International Journal of Computer Systems (ISSN: 2394-1065), Volume 03 Issue 01, January, 2016

Available at http://www.ijcsonline.com/

New Improved Feature Extraction Approach of IRIS Recognition

Jyoti pooniaA, Parvati BhuraniA, Sandeep Kumar GuptaB, Shubh Lakshmi AgrwalC
A

Government Mahila Engineering College, Ajmer

B
JECRC University, Jaipur, India
C
The ICFAI University, Jaipur, India

Abstract
Iris recognition is used for identify the person by recognizing iris pattern of eye of a person. Currently there is no method
of IRIS recognition system that have 100% recognition rate using Gabor filter. So research issues are to improve
recognition rate by improving the pre-processing of datasets, improving the feature extraction method and using the best
classifier for iris recognition. Feature extraction is the key step on which recognition rate depends for iris recognition.
Gabor filter extract the edge information of iris pattern using the projection on iris image but have a problem of huge
dimension and high redundancy. In the proposed technique, the dimension and redundancy is reduced effectively in
order to increases accuracy.
Keywords: IRIS Recognition, Gabor Filter, DWT, DCT.

I.

II.

INTRODUCTION

The iris recognition is better than other identification

technique due to not change over the years and age in iris
pattern. The iris recognition includes iris image
preprocessing, feature extraction and classification.
Wavelet transform and hybrid wavelet transforms
concluded that hybrid transforms are better than simple
transform wavelet transform for better accuracy. Local iris
features is used for iris recognition but Man Zhang [1] et.
al. Proposed both geometric and photometric features
extraction technique in order to get a better result. They
proposed that iris image should be decomposed into
lowpass and band pass components based on non sub
sampled contoured transform ant then the geometric
features were extracted from bandpass component and
ordinal measure of local iris region from lowpass
components. R. Rizal Isnanto [2] extracted the image
feature based on energy after the Wavelet transforms using
Haar and Daubechies and concluded that higher
recognition rate is achieved using Haar compared to
Daubechies wavelet transform. R. Rizal Isnanto [3]
suggested that textural characteristics of the iris pattern can
be extracted for unique identification. They used wavelet
transform feature extraction technique for extracting the
information from iris image. K. Nguyen et. al. [4]
presented feature-domain super resolution framework for
Gabor-based face and iris recognition. They used super
resolution to improve the resolution and the recognition
performance and suggested that Current existing featuredomain super resolution approaches are limited to simple
linear features such as Principal Component
Analysis(PCA) and Linear Discriminant Analysis(LDA),
which are not the most Discriminant features for
biometrics. Priya et. al [5], presented a new improved
feature extraction technique based on average Gabor scale.
They suggested two stage reduction technique of feature
extraction

RELATED WORK

A. Gabor Filter Feature Extraction

In spatial domain, Gabor filter is given by equation (1)
[6].

(1)
Where (x, y), the pixel location in the digital image and
Sx, Sy Standard deviation in the x & y directions
respectively. is inverse of central frequency and is angle
applied in the Gabor filter equation. The parameters x1 and
y1 are given as equation (2).
x1=xcos + ysin

y1= -xcos + ysin

(2)

The Gabor features are evaluated using convolution

operation of Gabor filter bank (x,y) with input image I
(x, y) which is defined in equation (3).
Gu,v(x,y) = I (x,y) * (x,y)

(3)

The Gabor filter bank Gu,v(x,y) is complex number so a

convolution operation of Gabor filter is performed
separately for real and imaginary part as defined in
equation 4 and 5.
Re(O(x,y))m ,n = I ( x, y ) * Re((x,y,m,n))
(4)
Im(O(x,y))m ,n = I ( x, y ) * Im((x,y,m,n))
(5)
The final amplitude of Gabor filter bank is calculated as
equation 6[7].
|O(x,y)|m,n=((Re(O(x,y))m ,n)2+ (Im(O(x,y))m ,n)2)1/2
(6)

1 | International Journal of Computer Systems, ISSN-(2394-1065), Vol. 03, Issue 01, January, 2016

Jyoti poonia et al

New Improved Feature Extraction Approach of IRIS Recognition

B. Discrete Wavelet Transformation

The term wavelet transform is explained as
decomposition of the data or the image into wavelet
coefficients, comparing the detail coefficients with a given
threshold value, and shrinking these coefficients close to
zero to take away the effect of noise in the data. The image
is reconstructed from the modified coefficient which is
known as the inverse discrete wavelet transforms [8]. DWT
transformation converts the iris image into four different
frequency sub band as LL, LH, HL and HH as figure 1.
Where range of frequency is represented as LL<LH<HL<
HH. The feature or characteristics of iris is represented by
low frequency coefficients or LL sub band so LL frequency
sub band is extracted for further feature reduction [9].

DWT transformation converts each Gabor feature

matrix into four different frequency sub band as LL, LH,
HL and HH as figure 1. Where range of frequency is
represented as LL<LH<HL< HH. The feature or
characteristics of iris is represented by low frequency
coefficients or LL sub band so LL frequency sub band is
extracted for further feature reduction.
After the DWT stage, DCT is applied on each LL sub
band which converts it into DCT feature matrix. In the
feature extraction, DCT extract the energy information of
the image from low frequency region and then it will focus
on some specific features located in the upper left Corner.
After above stage the resultant matrix is converted into
a single feature vector (FV) and feature vector is passed to
classifier.

IV.
LL Sub band

LH Sub band

HL Sub band

HH Sub band

Figure 1: frequency Sub band of DWT

EXPERIMENT AND RESULTS

The 7 different angles are applied for Gabor projection

with 3 scales of iris image. The CASIA dataset is used for
experiment with ratio 80/20 of training/testing. In the
proposed system, 3 different scales and 7 different
orientations or total 3*7= 21, Gabor matrices are generated.
The extraction features are reduced using proposed
reduction technique for iris recognition. The adaboost
classification technique is used with 70/30 training testing
ratio which provides the 92.5% correct recognition rate.
Table 1: Results and comparison with other techniques

C. Discrete Cosine Transformation

DCT is a closely related to discrete Fourier transform
(DFT), using the DCT a signal is categorized into its basic
frequency components. The workings of DCT coefficients
return the average energy of pixel blocks whereas the AC
components return the intensity of image[10].
Discrete cosine transform is defined [11] as equation
(7).

Avg.
S no.

DWT
1.

3.

5.

PROPOSED WORK

The iris image is preprocessed by converting from RGB

to gray scale and Gabor feature extraction is applied to
extract boundary information of iris pattern. In the Gabor
filter, Gabor equation is projected on iris image with 3
different scales and 7 different angles which generate
3*7=21 Gabor features matrices which are reduced using
following process.

feature

technique

extraction

based

iris

82%

recognition

4.

III.

Recognition
Rate (%)

2
(7)

Methods

DCT

feature

extraction

based iris recognition

Gabor filter + PCA based
iris recognition
Gabor Scale + PCA based
iris recognition
Proposed Feature Extraction
Technique

V.

75%

77%

82.5%

92.5%

CONCLUSION

The classification results of proposed work are

compared with results of iris recognition method based on
existing Gabor feature extraction technique and table 1.2
shows that proposed technique increased the accuracy of
iris system to 92.5% which is much better than existing
accuracy of Gabor Scale PCA and other feature extraction
technique.

2 | International Journal of Computer Systems, ISSN-(2394-1065), Vol. 03, Issue 01, January, 2016

Jyoti poonia et al

New Improved Feature Extraction Approach of IRIS Recognition

REFERENCES
[1]

Zhang, Man, Zhenan Sun, and Tieniu Tan. "Deformed iris

recognition using band pass geometric features and low pass ordinal
features", in International Conference on Biometrics (ICB), IEEE,
2013.
[2] Isnanto, R. R., Iris recognition analysis using biorthogonal
wavelets transform for feature extraction, In 1st International
Conference on Information Technology, Computer and Electrical
Engineering (ICITACEE), pp. 183-187, IEEE, 2014.
[3] Isnanto, R. R., Satoto, K. I., & Windasari, I. P. Constructing irislet,
A new wavelet type which matched for iris image characteristics,
In International Conference on Information and Communication
Technology (ICoICT), 2nd International Conference on pp. 232237). IEEE, 2014.
[4] Nguyen, K., Sridharan, S., Denman, S., & Fookes, C. , Featuredomain super-resolution framework for Gabor-based face and iris
recognition, In IEEE Conference on Computer Vision and Pattern
Recognition (CVPR), pp. 2642-2649, IEEE, 2012.
[5] Dosodia, P., Poonia, A., Gupta, S. K., & Agrwal, S. L.,New
Gabor-DCT Feature Extraction Technique for Facial Expression
Recognition, In Fifth International Conference on Communication
Systems and Network Technologies (CSNT), pp. 546-549, IEEE,
2015.
[6] George, A. M., & Durai, C. A. D., A survey on prominent iris
recognition systems, International Conference on Information
Communication and Embedded Systems (ICICES), pp. 191-195,
IEEE, 2013.
[7] Sandeep, Shubhlakshmi, Yogesh, Neeta, A Hybrid method off
Feature Extraction for Facial Expression Recognition , in n
seventh international conference on Signal image Technology and
Internet based systems (SITIS), page(s): 422 425, IEEE, 2011.
[8] L. Ma, T. Tan, Y. Wang, and D. Zhang, Efficient iris recognition
by characterizing key local variations, IEEE Trans on Image
Process., vol. 13, no. 6, pp. 739750,IEEE, 2004.
[9] Sharma, V. P., Mishra, S. K., & Dubey, D., Improved Iris
Recognition System Using Wavelet Transform and Ant Colony
Optimization, In 5th International Conference on Computational
Intelligence and Communication Networks (CICN), (pp. 243-246).
IEEE, 2013.
[10] Soni, K., Gupta, S. K., Kumar, U., & Agrwal, S. L., A new Gabor
wavelet transform feature extraction technique for ear biometric
recognition, In 6th IEEE International Conference on Power India
(PIICON), (pp. 1-3). IEEE, 2014.
[11] Monro, D. M., Rakshit, S., & Zhang, D., DCT-based iris
recognition, in IEEE Transactions on Pattern Analysis and
Machine Intelligence, , vol. 29(4), pp. 586-595. 2007.
[12] Jyoti poonia, Parvati Bhurani, Rohit Kumar, Shubh Lakshmi
Agrawal, "Performance Review of IRIS Recognition Systems",
International Journal of Computer Systems (IJCS), 2(12), pp: 564566, December 2015.

3 | International Journal of Computer Systems, ISSN-(2394-1065), Vol. 03, Issue 01, January, 2016