03-08-2013, 02:21 PM
Improved Feature Processing for Iris Biometric Authentication System
Iris Biometric.pdf (Size: 417.44 KB / Downloads: 37)
Abstract
Iris-based biometric authentication is gaining impor-
tance in recent times. Iris biometric processing however, is a complex
process and computationally very expensive. In the overall processing
of iris biometric in an iris-based biometric authentication system,
feature processing is an important task. In feature processing, we ex-
tract iris features, which are ultimately used in matching. Since there
is a large number of iris features and computational time increases
as the number of features increases, it is therefore a challenge to
develop an iris processing system with as few as possible number of
features and at the same time without compromising the correctness.
In this paper, we address this issue and present an approach to feature
extraction and feature matching process. We apply Daubechies D4
wavelet with 4 levels to extract features from iris images. These
features are encoded with 2 bits by quantizing into 4 quantization
levels. With our proposed approach it is possible to represent an
iris template with only 304 bits, whereas existing approaches require
as many as 1024 bits. In addition, we assign different weights to
different iris region to compare two iris templates which significantly
increases the accuracy. Further, we match the iris template based on
a weighted similarity measure. Experimental results on several iris
databases substantiate the efficacy of our approach.
INTRODUCTION
Ecent advances in information technology and increasing
emphasis on security have resulted in more attention to
automatic personal identification system based on biometrics.
Biometric technology is an automated method for recognizing
an individual based on physiological or behavioral character-
istics. Among the present biometric traits, iris is found to be
the most reliable and accurate [1] due to the rich texture of
iris patterns, persistence of features through the life time of
an individual and it is neither duplicable nor imitable. These
characteristics make it more attractive for used as a biometric
feature to identify individuals.
RELATED WORK
The iris feature extraction process is roughly divided into
three major categories: the phase-based method [1], [2], [3],
[5], [10], zero-crossing representation [4] and texture analysis
based method [8], [11], [9], [12], [6], [13], [14], [7]. The well-
known phase based methods for feature processing are Gabor
wavelet, Log-Gabor wavelet. The 1 D wavelet is known for the
zero-crossing representation. The Laplacian of Gaussian filter
and Gaussian-Hermite moments are used in texture analysis
based method.
Daugman [1], [2], [3] uses the 2D version of Gabor fil-
ters [15], [16] to extract the iris features and demodulates
the output of the Gabor filters in order to compress the data.
Demodulation is done by quantizing the phase information
into four levels for each possible quadrant in the complex
plane. These four levels are represented using two bits of
data. In other words, each pixel in the normalized iris pattern
corresponds to two bits of data in the iris template. A total
of 2,048 bits are calculated for the template, and an equal
number of masking bits are generated in order to mask out
corrupted regions within the iris. This creates a compact 256-
byte template. Vasta et al. [5] use Log-Gabor filter for iris
feature extraction. Boles and Boashash [4] calculated zero-
crossing representation of 1D wavelet transform at various
resolution levels of a virtual circle of an iris image to char-
acterize the texture of the iris.
CONCLUSION
In this paper, we present a novel and efficient approach
to extract iris features and matching technique to compare
iris features. Our approach uses Daubechies wavelet transform
with four coefficients. The Daubechies wavelet transform is
easy to compute and fast compared to the other methods on
texture analysis. Further, Daubechies wavelet transform allows
to keep the count of feature vectors into a significantly lesser
numbers. This is indeed without affecting the accuracy of the
results. This way we are able to represent feature vectors with
only 304 bits. Although Ali et al. [6] propose an approach
to represent feature vectors with 87 bits but the accuracy
rate is very poor. In addition to the savings in the number
of feature vectors, another contribution in our approach is to
assign a weight to each feature vector and hence to increase
more accuracy in the similarity measure.