Vol.7 (2017) No.

5
ISSN: 2088-5334

Feature Matching in Iris Recognition System using MATLAB

Imran Naguru#1, Narendra Kumar Rao B#2
#
Department of Computer Science & Engineering,
Sree Vidyanikethan Engineering College (Autonomous), Tirupati, Chittor District-517102, A.p, India.
E-mail:[email protected]; [email protected]

Abstract - Iris recognition system is a secure human authentication in biometric technology. Iris recognition system consists of five
stages. They are Feature matching, Feature encoding, Iris Normalization, Iris Segmentation and Image acquisition. In Image
acquisition, the eye Image is captured from the CASIA database, the Image must have good quality with high resolution to process
next steps. In Iris Segmentation, the Iris part is detected by using Hough transform technique and Canny Edge detection technique.
Iris from an eye Image segmented. In normalization, the Iris region is converted from the circular region into a rectangular region by
using polar transform technique. In feature encoding, the normalized Iris can be encoded in the form of binary bit format by using
Gabor filter techniques. In feature matching, the encoded Iris template is compared with database eye Image of Iris template and
generated the matching score by using Hamming distance technique and Euclidean distance technique. Based on the matching score,
we get the result. This project is developed using Image processing toolbox of Matlab software.
Keywords - iris recognition; biometric identification; feature matching; iris normalization; image acquisition feature encoding; iris
segmentation.

I. INTRODUCTION
Iris recognition system is a secure human authentication
in biometric technology. This is one of the most accurate,
reliable and leading technology for providing security. First
Iris recognition system was proposed in 1939 by Dr. Frank
and it has been implemented in 1990 by Dr. Daugman. At
present Iris Recognition System has become more accurate
and reliable biometric identification technology. Iris
Recognition System consists of five stages to define or
authenticate the user. They are Feature matching, Image
acquisition. Feature encoding, Iris normalization and Iris
segmentation. Image acquisition is the first stage in Iris
recognition system. In this stage, first we capture two eye
Images from CASIA database. The captured eye Images
should be with good quality and high resolution for Image
processing. Iris segmentation is the second stage in Iris
recognition system. In this stage, Iris part is segmented from
an eye Image by using canny edge detection technique and
Hough transform technique. The third stage in Iris Fig. 1. Working flow of Iris Recognition System
recognition system is Iris normalization. In this stage, the
segmented Iris is converted from the circular region into a Feature matching is final stage in Iris recognition. It is
rectangular region by using polar transform technique and defined the integrity match score based on the two encoded
rubber sheet model. In Iris recognition system, the Feature Iris Images of Iris templates. Based on that matching score
encoding is fourth stage. In this stage, the normalized Iris is we get the result.
encoded in the form binary bit format and generates an Iris
template by using Gabor filter technique.

1748
 II. MATERIAL AND METHOD Sometimes system fails to detect Iris from an eye Image due
The Iris recognition system is defined in 5 stages. They to several eyelashes and deals with statistical interference
are, Feature Matching, Iris Normalization, Iris Segmentation, and threshold. Eyelid can deal by implementing hyperbolic
Image Acquisition and Feature Encoding. The above Figure Hough Transform. In the result, suggested wavelet features
shows the Iris recognition system process. and unique codes are generated for an individual eye Images.
Iris segmentation can improve the Iris recognition system
A. Methods for Image Acquisition dramatically.
Samuel et. al. [2] defined horizontal and vertical
In Image acquisition stage, we capture the Iris Images derivatives in Iris recognition by developing Circular Hough
using different cameras or from the databases like CASIA transform technique. Provides a very accurate and reliable
and UBIRIS. The captured eye Image must have good biometric technology to identify or authenticating the person.
quality along with a high resolution for Image processing. In Iris consists of different Iris patterns for comparing one
table 1, we mentioned the different Image acquisition person with the other person. Iris templates are depends on
cameras with format and Resolution. Particularly Iris and reliability human Iris up to some level, which are identical
pupil part must have good quality with high resolution for twins. The eyelashes and eyelids are indicated by the lower
Image processing. Some Iris Images are available in the and upper parts of Iris boundary, which are removed
form of datasets which are freely available in the database. perfectly by using Hough transform technique. The accuracy
The Iris datasets are UBIRIS, CASIA, etc. This project of the system is calculated by converting Iris region into
implemented by using CASIA database. biometric template matching. The accuracy of proposed
system is 95.6%.
TABLE I Sharma et al., [22] mainly focused on Hough transform
CAMERAS USED FOR IMAGE ACQUISITION method and developed efficient algorithm for Iris
Cameras used for Image
recognition in Iris segmentation stage. Over many years the
Format Resolution Iris recognition provides a reliable verification and
Acquisition
identification for an individuals and increased the biometric
Nikon E5700 camera JPEG 400*300
system accuracy. This process can captures an eye Image
Canon EOS 5D JPEG 800*600 from user and segmented the Iris from an eye Image and
Iridian LG EOU2200 TIFF 640*480 normalized the segmented Iris using rubber sheet technique
later we encode the normalized Iris by using Gabor filter
AD-100 Iris Guard Dual
Eye autofocus camera
BMP 1280*960 technique and matching score is defined to the encoded Iris.
The performance of this system depends on the segmentation
SONY DXC-950P 3CCD and normalization. Firstly the Iris Images are selected from
PNG 576*768
Camera
the database after removed the noises from an eye Image
LG Iris Access 2200 BMP 320*280 later the pupil and Iris boundaries are detected. The Iris
Panasonic BMET100US region that is segmented for Iris normalization in order to
BMP 320*280
Authentic cam minimize the dimensional inconsistencies. For Image
processing and pattern recognition used Biometric
B. Methods for Iris Segmentation recognition algorithms. Based on histogram threshold and
In segmentation stage, Iris part is segmented from an eye gamma correction defined matching score to generate the
Image using different segmentation techniques. The Iris result. By using Image processing toolbox of Matlab
Segmentation techniques are Hough transform, Integro- software this project is developed.
differential operator and gradient based edge detection, Amit et al., [23] applied Hough transform technique for
removal of eyelash & eyelids and bisection methods. To detecting eyelashes, eyelids on Iris Image, but in order to
segment the Iris, used Hough transform technique from an improve the accuracy need to eliminate noise factors from
eye Image. the eye Image. Based on the physical characteristics and the
1. Hough Transform Technique human behaviour, the Biometric system is developed. Iris
has a unique structure, which contains high accuracy and
Caroline Houston [1] defined Iris segmentation is will be stable for each individual. A person recognized based
performed by using Canny edge detection and Circular unique features of Iris patterns. In the eye Image the Iris is
Hough Transform techniques. Iris is segmented by using found based on Iris patterns uniqueness. Iris recognition is
pupil and Iris center offset and Iris stretch methods. In Iris defined in four stages; they are acquiring an eye Image from
segmentation, Iris part is detected two circles from an eye user, segmenting Iris, Feature extraction and Feature
Image, one is Iris boundary and other is sclera boundary. matching. To detect the eyelids and eyelashes, eyelashes
Hough transform is used to detect the centers and radius of detection algorithm is used. To identifying the lines and
two boundaries. In General Hough transform technique is circles of particular radius from an eye Image used Hough
used to detect geometric shapes in the form of circles, lines, transform technique. The performance of the Iris recognition
parabola, and hyperbolas. For detecting circle of Iris, we is improved based on Iris Image of canny edge detection
used Circular Hough transform technique from an eye Image. algorithm.
Iris edges are detected by using a Hough transform Verma et al., [24] uses Hough transform technique to
technique. Iris segmentation technique performs well on segment the Iris by eliminating noisy and detecting Iris and
most Iris Images for detecting Iris from an eye Image, pupil boundaries from an eye Image. Iris recognition system

1749
is the most accurate and reliable biometric identification template results will be corrupted. To avoid this drawback,
system in the present situation. This system captures an Daugman’s integrodifferential operator method is used.
Image of the human eye then the Iris Image is to be 3. Eyelash Detection Technique:
processed for segmentation, normalization and for the Yuan et al. [5] proposed a method for detecting eyelashes
feature extraction process. For detecting the correct Iris need accurately. Eyelashes are two types; one is separable eyelash
to remove the pupil noises reflection and the eyelashes and the other is multiple eyelashes. The main intention is to
present in Iris region correctly and accurately in Iris find out the eyelash without affecting outside and inside
Segmentation stage. In Iris Normalization, the Segmented boundaries of Iris Image. In this, eyelash detection is a very
Iris is converted from circular region into rectangular region significant part in Iris segmentation process to get an
by using Rubber Sheet Model in order to minimize the accurate and desired results. Firstly, he is used local intensity
dimensional inconsistencies in between the Iris regions. The minimum algorithm to detect the eyelashes, later it is
normalized Iris region is converted into binary format and detected based on the template of standard deviation and
form a bit-wise biometric template produces the output like mean. Iris region consist of high-intensity values than
phase quantization by using the 1D Log-Gabor filters Separable eyelash region. To obtain eyelash points, simple
technique. By using Iris template we calculated matching threshold can be used. Multiple eyelashes are overlapped in
score for authentication of an Iris Image. a small region. To detect multiple eyelashes, the standard
Naveen et al., [25] presents the novel techniques in order deviation and means are used. In the result, it is proved that
to create the Iris recognition system for the analysis and used the eyelash noise detect ed with an proposed approach
fusion mechanism. To detect the Iris boundaries used both accurately.
Circular Hough Transform and Canny Edge detection Zemao et al. [6] proposed an eyelash recognition
techniques in the eye Image. In recent day’s automatic algorithm based on adaptive threshold. It is used for
identification provides different feature or the characteristic reducing eyelash influence on Iris recognition rate. Adaptive
of the individuals in Iris recognition system which is more threshold is used to segment the eyelashes, according to gray
accurate and reliable for biometric identification. Wavelet distribution of the parabolic pixel point, upper eyelid
transform technique forms a feature vector by determining parabolic moved down. In the results, we can observe that
Iris patterns. The two Irises are similar when quantized the improvement of the threshold is better in order to detect
vectors is compared with the Hamming Distance and proved eyelash edges by comparing with traditional eyelash
that system is quite effective. detection algorithm, it is not necessary to set up more
2. Integro-differential operator Technique parameters for the algorithm. It can embed with simplicity to
maximize the result of eyelash segmentation. The algorithm
R.B. Dubey et al. [3] proposed an approach for automatic reduces the three-dimensional searching space of the
segmentation that is based on different algorithms in Iris traditional algorithm with one dimension, So that the speed
recognition system. It can localize the circular pupil region has been increased significantly. The eyelid detection
and Iris region by removing eyelashes and eyelids. The Iris algorithm is used for the eyelash threshold segmentation of
recognition System is defined as most precise and reliable the adaptive generation, and compared with the traditional
biometric authentic system. Patented algorithm is used for algorithm which doesn’t need to set up more parameters to
producing perfect Iris recognition rates in Iris recognition. segment effectively. The eyelash maximization can improve
These algorithms are based on linear search methods which further aspects like the gray morphological operation. The
are extremely slow to identify the process and raise the false palpebral eyelash interference will be better to remove seek
acceptance rate. The Iris segmentation is achieved through for adaptive structure elements. The adaptive generation
by finding Iris boundaries as well as outer and inner algorithm can improve the threshold to detect eyelash edges.
boundaries of Iris region. The Iris region is approximately
4. Eyelid Detection Technique:
in between two circles, one is the pupil and other is sclera. In
Adam et al. [7] proposed two steps algorithm. One is
the eye except Iris remaining parts are normally removed
possible edges are detected by selecting the inside area of the
from the lower and upper parts of the Iris region. Depending
Iris. To refine parameters of a larger area, applied
on the quality of eye Image, defined the success rate of Iris
maximization at every parabola. The segmentation can be
segmentation. Poor recognition rates will induce due to
done manually. The performance evaluation of segmented
noisy or falsely data. This work is mainly focus on exact Iris
Iris is compared with the manual segmentation using
Segmentation method for improving better biometric
proposed method. More than 90% of the cases are satisfied
authentication system in widespread application areas. The
with the results. In this, based on parabolic curve fitting
proposed method achieves in 12.056 seconds of computation
proposed a new eyelid localization algorithm. Due to Iris
time in Iris segmentation but Hough transform method
texture, low contrast, the false detection is calculated by
achieved in 248.23 seconds.
dealing with eyelashes. On a restricted Image area, we used
Maeva et al. [4] proposed an improvement method for Iris
the edge detection method. Parabolic arcs are estimated
segmentation process. Firstly, the Iris boundaries are
based on edges. A gradient maximization is applied on the
identified by using Integro differential operator method and
parabola curvature of a larger Image area is restricted by
identified eyelashes and eyelids are removed by using Linear
using maximization of every candidate. Accurately is
Hough transform technique based on threshold. Iris
demonstrated by detecting the final boundaries.
segmentation method is improved in order to find and
Ismail et al. [8] proposed a simple algorithm for
remove the eyelids and eyelashes with a precise method.
identifying boundaries of eyelids in an eye Image and
Due to poor recognition of Iris regions the biometric

1750
efficient eyelid detection algorithm is applied on eye Image. track can be extracted by using Gabor filter technique and
Under less constrained Imaging conditions, the Iris Images forms a binary bit code (0 and 1). The extracted Iris codes
are gained. Based on live-wire technique developed a are used for enrolment to identify the person. The proposed
detection algorithm for detecting eyelid. The eyelid detection method can be defined in two steps. In step 1, Iris patterns
algorithm main advantage is very simple computationally are automatically detected with Non-Linear positions by
because the detection algorithms is compared with other changing illumination conditions in a radial direction when
prior eyelid algorithms. The Iris Image is determined in contracted or dilated the pupil. In step 2, differentiated
between outer Iris boundary and each eyelid by an two tensile properties of Iris patterns individually and this
intersection points. The two intersection points are detected method determines based on the user dependent method of
by using live-wire technique. In this, less constrained Iris Non-Linear dynamic track allocation, in which the Iris
Images accuracy is improved in Iris segmentation stage. recognition accuracy is greatly improved. Based on the
5. Non –Linear Circular Active Contour Technique: linear rubber sheet model the proposed method will have
Satish et al. [9] improved the efficiency and accuracy to high accuracy and efficiency than conventional methods
the UBIRIS and CASIA V4.0 Iris databases with a proposed which will use consistent track allocation.
technique and also uses Geodesic Active Contour (GAC) Zhenan et al. [12] defined a novel method for correcting
technique for segmenting Iris from eye structures. Active Non- Linear Iris Deformation. In this, Iris Images are
colours can assume shapes and segment multiple objects corrected by an addictive deviation between Non-Linear and
simultaneously. In the survey most segmentation models Linear Iris stretch. The corresponding distance under linear
assumed that the eyelid, limbic and pupillary boundaries are stretch plus an addictive deviation is calculated in between
elliptical or circular in shape. The constant unique pattern of an Iris region to pupil boundary. In the result, the
the Iris makes it a rich biometric characteristic for human effectiveness algorithm is demonstrated based on two
recognition. The 1st stage of Iris recognition is to keep a part databases with non-linear deformation. In the survey, lower
of the actual Iris region in a digital eye Image. Iris Equal Error Rate (EER) is defined by using proposed Iris
segmentation stage is significant to achieve Iris pattern deformation algorithm by comparing with two Non-Linear
which are falsely represented in an Iris recognition will and linear normalization methods. It can make the system
corrupt the biometric templates. In the result, less more robust in a real environment. The proposed method is
recognition rates of data is generated. more efficient than other approaches.
Mohammed et al. [10] proposed a scheme for a non-ideal 2. Polar Transform Technique:
Iris Images to find the exact pupil boundary. The Iris
Anicham et al. [13] used different techniques for Iris
recognition system affected significantly in non-ideal Iris
recognition system. It is consists of five stages. They are
Images, and also affected the segmentation accuracy. To
Feature matching, Iris normalization, Iris segmentation,
segment the pupil shape proposes an accurate and efficient
Image acquisition and Feature encoding. Each step is
method by using active contour and morphological operation
defined by using various techniques of Iris recognition
methods. The rough pupil boundary is detected by using
system. To segment the pupil and Iris boundaries proposed a
Morphological operations method. To identify the actual
method that are detected all boundaries of Iris by using
boundary, Active contour method is applied. The inner Iris
Hough Transform technique from an eye Image. Daugman’s
boundary is segmented by proposed method in real shape,
rubber sheet model defined the Iris normalization from
which is more accurate and more reliable to contribute a rich
segmented Iris. The active contour transform technique is
Iris recognition. The main advantage of the active contour is
extracted or encoded the normalized Iris. To classify the Iris
to achieve more accuracy and efficiency by generating the
Image SOM, LDA techniques are used. Iris recognition is a
more accurate active contour of pupil boundary than the
secure and sufficient than using the user id and security code
circular and elliptic approximation. In the result, high-level
and the interlopers prevented by the malicious action. This
accuracy is found on CASIA V.4 database.
experiment can be executed by using MATLAB.
C. Methods for Iris Normalization Mayank et al. [14] proposed an algorithm to increase the
In Iris recognition system, the Iris normalization is third speed and the accuracy of Iris recognition, indexing, match
step. In this, the localized Iris is converted from the circular score fusion and quality enhancement, for Iris segmentation.
region into a rectangular region. It has some techniques to In this, the modified Mumford–Shah functional is used to
normalize the segmented Iris. The normalization techniques propose Curve evolution approach to segment the Iris.
are the non-linear model, polar transform and rubber sheet Multiple enhanced versions of the Iris Images are produced
model. In this, we used Rubber Sheet technique and polar in Iris Segmentation by using Different improvement
transform technique to normalize the Iris. algorithms which are concurrently applied. SVMBLA
enhanced the local region from global region. The high-
1. Non-Linear Technique
quality Iris Image is created by using SVMBLA. The high-
Chul et al. [11] proposed non-linear tensile properties of quality Iris Images are encoded from two distinct features.
Iris patterns. It can detect outer and inner regions of Iris by Euler number and 1-D log-polar Gabor filter encode the
eliminating specular reflections, eyelash regions and eyelid. local topological feature and global textural features. An
To Normalize the Segmented Iris, normalization technique is intelligent fusion algorithm is improve the performance of
applied. Which converts the Iris from the circular region into matching score and textual score in Iris recognition system
a rectangular region. In Iris region, we can able to allocate and reduces the FRR. Supports accurate and efficient Iris
the tracks Non-Linearly and dynamically, in which each identification in an indexing algorithm. Performance of

1751
Identification is verified and validated by using a proposed transform method for detecting the feature points from
algorithm and other algorithms are comparing with proposed segmented Iris by using log-Gabor filter technique to
algorithms by using the UBIRIS, ICE 2005, and CASIA produce a rotation invariant descriptor at every detection
Version 3 Iris database. point. In feature matching, proposed a technique based on
3. Rubber Sheet Model technique: PFM method and compared with 2 sets of feature points by
using local feature point and position of every point. In the
Qureshi et al. [15] proposed conventional Iris recognition result, achieved more accuracy percentage in curve let
method with three serious modifications which are very transform than log-Gabor filter curve let transform.
efficient and secure for Iris recognition. To Segment the
pupil from eye Image by using canny edge detection E. Methods for Feature Matching:
technique and which is specially designed for handling pupil In Iris recognition system, The Feature matching is last
occlusion problems. Rubber sheet model is used for feature stage. In this, feature matching is defined by the integrity
extraction, which can encode the normalized Iris into binary match score based on the two encoded Iris Images. It has
bit format. Feature matching determines the matching score. some techniques to define the match score. The feature
In this, some serious modifications are done to the rubber matching techniques are hamming distance, Euclidean
sheet model and k-nearest method for achieving accuracy distance and K-NN classifier.
and efficiency in Iris recognition system. The average Iris 1. Euclidean Distance Technique
recognition time is found as 0.8254 sec, which is very less Win et al. [19] used different methods for Iris recognition
when compared to state of other techniques. system and defined Feature matching, Iris normalization, Iris
Rana et al. [16] proposed Daugman’s rubber sheet model segmentation, Image acquisition and Feature encoding
based on Independent Component Analysis (ICA) and stages for Iris recognition system by using different
Principle Component Analysis (PCA). Iris recognition is a techniques. In Image acquisition, Eye Images are captured
fast biometric technology to authenticate or identify the from CASIA database and eyelashes, eyelids are removed
person based on unique pattern of Iris recognition. The from an eye Image by using Hough transform method in Iris
proposed algorithm increases the accuracy and reliability by recognition. In Iris normalization, the Iris part is normalized
comparing over two hundred million of Iris Images. A by using rubber sheet model to minimize the dimensional
working prototype of techniques is used to implement the inconsistencies between Iris regions. In feature extraction,
Iris recognition. In this work used three different algorithms one- dimensional object detection method is used. In feature
in order to increase the efficiency of Iris recognition. The matching, Euclidean distance method is used and compared
algorithms are Daugman’s rubber sheet model, PCA and with two Iris textures. In the result achieved more accuracy
ICA. The UBIRIS database is used to capture the Iris Images, in Iris segmentation and feature extraction by reducing the
in which each database there are 230 sets of Images by computational time. The performance of Iris segmentation
consists of 5 Images in each set of a single person. In the and feature extraction is defined based on accuracy of Iris
result, efficient results are obtained by comparing with recognition.
existing methods. Sahoo et al. [20] proposed a feature matching technique
D. Methods for Feature Encoding: for Normalized Iris based on wavelet decomposition of
In Iris recognition, the Feature Encoding is fourth stage. optimized Iris. The unique Iris patterns of the people, which
In this stage, the encoded binary bits are formed by are highly stable from starting birth to death duo to this
normalizing the Iris. It has some techniques for encoding the defined Iris recognition system. For this we proposed many
normalized Iris. The techniques are Gabor filter, Cumsum methods and techniques from last few years. In this, the
based CPA and Ridge let. In this, we used Gabor filter evaluation of normalized correlation is an optimum sub band
technique for encoding normalized Iris into binary bit format. decomposition, the accuracy is improved by determining the
1. Gabor Filter Technique Iris recognition system and made the robustness and simpler
Avila et al. [17] proposed Hamming distance and Gabor by increasing the Iris recognition system. The matching
filter techniques for feature matching and feature encoding. score enhances the selective matching Iris set and reduces
To develop biometric technique used Hamming distance and FRR of registered user and FAR of the imposter in the
Gabor filters technique to describe different approaches. The recognition system. In the result, reliability and accuracy of
wavelet transform is applied for zero-crossing representation Iris recognition system are improved efficiently.
and Iris signatures are defined in two ways: one is based on 2. Hamming Distance Technique
annular region and another is based on a circle of the Iris and Hesham et al. [21] proposed an algorithm to generate a
also compared with some other metrics in the results. Iris template for Iris recognition system by using most
Proposed approach is a scale invariant, a translation and consistent feature data. A new metric measure formula is
rotations are also worked in 0 crossing representation of used to calculate hamming distance in Feature matching
wavelet transform. In the result, the Classification up to and which the noisy level presented in the considered images.
99.6% of Success Rate is achieved and down to 0.12% of In this, to increase the accuracy of an Iris recognition system
Equal Error Rate achieved and contained low false rejection required many eye Images of same person and to generate
rates, null false acceptance rates. final Iris template used an algorithm in which base templates
Dhinesh et al. [18] defined the Feature Matching, Feature are mixed with the given Iris templates set. In this work,
Extraction and Iris segmentation. In Iris segmentation, to revealed a reduction in size by using 45 persons of 450
detect the outer and inner boundaries proposed a method for Images from MMUI database almost 78% and increases in
Iris curves. In feature extraction, proposed a curve let

1752
verification speed around 80% while maintaining 99.7% segmented Iris is converted from the circular region into a
accuracy of matching. rectangular region by using polar transform technique in Iris
Gayathri et al. [26] used hamming distance measurement normalization stage. In feature encoding, the normalized Iris
to authenticate the users by comparing quantized vectors. In can be encoded in the form of binary bits by using Gabor
this, discussed Feature matching, Iris normalization, Iris filter techniques. In feature matching, the encoded binary bit
segmentation, Image acquisition and Feature encoding are form of Iris template is compared with database eye Image
based on user Iris. Circular Hough transform methods are of Iris template, based on the matching score of both Iris
used to identify the Iris boundaries in eye Image. Image templates output is displayed.
registration technique is used to normalize the segmented
Iris. For Iris Image registration, correlation base method is
used. Registration method adjusts points with corresponding
points by replacing one object view with another object view
to determine the geometrical transformation. In feature
encoding phase, the normalized Iris is encoded by using 2-D
Gabor filter technique. In this paper developed a new Iris
recognition system by using phase correlation technique to
compare the two eye-Images. These techniques has been
calculated individually for several different applications. In
the result, the proposed algorithm increases the security by
using encryption techniques and encourages the performance
of Iris recognition system.
Prashanth et al. [27] described an efficient biometric
security based on Iris recognition system. Iris Images are
verified or identified by using Integral Wavelet Transform
(IWT) technique. Iris recognition is one of the most secure Fig.2. Working process of Iris recognition system
and authentic method for human recognition by comparing
with other biometrics. Integro–Differential operator method Steps: Iris Recognition System
is used to identify the Iris and pupil boundaries in an eye
Image. Extracted the Normalized Iris by using Discrete Input: Eye Image
Wavelet Transform (DWT) and IWT. For Feature matching Output: Authenticating the Iris
of two Iris feature vectors used Hamming Distance
Step 1: Input Eye Image
technique. In the result, we detected that the values of Step 2: Iris segmentation is done by using
computation time, ERR, FRR and FAR are required to Hough transform and canny edge detection
improve Human Iris Recognition system as compared with technique.
the IWT to DWT. The Iris recognition system is Step 3: Segmented Iris converted from the
implemented by using DWT and IWT. The efficiency and circular region into a rectangular region by
computation speed is improved based on Iris recognition using rubber sheet technique.
system. The mean of Hamming Distance is calculated for Step 4: The normalized Iris is converted into
making decisions and selected only one matched Image from binary bit format by using Gabor Filter
Technique.
out of seven sets of Iris Images. The efficiency algorithm is Step 5: calculated matching score for input
improved the Iris recognition system. Iris template and database eye image of
another input Iris template by using Hamming
III. RESULTS AND DISCUSSION distance.
Step 6: perform left and right shift
The Iris recognition system is defined in five steps. They
operations to get similarity in between two
are Feature matching, Iris normalization, Iris segmentation, images.
Image acquisition and Feature encoding. In literature survey Step 7: decision is obtained like Match/No
we defined all techniques of Iris recognition system based match based on matching score.
each stage. In this, Hough transform technique is used for
Iris Segmentation, for normalizing the Iris used rubber sheet B. Case Study:
model, for feature encoding used Gabor filter technique and The Iris recognition system is defined in five stages. They
for feature matching used Hamming distance technique. are Feature matching, Iris normalization, Iris segmentation,
A. Block Diagram Image acquisition and Feature encoding. For example, Fig. 1
shows Iris recognition system process. These all stages are
Iris recognition system consists of five stages. They are explained clearly below.
Feature matching, Iris normalization, Iris segmentation,
Image acquisition and Feature encoding. In Image 1. Image Acquisition
acquisition, the eye Image is captured from databases of eye
Images, the Image must have good quality with high In Iris Recognition System, Image acquisition is first
resolution to process for next steps. In Iris segmentation, stage. In this stage, we capture the Iris Images using
detected the Iris region from an eye picture. The Iris part is different cameras or from database. The capture eye Image
detected by using Hough transform technique. The must have good quality with high resolution for Image

1753
processing. The different Image acquisition cameras are from the Hough space and edge map. From each edge point
mentioned in table 1 with format and resolution. Particularly the parameters of the circles are passed. The radius r, centre
Iris and pupil part must have good quality with high coordinates xc and yc, which are defined and representing
resolution for Image processing. Some Iris datasets are parametric circle equation. It is given below:
available in database to test Iris recognition. The available ( X − X C ) 2 − (Y − YC ) 2 = r 2
Iris datasets are UBIRIS, CASIA, etc.

Fig.4. Segmented Iris

The equation of parametric circle representation:
x = xc + r cos θ
y = y c + r sin θ
The Line equation is:
r = x cos θ + y sin θ
Where,
r is the distance between origin and line
Fig.3. Output Image Acquisition θ is the angle of the vector

2. Iris Segmentation:
In Iris recognition system, the Iris segmentation is second
stage. In this stage, the Iris is segmented from an eye Image
by using different segmentation techniques. The
segmentation techniques are Hough transform, Integro-
differential operator and gradient based edge detection,
removal of eyelash & eyelids and bisection methods. The
Iris part is detected by eliminating the upper & lower
boundaries of eyelids & eyelashes and central part of pupil
boundary. In this we used Hough Transform Technique.

Algorithm 1: Detecting the boundary of Pupil

Step 1: Begin
Step 2: Give Input Eye Image Fig.5. Output of Segmented Iris
Step 3: If p = r =q = 0
Step 4: Calculate Inner Iris boundary Algorithm 2: Detecting Iris boundaries
Step 5: Calculate outer Pupil boundary
Step 6: If r >=Eye Image [p] [q] then Step 1: Begin
Eye _ Image[p] [q] ← 0 = Step 2: Take Input _ Image (Input Eye []
True [], Iris _ r, pupil _ r, width, Height, pC,
Eye _ Image[p] [q] ← 255 = False qC,)
Step 7: End If Loop Step 3: Consider r1 = 0, pupil _ r= pupil
Step 8: If q ← q+1 _ r, Iris _ r= Iris_r1.
Step 9: Run from step 6 to Step 8 until q
Step 4: Initially take p= 0
= width-1 Executed
Step 10: End If Loop Step 5: do p← 0
Step 11: if p ← p+1 Step 6: Initially take q= 0
Step 12: Run from Step 5 to Step 10 until Step 7: do q← 0
p = height-1Executed Step 8: Calculate radius using (p-pc)2+(q-
Step 13: End If Loop qc) = r12
2
Step 14: End If Loop
Step 9: If r1<pupil_r1 or r1>Iris_r1
Step 15: Result Threshold Image, Eye Image
Step 16: End. True: Eye[p] [q] = 0
Step 10: end if Loop
Hough Transform technique: Step 11: Display Input _ Eye
Hough transform technique is defined the intensity values Step 12: End
of an eye Image by calculating the first derivatives of edge
map and detected the outer Iris boundary and pupil boundary

1754
3. Iris Normalization:
In Iris recognition system, Iris normalization is third stage.
In this stage, the segmented Iris is converted from circular
region into rectangular region. It has different techniques to
normalize the segmented Iris. The normalization techniques
are non-linear model, polar transform and rubber sheet
model. In this we used Rubber Sheet Technique.

Rubber Sheet Technique:

Once Segmented the Iris, the next stage is Iris
Normalization. In this the Segmented Iris is converted from
circular region to rectangular region by using rubber sheet
technique.
Iris Normalization is a process of converting the Iris into
its polar equivalent. The centre of the pupil is identified and Fig.8. Iris Normalization and Feature Encoding
removed from an eye Image, and radial vectors are passing
through the Iris region. The number of points are chosen in 4. Feature Encoding
each radial line with the radial resolution. The Iris region is In Iris recognition system, the feature encoding is fourth
defined as the number of radial lines and angular resolution stage. In this stage, the normalized Iris is encoded in the
going around Iris region. form of binary bits (0 & 1). It has different techniques for
encoding the normalized Iris. The techniques are Gabor filter,
r'= αβ ± α β 2 − α − rI 2 Cumsum based CPA and Ridge let. In this we used Gabor
Filter Technique.
With
α = o x2 + o 2y Gabor Filter Technique:
The most discriminating features of Iris is created an Iris
  oy   template. Extracted the feature extraction is done by Gabor
β = cos π − arctan   −θ 
  ox   filter technique to encode the normalized Iris region into
binary bit format. A Gabor filter technique is developed by
Where r is the radius of Iris
Gaussian and it is a sine wave modulated. Extracted the
Ox, Oy is Displacements center of pupil to Iris
unique features of Iris once applied on entire Image. Feature
θ is the angle
encoding is processes of converting the normalized Iris into
binary bit format by using 1D Gabor wavelets transform
technique. In this, Iris normalisation is done by using polar
coordinates, the polar form filter equation is given below
(r , θ ) = − i w (θ − θ ) − ( r − r0 )2 / α − i (θ − θ )2 / β
2 2
H e 0
e e 0

Where (r,θ) specify that center frequency of the filter

(α,β) specify effective width and lengh
Fig.6. Unwrapping the Iris

The Gabor filters technique defines output in order to

compress the data into four levels. Each possible quadrant
data in the complex plane is quantised by the phase
information.
I ( ρ ,φ ) e
−iw(θ0 −φ ) −( r0 − ρ ) /α 2 −(θ0 −φ ) / β 2
2 2

h{Re,Im} = sgn{Re,Im} ∫ ∫ e e ρ d ρ dφ
ρ φ

Fig.7. Normalized Iris where, h{Re, Im} Im is the imaginary part and Re is the real
part, depending on quadrant of Iris assign the values 1 or 0,
in dimensionless polar coordinate system I( ρ,θ ) is lies in
raw Iris image.

Using imaginary and real values, extracted the phase

information by encoding the bit pattern.
 Im Ig ( r , θ )
P (r ,θ )= tan − 1  
 R e Ig ( r , θ ) 

1755
  [1,1]if 0 0 < P ( r , θ ) ≤ 90 0  Template1100011001001
 0  ⇒ HD = 0.83
 [0,1]if 90 < P ( r , θ ) ≤ 180 
0
I P [r ,θ ] =  0
Template 2 001100100110
[0, 0]iff 180 < P ( r , θ ) ≤ 270 
0
← Shift 2bitsleft
[1, 0]iff 270 0 < P ( r , θ ) ≤ 360 0 
Template1 001100100110
⇒ HD = 0.00
Template 2 001100100110
→ Shift 2bitsright
Template1 011000110010
⇒ HD = 0.33
Template2 001100100110

C. Results
Iris recognition system is developed by using Image
processing toolbox of MATLAB software. In this project we
used canny edge and Hough transform algorithms to find
Fig.9. Phase Quantisation the Iris region. Daugman’s algorithm applied to convert the
segmented Iris from circular region into rectangular region.
The number of filters in the radial resolution is lesser than The normalized Iris is encoded in the form of binary bits (0
angular resolution. Total number of bits in the Iris template & 1) by using Gabor filter technique. In feature matching
will be n times more than Iris template. Parameters and defines the integrity match score based on the two encoded
centre frequencies of modulating Gaussian functions must be Iris Images by using Hamming distance technique. In this we
detected the number of filters. compared two (CASIA and UBIRIS) dataset of eye images
by calculating the FAR and FRR. The formulas of FAR and
FRR given below
W r o n g ly A c c e p te d In d iv i d u a l s
FAR =
T o ta lN u u m b e r o fW r o n g M a tc h in g

WronglyRejectedIndividuals
FRR =
TotalNuumberofCorrectMatching

TABLE II
FAR AND FRR VALUES BY APPLYING HAMMING DISTANCE TECHNIQUE

Threshold FAR% FRR%

0.12 0.000 98.947
0.18 0.000 81.977
Fig.10. Feature encoding
0.24 0.000 36.980
5. Feature Matching:
0.30 0.000 9.548
This is last step in Iris recognition system. In this step,
0.36 0.005 0.827
feature matching is defines the integrity match score based
on the two encoded Iris images. It has some techniques to 0.42 5.859 0.000
define the match score. The feature encoding techniques are 0.48 90.549 0.000
hamming distance, Euclidean distance and K-NN classifier. 0.54 99.755 0.000
In this we used Hamming Distance technique.
Hamming Distance technique: TABLE II
Based on two Iris templates calculated matching score by FAR AND FRR VALUES BY APPLYING EUCLIDEAN DISTANCE TECHNIQUE
using the Hamming Distance which gave a measure of how Threshold FAR% FRR%
many bits disagreed.
0.18 0.000 92.407
Hamming distance is calculated by using two Iris
templates, Iris template bits are shifted one shift right and 0.24 0.000 80.256
one shift left to determine rotational inconsistencies. Two 0.30 0.000 45.350
Iris images of the same eye will be determined by rotational
0.36 0.000 10.650
inconsistencies with the difference of maximum angle
between same Iris images. 0.42 4.329 1.830
Bit shift concept is applied to get approximate matching 0.48 18.627 0.000
score by shifting one bit left and one bit right.
0.54 91.812 0.000

1756
 IV. CONCLUSION [15] S. Pardhi and S. G. Qureshi, “Designing and Implementation of Iris
recognition System Using Morphological Bridged Canny Edge
The Iris recognition system is developed by using image Detection and KNN Classifier,” International Journal of
processing toolbox of Matlab software. In this, we used Engineering and Computer Science, ISSN: 2319-7242, vol. 4(6), pp.
12604-12609, Jun. 2015.
Hough transform technique to find the outer and inner
[16] P. P. Chitte, J. G. Rana, R. R. Bhambare, V. A. More, R. A. Kadu,
regions of Iris. After that applied rubber sheet technique to and M. R. Bendre, “IRIS Recognition System Using ICA, PCA,
the segmented Iris for converting from circular region into a Daugman’s Rubber Sheet Model Together,” International Journal of
rectangular region. In feature encoding, the normalized Iris Computer Technology and Electronics Engineering, ISSN: 2249-
is encoded and forms binary bit of Iris template by using 6343, vol. 2(1), pp. 16-23, 2010.
Gabor filter technique. In feature matching, defines the [17] C. Sanchez-Avila and R. Sanchez-Reillo, “Two different approaches
integrity match score by using hamming distance technique for Iris recognition using Gabor filters and multiscale zero-crossing
representation,” Pattern Recognition, vol. 38, pp. 231–240, 2005.
to the two encoded Iris templates. In feature matching, the [18] K. Sathiyaraja, M. Dhineshkumar, and N. Thiyagarajan, “Iris
encoded Iris template is compared with database eye Image Segmentation and Recognization Using Log Gabor Filter and
of Iris template and generated the matching score by using Curvelet Transform,” International Journal of Engineering and
Hamming distance technique. Based on the matching score, Computer Science, ISSN: 2319-7242, vol. 2(9), pp. 2709-2714, Sep.
we get the result. The future research work is to implement 2013.
this project for real time personal authentication for security [19] E. P. Win and N. Aye, “An Effective Iris Recognition System,”
International Conference on Advances in Engineering and
purpose.
Technology, pp. 29-30, Mar. 2014.
REFERENCES [20] S. Dalal and T. Sahoo, “A Selective Feature Matching Approach for
Iris Recognition,” International Journal of Computer Applications,
[1] C. Houston, “Iris Segmentation and Recognition Using Circular
ISSN: 0975–8887, vol. 41, no. 20, pp. 34-39, Mar. 2012.
Hough Transform and Wavelet Features,” pp. 1-4, 2010.
[21] A. I. Desoky, H. A. Ali, and N. B. Abdel-Hamid, “Enhancing Iris
[2] M. S. Monaheng and K. Padmaja, “Iris Recognition using Circular
Recognition System Performance using Templates Fusion,” Ain
Hough Transform,” International Journal of Innovative Research in
Shams Engineering Journal, vol. 3, pp. 133–140, 2012.
Science, Engineering and Technology, ISSN: 2319-8753, vol. 2(8),
[22] K. Sharma and H. Monga, “Efficient Biometric Iris Recognition
pp. 3546-3553, Aug. 2013.
Using Hough Transform with Secret Key,” International Journal of
[3] R.B. Dubey and M. Abhimanyu, “Iris Localization using Daugman’s
Advanced Research in Computer Science and Software Engineering,
Intero-Differential Operator,” International Journal of Computer
vol. 4(7), pp. 632-640, Jul. 2014.
Applications, ISSN: 0975–8887, vol. 93, no. 3, pp. 6-12, May 2014.
[23] M. W. Amit and R. T. Satish, “Eyelids, Eyelashes Detection
[4] D. Maeva, “Iris Segmentation using Daugman's Integro-Differential
Algorithm and Hough Transform Method for Noise Removal in Iris
Operator,” 2010.
Recognition,” International Journal of Computer Applications,
[5] W. Yuan and W. He, “A Novel Eyelash Detection Method for Iris
ISSN: 0975 – 8887, vol. 112, no. 3, Feb. 2015.
Recognition,” IEEE Engineering in Medicine and Biology 27th
[24] P. Verma, D. Maheedhar, S. Basu, and P. Verma, “Hough Transform
Annual Conference Shanghai, China, pp. 6536- 6539, Sep. 2005.
Method for Iris Recognition-A Biometric Approach,” International
[6] W. Zhu, Z. Zhao, and Y. Wu, “An Algorithm of Eyelashes Detection
Journal of Engineering and Innovative Technology, ISO 9001:2008
for Iris Recognition,” International Journal of Security and Its
Certified, ISSN: 2277-3754, vol. 1(6), PP. 43-48, Jun. 2012.
Applications, vol. 10, no. 7, pp. 195-202, 2016.
[25] N. Singh, G. Dilip, and P. S. Krishna, “Iris Recognition System
[7] M.Adam, F.Rossant, F.Amiel, B.Mikovikova, and T. Ea, “Eyelid
using A Canny Edge Detection and A Circular Hough Transform,”
Localization for Iris Recognition,” Radio Engineering, vol. 17, no. 4,
International Journal of Advances in Engineering & Technology,
pp. 82-85, Dec. 2008.
ISSN: 2231-1963, vol. 1(2), pp. 221-228, May 2011.
[8] R. Abduljalil, Z. Nasharuddin, and M. Ismail, “Efficient Iris
[26] S. Nithyanandam, K. S. Gayathri, and P. L. K. Priyadarshini, “A
Segmentation Based on Eyelid Detection,” Journal of Engineering New Iris Normalization Process for Recognition System with
Science and Technology, vol. 8, no. 4, pp. 399–405, 2013. Cryptographic Techniques,” International Journal of Computer
[9] R. Satish, P. Rajesh Kumar, and I. Praneeth, “Iris Recognition Science, ISSN: 1694-0814, vol. 8(4), no. 1, pp. 342-348, Jul. 2011.
System Using Geodesic Active Contours for Non-Ideal Iris Images,” [27] C. R. Prashanth, D. R. Shashi kumar, K. B. Raja, K. R. Venu gopal,
International Journal of Engineering Research in Electronics and and L. M. Patnaik, “High Security Human Recognition System using
Communication Engineering, vol. 2(6), pp. 81-87, Jun. 2015. Iris Images,” International Journal of Recent Trends in Engineering,
[10] A. Mohammed, M. Abdullah, S. S. Dlay, and W. L. Woo, “Fast and vol. 1, no. 1, pp. 647-652, May 2009.
Accurate Pupil Isolation Based on Morphology and Active Contour,” [28] Open source Libraries for eye Images of CASIA database link:
International Journal of Information and Electronics Engineering, http://biometrics.idealtest.org/dbDetailForUser.do?id=4.
vol. 4, no. 6, pp. 418-422, Nov. 2014. [29] Open source code Libraries for Iris Recognition System
[11] D. S. Jeong, D. Cho, J. Jo, M. K. Bae, M. W. Park, and E. C. Lee, link1:http://www.peterkovesi.com/studentprojects/libor/sourcecode.h
tml.link2:http://www.advancedsourcecode.com/Iris.asp.
“Compensation for Non-linear Iris Pattern Deformation based on the
Tensile Properties of Iris,” Wseas Transactions on Information
Science and Applications, ISSN: 2224-3402, vol. 12, pp. 315-323,
2015.
[12] Z. Wei, T. Tan, and Z. Sun, “Nonlinear Iris Deformation Correction
Based on Gaussian Model,” S.-W. Lee and S.Z. Li (Eds.): ICB 2007,
pp. 780–789, 2007.
[13] S. Anicham and C. Murukesh, “An Efficient Iris Recognition System
Using Contour let Transform and Neural Networks,” International
Journal of Innovative Research in Science, Engineering and
Technology, ISSN: 2319-8753, vol. 3(4), pp. 11876-11881, Apr.
2014.
[14] V. Mayank, S. Richa, and N. Afzel, “Improving Iris Recognition
Performance Using Segmentation, Quality Enhancement, Match
Score Fusion, and Indexing,” IEEE Transactions on Systems, pp. 1-
15, 2008.

1757