iris image recognition method based on...
TRANSCRIPT
Iris Image Recognition Method based on Multi-feature
Lv Hanfei
Department of Information Technology and Management
ZheJiang Police Vocational Academy
Hangzhou, Zhejiang Province, China
Abstract. We raise one multi-feature combination localization method
(MCLM) for iris image recognition in this paper. MCLM is a novel approach
out of the ordinary iris localization algorithm. Our comprehensive utilization of
the advantages of both linear filtering and non-linear filtering methods can not
only remove the interference but also make up the defect of iris image edge
information. We adopt hamming distance approach as the template matching
method for iris matching.
Keywords: Multi-feature Combination, Iris Localization, Iris Image
Recognition, Template Matching
1 Introduction
Recently iris recognition is more and more popular, and Samsung even wonder to use
this technology in its flagship mobile phone. The iris is unique, even the twins have
different iris. And it is stable from the baby to the aged for one person. High security
and high accuracy are the two typical advantages of iris recognition technology. And
it has broad prospects in the security applications.[1,2] The bank, the hospital, the
police station, and the mine have deployed the iris image recognition system.
In one iris recognition system the first and the kernel step is the iris localization.
We must detect and isolate the iris components from an image of the eye
segmentation. In fact the iris located in the vicinity of the sclera, the pupil and the
eyelids. It is hard to precisely detect the boundaries separating the iris from these
other components in the segmentation process.
Many scholars have proposed different methods to enhance the performance of iris
localization and iris image recognition. Daugman [3] presented 2D Gabor filter based
method for the iris identification system and he used hamming distance method to do
the iris matching. Wildes et al. [4] adopted Laplacian pyramid for highly-efficient
execution of gradient-based iris segmentation. Based on zero-crossing representation
from the wavelet transform decomposition, Boles [5] has raised fine-to-coarse
approximation at diverse resolution levels. He also adopted European distance method
in his system for iris template matching.
Advanced Science and Technology Letters Vol.81 (CST 2015), pp.47-50
http://dx.doi.org/10.14257/astl.2015.81.10
ISSN: 2287-1233 ASTL Copyright © 2015 SERSC
2 Iris Localization and Template Matching
Iris localization is to determine boundary of the iris and pupil, the boundary of the iris
and sclera in eye images. The main purpose is to determine the iris inner boundary
and outer boundary. The first step is inner edge localization of iris image, and the
second step is outer edge localization of iris image.
2.1 Iris Inner Edge Localization
The inner boundary of the iris includes the limbus part at the junction of iris and pupil.
The gray of iris inner boundary vary apparently. When the person does not blink the
eyelash and eyelid occlusion trouble exists. The contrast of iris image is higher than
other parts of the eye in the infrared illumination. The localization for the inner
boundary of the iris is easy. Firstly we adopt the traditional edge detection method for
iris inner boundary detection of image after pre-processing. Secondly we use the
connected domain characteristics of multistage denoising approach to remove the
eyelash, eyelid and other interfering parts of the eye. Thirdly we adopt the circle
fitting approach based on the precise location of the inner iris boundary to locate the
circle center. Finally we do the localization work of the iris inner boundary.
We use the Canny operator for iris boundary extraction. The Canny operator is
used to extract the iris inner edge and it can preserve the iris inner edge well. And the
detail characteristics are suitable for the follow-up precise localization.
2.2 Iris Outer Edge Localization
The outer boundary of iris includes the sclera limbus edge and a part of the iris. The
edge is unclear and the part of the transition band is wide. At the same time it is hard
to get the accurate outer iris boundary with the interference of eyelids and eyelashes.
It is difficult to extract the iris outer boundary features because they are fuzzy. But it
is easy to distinguish the iris and sclera for the reason of the obvious contrast. And we
use one segmentation method which combines OTSU and the contrast to discern
between the iris and sclera in the first step. We adopt the LOG operator to extract
the weak iris outer boundary data in the next step. Then the features set filtering
approach is used to cut off the interference outside the region of iris outer boundary.
The last step is to achieve the localization of the iris outer boundary.
3.3 Iris Outer Edge Localization
In the iris image after features set filtering, the main interference region includes
sclera, canthus, eyelid and other regions in the iris image. In iris image the features of
outer boundary are different from the inner boundary. It is not a closed region for
outer boundary. The sclera is just a small connected region. The sclera interference
can be removed after we calculate the closed region number. We find that most of the
target domains for the eyelid and canthus connected with the outer boundary are
Advanced Science and Technology Letters Vol.81 (CST 2015)
48 Copyright © 2015 SERSC
independent neighborhood. We can use the similar approach to fit the outer boundary
from the target point of the independent neighborhood. Only the fitting target of
solitary neighborhood without neighborhood is selected for subsequent processing.
3.4 Iris Template Matching Method
Matching is the process of calculating the similar degree between the input iris image
and collected iris image from the iris image database. Here we use the traditional iris
matching method--Hamming Distance method.
M
i
ii BAM
HD
1
1
(1)
Ai and Bi are two different iris encoding. is EOR calculation. M is the total
bits. The iris normalization only corrects the handle the translation and zooming. It
can not correct the iris rotation for the reason of the head movement. Iris matching
method usually uses horizontal displacement to eliminate the rotation. In general we
got the minimum HDmin when the best iris match happened.
HDmin = min ( HDi) , i = -n,-n + 1,…, n,
Here HDi represents the HD value when we horizontal displace i bits. + and -
means the direction. From right to left is -, from left to right is +. And n is the largest
offset of the pixel bits.
4 Conclusion
We collected the iris features by edge extraction algorithm and the connected domain
characteristics of multistage de-noising algorithm. Then we raise one multi-feature
combination localization method (MCLM ) to precisely locate the boundary of the iris
image. MCLM can effectively remove the noise and keep the useful edge information
of the iris image. We use the hamming distance method for template matching of the
iris recognition in the second step. Therefore MCLM is an excellent localization
algorithm for iris image recognition. In the future we will try other template matching
methods used with MCLM together to get better performance for iris image
recognition.
References
1. Gale1, A., Salankar, S. S.: A Review on Advance Methods of Feature Extraction In Iris
Recognition System. IOSR Journal of Electrical and Electronics Engineering (IOSR-JEEE)
e-ISSN: 2278-1676, p-ISSN: 2320-3331 pp. 65-70 (2014)
2. Alotaibi, A., Hebaishy, M. A.: Increasing the Efficiency of Iris Recognition Systems by
Using Multi-Channel Frequencies of Gabor Filter. Journal of Remote Sensing Technology
Feb. vol. 6, no 1, pp. 98-107, (2014)
Advanced Science and Technology Letters Vol.81 (CST 2015)
Copyright © 2015 SERSC 49
3. Daugman, J. G.: High Confidence Visual Recognition of Persons by a Test of Statistical
Independence. IEEE Trans on Pattern Analysis and Machine Intelligence, Vol. 15, No 11 ,
pp.1148-1161 (1993)
4. Wildes, R. P.: Iris Recognition: An Emerging Biometric Technology. Proc of the IEEE,
Vol. 85, No 9 , pp.1348-1363 (997)
5. Boles, W. W., Boashash, B.: A Human Identification Technique Using Images of the Iris
and Wavelet Transform. IEEE Trans on Signal Processing, Vol. 46, no 4, pp.1185-1188,
(1998)
Advanced Science and Technology Letters Vol.81 (CST 2015)
50 Copyright © 2015 SERSC