Page 1 of 11
DOI: 10.14738/tmlai.86.9687
Publication Date: 14th February, 2021
URL: http://dx.doi.org/10.14738/tmlai.86.9687
Volume 8 No 6
A Review of the Iris Recognition Methods Used for the
Individual Authentication
1
Amina A. Abdo, 2
Ahmed O. Lawgali, and 3
Mohamed A. E. Abdalla 1,2,3 Faculty of Information Technology, University of Benghazi, Benghazi, Libya;
amina.abdo@uob.edu.ly, ahmed.lawgali@uob.edu.ly, and mohamed.abdalla@uob.edu.ly
ABSTRACT
The automatic iris recognition has become one of the most important techniques for authenticating the
identity of individuals. The analysis of human iris is a reliable tool for the individual authentication due to
the iris structure. Iris patterns constitute one of the uniqueness, permanence, and performance biometric
traits. Moreover, the iris is considered as not easily tampered biometric traits. Therefore, this paper
considers investigating the common automated methods of iris recognition. It surveys the development
of utilizing iris images as an authentication means through the explanation of the historical improvement
of the processes of the iris analysis. The contribution of this paper is to provide readers with huge
information collected and discussed from more than 40 papers of iris recognition studies which have been
published in a period of more than 20 years.
Keywords: Biometric traits; iris normalization; feature extraction; iris recognition and classification.
1 Introduction
Automatic biometric systems indicate to the measurable physiological or behavioral characteristics used
to verify the identity of individuals. Several biometric traits are used to authenticate the identity of
individual such as face, iris, fingerprint, handwriting, signature, voice, gait, and DNA [1]. Biometrics should
meet the specified recognition accuracy, speed, and resource requirements. In other words, biometric
systems are important for many applications that are related to security such as passenger control in
airports, access permission in restricted areas, border control, database access, and financial services [2].
The biometric features are unfeasible to be lost, forgot, or stolen easily in contrast with other methods
such as passwords or bin codes. Furthermore, biometrics of every human being are unique, accurate and
stable information [3], [4]. The analysis of iris provides rich and distinctive information which can be used
to identify individuals. As shown in figure 1, the iris of human eyes is a thin, colored ring around the pupil.
It is bordered between an inner and outer circles. The inner circle locates between the iris and pupil
boundary, the outer circle locates between the sclera and iris boundary [4]. There are five parts inside the
iris: crpyts, radial area and furrows, radial, ciliary area, and collarette. These parts together ensure
achieving the biological function of iris which controls the amount of light entering through the pupil [5].
The exploitation of the automatic recognition of iris has been rapidly increased and become one of the
most reliable and accurate techniques for biometric authentication. The fundamental and theoretical idea
of the recognition the analysis of the human iris was proposed in 1939 [6]. Whilst the implementation of
Page 2 of 11
Transactions on Machine Learning and Artificial Intelligence Volume 8 , Issue 6, Dec 2020
Copyright © Society for Science and Education United Kingdom 17
that idea was carried out in 1990 [7]. According to the previous studies collected in this paper, it would
be noticed that, for over two decades, the recognition of iris images has broadly been a reliable biometric
authentication technique.
Considerable studies state that an automatic recognition of iris is achieved by following five steps [8]. The
first and foremost is the acquisition of iris images. It is done by standard iris camera operated invisible
infrared light band. Due to the large number of available iris datasets, researchers often rely on published
images such as CASIA iris, UBIRIS, MMU, and UPOL [8], [9].
The image pre-processing is the second step. It is simply adopted for the enhancement of the region of
interests. It includes enhancement, segmentation, and normalization [10]. Followed by the feature
extraction as a third stage to collect the effective descriptors that present the iris pattern. Finally, the
classification models are implemented to measure similarities and dissimilarities between two or more
iris images. Figure 2 depicts the steps of iris recognition algorithm. The discussion presented in this paper
is to explain the sequential phases in several studies of the iris identifications: image pre-processing,
feature segmentation, normalization, feature extraction, and finally the classification. The discussion also
involves the common datasets examined in those studies.
2 The Development of Iris Recognition
As mentioned earlier and shown in figure 2, the framework of the process of iris recognition requires
image requisition, image pre-processing, Feature Extraction, and the classification.
Iris
Crypts
Radial furrows
Radial area
Ciliary area
Collarette
Pupil
Figure 1. Iris structure
Figure 2. Diagram of the recognition of iris system
Image result for structure of iris
Page 3 of 11
Amina A. Abdo, Ahmed O. Lawgali, and Mohamed A. E. Abdalla; A Review of the Iris Recognition Methods Used
for the Individual Authentication. Transactions on Machine Learning and Artificial Intelligence, Volume 8 No 6
December (2020); pp: 16-27
URL:http://dx.doi.org/10.14738/tmlai.86.9687 18
2.1 Image Acquisition
Capturing valid iris images for the recognition process requires acquiring photos with good resolution and
clarity. Furthermore, any noise must be removed, because noises such as illumination causes poor
decisions [6]. To avoid noises, infrared cameras are recommended to be used for this issues [2]. There are
several available datasets of digital images of human iris. These datasets have been analyzed and
examined in huge number of studies. Table I shows the common datasets and the number of images and
the classes on every database.
2.2 Pre-processing Phase
The image pre-processing phase aims improve the image features by removing unwanted distortions and
enhancement of informative features of images. This task is to ensure exactly detecting the region of iris.
Majority of iris recognition studies include four stages: enhancement, localization/segmentation, and
normalization [1].
The pre-processing is the initial and important stage for yielding accurate recognition rates. Applying
similar techniques was quite familiar among various studies. Generally, those techniques started with
converting the images from color to grayscale images [8]. Followed by converting to the black and white
images using threshold values [4]. Those two stages were dominant in a huge number of the available
studies. The following phases of the segmentation are not unified, because the image conditions in the
different datasets are not always similar due to the image resolution and clarity [3].
Some studies considered the detection of the inner and outer borders of the iris[16]. Other studies tried
to locate the eye pupil first and then went out to detect the iris [12]. In contrast, a large number of
previous studies utilized morphological processes to detect the region of the whole iris correctly [7], [17].
In some case, the detection is for partial parts of the iris circle especially when the eyelid covers the top
or the bottom of the iris. For example, one of the available studies concentrated on segmenting left and
right quadrants of the iris images [7].
Moreover, the normalization is considered as an image processing stage of iris recognition. It is as next
phase after the segmentation. The aim of the normalization is to enhance the segmented region from the
whole iris image [18]. Homogeneous rubber-sheet, which found by Daugman model, is one of most
common methods used to normalized iris images [10]. It is used to remap each point within the iris region
to a pair of polar co-ordinates [19].
Table 1. The common datasets of iris images used
in iris recognition studies
Reference Name Description Image
number
Class
number
[6] CASIA V1.0 Chinese Academy of Sciences Institute of Automation
(CASIA) version 1.0 756 108
[11] CASIA-Iris V3
Interval
Chinese Academy of Sciences Institute of Automation
(CASIA) version 3 Interval
http://biometrics.idealtest.org/ dbDetailForUser.do?id=3
249 99
[12]
CASIA- IrisV4-
Interval
Chinese Academy of Sciences Institute of Automation
(CASIA) version 4 Interval 2639 249
[13] CASIA-V3
Lamp
Chinese Academy of Sciences Institute of Automation
(CASIA) version 3 Lamp 1000 50