International Conference on Digital Image Processing
Iris Recognition Using Discrete Wavelet Transform Ms.Lenina Vithalrao Birgale
Manesh Kokare
Department of Electronics and Communication Engineering, ICFAI University, Institute of Science & Technology, 4 - 102, Jeedimetla, NH 7,Hyderabad, Andhra Pradesh State, 500055 (India) Email:
[email protected]
Department of Electronics and Telecommunication Engineering, SGGSIT, Vishnupuri, Nanded, Maharashtra State, 431606 (India) E-mail:
[email protected]
Abstract- We propose a novel approach for improved iris recognition system with reduced false acceptance rate (FAR) and false rejection rate (FRR). The technique developed here uses all the frequency resolution planes of Discrete Wavelet Transform (DWT). These frequency planes provide abundant texture information present in an iris at different resolutions. The accuracy is improved up to 98.98%. With proposed method FAR and FRR is reduced up to 0.0071% and 1.0439% respectively.
have used three level decomposition of the iris signature as shown in figure 2(e). It gives better results. We have found comparative reduction in false acceptance and false rejection rates by this method thus improving the system efficiency. Eye lid
Keywords- Biometrics, features, database, wavelets, FAR, FRR, Euclidian distance.
Pupil
1. Introduction The demand for security systems is increasing day by day. Rigorous search for different verification and identification techniques is the need of the day. Facial features, voice patterns, hand geometry, retinal patters, vein patterns, signature dynamics, voice verification, facial thermography, DNA matching nail bed identification, gait recognition, ear shape recognition and finger prints have all been explored as biometric identifiers with varying levels of success. However iris being, unique and stable for a life period is the most reliable biometric. Iris as recognition biometric for identification formed the active research area since 1992. The uniqueness of iris patterns was identified since then. This uniqueness property of iris can be quoted in the words of Daugman, [5] as, ‘An advantage of the iris shares with fingerprints is the chaotic morphogenesis of its minutiae’. Iris can be identified as the colored portion of the eye lying between pupil and sclera. Iris in an eye can be identified as shown in figure 1(a). A front-on view of the human eye is shown in Figure 1(b). A very important characteristic of an iris is that it is a naturally protected organ and is stable without any variations along with the age of an individual. The main motivation of this work is the multi resolution provided by wavelets as verified by our previous work [8]. It helps to form an iris signature of prominent length which can be used in the recognition system. Various research groups such as Ma et al [9], Tisse et al [3], Zhu et al [13] and Wildes et al [12] has proposed a recognition system based on Haar wavelets. Rydgren et al [4] has proposed a system based on wavelet packets. They all have not exploited all the frequency planes. But our paper shows that standard wavelets provide much texture information required in iris signature formation. Here we 978-0-7695-3565-4/09 $25.00 © 2009 IEEE DOI 10.1109/ICDIP.2009.30
Iris
Figure 1(a): Database iris image. Pupil Iris
Sclera Eye lashes Figure 1(b): Database iris image. Though system efficiency is low compared to that of Daugmans method, false acceptance rate outperforms the Daugmans and other methods. Also Haar wavelets give better results than that of daubchies wavelets. Obtained results are compared with the existing methods [2, 3, 5, 6, 7, 9, 10, 11]. This paper can be organized as follows. A brief overview of standard wavelets is presented in section 2. Section 3 reveals the proposed method. Section 4 and section 5 respectively presents experimental results and conclusions.
147
2. Standard wavelet The Short Time Fourier transform (STFT) represents a sort of compromise between the time- and frequencybased views of a signal. It provides some information about both when and at what frequencies a signal event occurs. However, one can only obtain this information with limited precision, and that precision is determined by the size of the window. But many signals require a more flexible approach where we can vary the window size to determine more accurately either time or frequency. Wavelet analysis is such an approach in which window with variable-sized regions is used. Wavelet analysis allows the use of long time intervals where we want more precise low-frequency information, and shorter regions where we want high-frequency information. However it is a time-frequency region, but rather a time scale.
W\V ( j , m, n) Vertical coefficients W\D ( j, m, n) Diagonal coefficients Here WI ( j , m, n ) is the original image whose DWT is to be computed. 3. Proposed method The proposed method gives better performance. It reduces the false acceptance rate and false rejection rate and improves the system efficiency. Though Daugmn’s algorithm out performs in terms of average accuracy our method out performs in terms of false acceptance rate. Also the feature vector size used by is very small as compared to that that used by Daugman [6]. Our method is discussed below. 3.1. Image database The database used in the experimentation consists of 756 different iris images from CASIA iris image database [1]. CASIA iris database is the only public domain database available. Size of each database iris image is 280 x 320. In this database there are 108 different subjects. Each subject has contributed to seven different images of same eye. Thus there are a total of 756 (108 x 7) images in the database. Among the seven images three are taken in first session and other four are taken in second session.
2.1. The Discrete Wavelet Transform Computing wavelet coefficients at every possible scale is a fair amount of work, and it generates an awful lot of data. That is why we choose only a subset of scales and positions at which to make our calculations. It turns out, rather remarkably, that if we choose scales and positions based on powers of two so-called dyadic scales and positions then our analysis will be much more efficient and just as accurate. We obtain such an analysis from the discrete wavelet transform (DWT) given by (1). f
DWT
2I 2 x n
(2)
3.2. Feature database creation Each point within the iris region can be mapped in to a pair of polar coordinates (r, ) where r is in the interval [0, 1] and is angle [0, 2]. It is shown in figure 2. The normalized iris image is used for iris signature extraction. Figure 2(a) and 2(c) represent segmented iris and 2(b) and 2(d) represents polar coordinates (r, ) mapping. Figure 2(e) represents three level, wavelet decomposition. To calculate the signature of iris we have used first level energy (4) and computed standard deviation (5) of each sub-band. Thus the signature vector size of database is of 756 x 24 and for query image it is of size 1 x 24. First level energy is computed by (4).
¦ h\ n 2\ 2 x n
(3)
Energy
f
¦ ¦ q k , l \ 2 k t l k 1 l f
(1)
An efficient way to implement this scheme using filters was developed in 1988. This algorithm is in fact a classical scheme known in the signal processing community as a two-channel sub band coder. This very practical filtering algorithm yields a fast wavelet transform a box into which a signal passes, and out of which wavelet coefficients quickly emerge. Let’s examine this in more depth. Let,
I x
¦ hI n n
\ x
n
x
M 1 N 1
\ x
(4)
Where X m, n) is a discrete function whose energy is to be computed. Similarly standard deviation is given by 5.
Both and can be expressed as linear combinations of double-resolution copies of themselves.
h
¦¦ X m, n)
m 0n 0
h
Here in (2) and \ in (3) the expansion coefficients are called scaling and wavelet vectors, respectively. They are the filter coefficients of fast wavelet transform (FWT), WI ( j , m, n) Approximate coefficients
Vk
1 MXN
Where,
N
N
¦¦ E>W i, j P @ k
k
i 1 j 1
Wk i, j is the kth wavelet decomposed sub-band.
H
W\ ( j , m, n) Horizontal coefficients
148
(5)
r
0 1
Figure 2(a): Segmented Iris
Figure 2(b): Normalized iris
Figure 2(c): Segmented Iris
Figure 2(d): Normalized iris
Original image
Level 1
Horizontal
Vertical
Diagonal
Approximate
Horizontal
Vertical
Diagonal
Approximate
Horizontal
Vertical
Diagonal
Approximate
Level 2
Level 3
Figure 2(e): Three level wavelet decomposition of normalized iris
149
M x N is the wavelet decomposed sub-band. P k is the mean value of kth decomposed sub-band.
better and reliable results. Though Daugman’s method gives best efficiency the iris signature length used by that algorithm is comparatively very high that is 1x 2048 phase vector. Also Daugman has used phase information in signature formation. Proposed method iris signature is just 1x24 vectors. Signature of this length can definitely improve the speed. However the proposed method can be further improved by avoiding iris segmentation by polar coordinate conversion and designing optimized mask for iris segmentation.
3.3. Matching A query signature is one of 756 signatures from image database signatures. Euclidean distance metric given by (6) is used to compute the similarity or match value for given pair of signatures. Zero distance implies a perfect match, and signature tends towards mismatch as the distance increases. N
D Eucli x, y
¦ x i 0
i
yi
TABLE 1. Comparison of results obtained with existing methods
2
(6)
3.4. Performance measures The threshold value used is the average distance of top seven signatures as they belong to different eye images of same person’s eye. Three images are taken in first session and four are taken in second session. The threshold thus set is used to compute False Acceptance Rate (FAR) and False Rejection Rate (FRR). FAR and FRR are the performance measures used. FAR is computed by (7) and FRR is computed by (8).
FAR
NFA NIVA
Algorithms
Avila [11] Ma [9] Tissue [3]
FRR
--------
% FAR
%FRR
% Average Efficiency
0.0300
2.0800
97.89
0.0200
1.9800
98.00
1.8400
8.7900
89.37
(7)
Where, NFA is the Number of False Acceptances. NIVA is the Number of Imposter Verification Attempts. FAR is the measure of likelihood that the biometric security system would incorrectly accept that an attempt by an unauthorized user.
NFR NEVA
Feature vector length
Daugman [6]
1x2048
0.0100
0.0900
99.90
Chu [2]
---
0.0000
0.6900
99.14
1.2000
31.500
-
0.0100
0.8800
-
0.0071
1.0427
98.95
0.0069
1.0207
Cui [7]
(8)
Maheshwari [10]
Where, NFR is the Number of False Rejections. NIVA is the Number of Enrollee Verification Attempts. FRR is the measure of likelihood that the biometric security system would incorrectly reject that an attempt by an authorized user.
-----
The proposed method (using daubchies wavelet)
1x24
The proposed method (using haar wavelet)
1x24
4. Experimental results Results obtained by the proposed method are much interesting and reliable. They are listed below in table. Table 1 gives comparative study of different methods along with the results obtained by the proposed method. From this we can see that FAR obtained is 0.0071 for Daubchies wavelets and 0.0069 for Haar wavelets. It is evident from the table that both the results are comparatively out performing over the other methods.
[1]
5. Conclusions From the listed table it can be concluded that Discrete wavelet transform used for iris signature formation gives
Chinese Academy of Sciences – Institute of Automation. Database of 756 Greyscale EyeImages. http://www.sinobiometrics.com Version 1.0, 2003.
[2]
Chia Te Chu and Ching-Han Chen, “High Performance Iris Recognition Based on LDA and LPCC”, Trans IEEE, Proceedings
98.98
6. References
150
of Conference on Tools with Artificial Intelligence (ICTAI ’05) 2005. [3]
Christel – Loic Tisse, Lionel Torres and Michel Robert, “Person Identification based on iris patterns”, Proceedings of the 15th International Conference on Vision interface, 2002.
[4]
Erik Rydgren, Thomas EA, Federic Amiel, Florence Rossant and Amara Amara, “Iris Features Extraction using wavelet Packets”, IEEE Trans. an International Conference on Image Processing(ICIP), pp 861-864, 2004.
[5]
J.G.Daugman, High Confidence Visual Recognition of Persons by a Test of Analysis of Statistical Independence, IEEE Trans on Pattern Analysis and Machine Intelligence, Vol. 15, No. 11, pp 1148-1161, Nov 1993.
[6]
J.Daugman (2001c), “Statistical Richness of Visual Information: Update on Recognising persons by Iris Patterns”, International Journal on Computer Vision, vol. 45, No. 1, pp 25-38.
[7]
Jiali Cui, Yunhong Wamg, Ju Zhou Huang, Tieniu Tan and Zhenan, “An Iris Image Synthesis Method Based on PCA and Super-resolution”, Proceeding of the 17th International Conference on Pattern Recognition (ICPR’04) 2004.
[8]
Lenina Birgale, Manesh Kokare and Dharmpal Doye, “Color and Texture Features for Content Based Image Retrieval”, Proceeding of 3rd IEEE Int Conference CGIV06, pp 146-149, Aug 2006, Sydney, Australia.
[9]
Li Ma, Y. Wrag and T.Tam, “Iris Recognition Using Circular Symmetric Filters”, Pattern Recognition, 16th International Conference on, vol 2, pp 414-417, 2002.
[10] Maheswari, P.Anbalagan and T.Priya, “Efficient Iris Recognition through Improvement in Iris Segmentation Algorithm”, ICGSTGVIP Journal, ISSN: 1687-398X , vol 8, Issue 2, pp. 29 – 35, 2008. [11] R. Sanchez- Reillo, C. Sanchez- Avila and De Martin- Roche D, “Iris Recognition for Biometric Identification usind Dyadic Wavelet Transform Zero Crossing”, Proceeding of the IEEE 35th Carnahan International Conference on Security Technology, pp 272 – 277,2002. [12] R. P. Wildes, J.C.Asmultr, G.L.Green., S.C. Hsu, R.J.Kolczynski, J.R. Matey and S.E.McBride, “A Machine Vision System for Iris Recognition”, Machine Vision and Application, Vol.9, pp 1-8, 1996. [13] Y. Zhu, T. Tan and Yusag ,“Biometric Personal Identification based on Iris Patterns, pattern Recognition”, 15th Internal Configuration, vol. 2, pp 801-804, 2004.
Acknowledgements- Authors would like to thank to Iris Recognition Research Group, Center for Biometrics and Security Research, National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences for providing the iris database for research.
151
