical direction
and is therefore relatively inexpensive
in terms of computations. On the other
hand, the gradient approximation which
Fig: Histogram Equalized Image it produces is relatively crude, in
particular for high frequency variations
2.1.3Edge Detection in the image.
Edge detection is a fundamental tool in 2.1.3.2 Prewitt Filter
image processing and computer The Prewitt operator is used in image
vision[6], particularly in the areas of processing, particularly within edge
feature detection and feature extraction, detection algorithms. Technically, it is a
which aim at identifying points in a discrete differentiation operator,
digital image at which the image computing an approximation of the
brightness changes sharply or, more gradient of the image intensity function.
formally, has discontinuities. At each point in the image, the result of
The purpose of detecting sharp changes the Prewitt operator is either the
in image brightness is to capture corresponding gradient vector or the
important events and changes in norm of this vector. The Prewitt
properties of the world. It can be shown operator is based on convolving the
that under rather general assumptions image with a small, separable, and
5. integer valued filter in horizontal and For Testing Phase:
vertical direction and is therefore
1. The user has to load his image
relatively inexpensive in terms of
using the Load Image option
computations. On the other hand, the
gradient approximation which it 2. The user then has to click the
produces is relatively crude, in compare option through which
particular for high frequency variations the new image is compared with
in the image images present in the database
3 Methodologies 4. Algorithm Steps
Initially the whole operation is divided The different modules algorithms are
into two phases the learning phase and as listed below:-
the testing phase. In learning phase we 1) Image Pre-processing
try to store images of an authorized • Histogram
person to the database so that it can be Equalization.
used for comparison when he comes for
authentication .The learning phase can 2) Edge detection
only be carried out by the administrator • Sobel mask
so he has to login first and then only he
can do further operations. • Prewitt mask
The testing phase is one where an
• Robert mask
authorized or an unauthorized person
comes for authentication so he first 4) Image comparison
inputs his image and then comparison 5) Saving Image to the database
of the image is done with the database 4.1.1 Image Pre-processing
to check his authenticity.The main user Image pre-processing is defined as the
requirements are for learning and stages before an image is processed in
testing phase are:- order to get an enhanced image through
For Learning Phase: which we can get better outputs. The
1. The user has to login if he image pre-processing steps used here
wants to store new images to is:-
the database through the Login 1) Histogram Equalization
option. Algorithm:
• Input the image
2. The user then has to specify
the size and the name which is • Convert to double dimensional
considered as the name of the array
output image that is to be • Compute Cumulative
stored in database using Image distributive function
Size option. • Compute Probability density
function
3. The user then has to load the • Output the image
image which he wants to store 4.1.2Edge Detection
in the database through the Sobel Mask
Load Image option. Algorithm:
• Input the image
4. Then the processing of the
image is done through the • Convert to double dimensional
Process option using which the array
output image is obtained. • Specify the threshold
• For each pixel in the image,
6. • Apply Sobel mask
• Compare the each resultant pixel
to threshold
• If greater than threshold make
pixel black, else white
• Output the image
The Sobel edge detector or the Sobel
filter can be implemented in three
orientations Sobel X, Sobel Y and Sobel
Fig:Implementation results of Prewitt Operator
XY. We have implemented all the three
orientations and found that Sobel XY
gave the best result. The output images
obtained from Sobel are as follows
Robert Algorithm
• Input the image
• Convert to double dimensional
array
• Specify the threshold
Fig: Implementation results of Sobel operator • For each pixel in the image,
Prewitt Mask • Apply Robert mask
Algorithm: • Compare the each resultant pixel
• Input the image to threshold
• Convert to double dimensional • If greater than threshold make
array pixel black, else white
• Specify the threshold • Output the image
• For each pixel in the image,
• Apply Prewitt mask
• Compare the each resultant pixel
to threshold
• If greater than threshold make
Fig: Implementation results of Robert Operator
pixel black, else white 5 Image Comparisons
• Output the image
The Prewitt edge detector or the Prewitt Learning Phase Algorithm:
filter can be implemented in three • Input the image
orientations o Prewitt X,Prewitt Y and • Convert to double dimensional
Prewitt XY. We have implemented all array
the three orientations and found that • Apply pre processing methods
Prewitt XY gave the best result. The • Apply edge detection
output images obtained from Prewitt are
• Compare the edge detected
as follows
image with the images in the
database
• If the image exists then discard
it and print error message
• Otherwise save it in the database
• End
Testing Phase Algorithm:
• Input the image
7. • Convert to double dimensional
array
• Apply pre processing methods
• Apply edge detection
• Compare the edge detected
image with the images in the
database
• If the image exists then print
authentication success message
• Else print authentication failure
message
• End
Saving Images in database
Algorithm:
Input the image 6. Conclusion
• Convert the image into two In this paper we illustrate a complete
dimensional array Retina Vascular Network Identification
• Apply pre processing and edge Algorithm for Human Recognition.
detection This system authorizes a person based
on his retinal vascular characteristics.
• Then convert the obtained two
The system takes fundus image of the
dimensional array into a raw
person as input, performs pre-
image
processing and produces edge detected
• Store the raw image with the image. This resultant image is
persons name obtained from compared with the images stored in the
input database. If the image exists, then the
• Then append the name to a file person is authorized, else unauthorized.
required for comparison We have also implemented some
5. Results fundamental pre-processing techniques
Experiment – A raw fundus image is such as Histogram Equalization and
given as input and then processing Image thresholding to alter certain
option is clicked which produces the characteristics of the image. For edge
following output which is compared detection we have used the various
with the database if it exists then a masks such as Robert mask, Prewitt
success message is shown or is the mask and Sobel mask. These filters are
image does not exist in the database valuable in detecting edges of various
then a failure message message is parts in the image.
shown We have provided the administrator an
option to add and remove the authorized
person to the database. Where as a
person other than administrator can only
check whether he is authorized or not.
Access rights to perform adding and
removing a person from database is
8. given to the administrator via username 2. A. Can, H. Shen, J. N. Turner, J. L.
and password. Tanenbaum, and B. Roysam, “Rapid
automated tracing and feature extraction
References from retinal fundus images using direct
1A. Pinz, S. Bernogger, P. Datlinger, exploratory algorithms,” IEEE
and A. Kruger, “Mapping the human tansactions on Information Technology
retina,” IEEE Transactions on Medical in Biomedicine, vol. 3,no. 2, pp. 125–
Imaging, vol. 17, no. 4, pp. 606–619, 137, June 1999.
1998. 3. Locating the Optic Nerve in a Retinal Image
Using the Fuzzy Convergence of the Blood
Vessels By Adam Hoover, Michael Goldbaum
4. Detection of Vascular Intersection in
Retina Fundus Image Using Modified
Cross Point Number and Neural
Network Technique By M. I. Iqbal, A.
M. Aibinu, M. Nilsson, I. B. Tijani,
and M. J. E. Salami.
5. A fuzzy impulse noise detection and
reduction method chulte,
S.Nachtegael, M.; De Witte, V.;
Van der Weken, D.; Kerre,
E.E.Dept. of Appl. Math. &
Comput. Sci., Ghent Univ., Gent,
Belgium
6. Local scale control for edge
detection and blur estimation IEEE
Trans. Pattern Anal. Mach. Intell., 20
(7) (1998), pp. 699–716
7. An improved Sobel algorithm based
on median filter hunxi Ma; Lei
ang; Wenshuo Gao; Zhonghui
Liu;
Digital Media Dept., Commun.
Univ. of China, Beijing, China
8. design of an image edge detection
filter using the sobel operator nick
kanopoulos, member, ieee,nagesh
vasanthavada, member, ieee,androbert
baker
9. given to the administrator via username 2. A. Can, H. Shen, J. N. Turner, J. L.
and password. Tanenbaum, and B. Roysam, “Rapid
automated tracing and feature extraction
References from retinal fundus images using direct
1A. Pinz, S. Bernogger, P. Datlinger, exploratory algorithms,” IEEE
and A. Kruger, “Mapping the human tansactions on Information Technology
retina,” IEEE Transactions on Medical in Biomedicine, vol. 3,no. 2, pp. 125–
Imaging, vol. 17, no. 4, pp. 606–619, 137, June 1999.
1998. 3. Locating the Optic Nerve in a Retinal Image
Using the Fuzzy Convergence of the Blood
Vessels By Adam Hoover, Michael Goldbaum
4. Detection of Vascular Intersection in
Retina Fundus Image Using Modified
Cross Point Number and Neural
Network Technique By M. I. Iqbal, A.
M. Aibinu, M. Nilsson, I. B. Tijani,
and M. J. E. Salami.
5. A fuzzy impulse noise detection and
reduction method chulte,
S.Nachtegael, M.; De Witte, V.;
Van der Weken, D.; Kerre,
E.E.Dept. of Appl. Math. &
Comput. Sci., Ghent Univ., Gent,
Belgium
6. Local scale control for edge
detection and blur estimation IEEE
Trans. Pattern Anal. Mach. Intell., 20
(7) (1998), pp. 699–716
7. An improved Sobel algorithm based
on median filter hunxi Ma; Lei
ang; Wenshuo Gao; Zhonghui
Liu;
Digital Media Dept., Commun.
Univ. of China, Beijing, China
8. design of an image edge detection
filter using the sobel operator nick
kanopoulos, member, ieee,nagesh
vasanthavada, member, ieee,androbert
baker