Performance Analysis of Iris Recognition by using Classical and Trapezoidal Shaped Templates with a State-of-Art Iris Segmentation Technique
Iris recognition is an automated process of authentication of individuals based on their unique characteristics of
iris patterns. These characteristic features can be extracted and encoded by projecting them onto Gabor wavelets and
transforming their phasor information into binary codes. These codes are used to identify or match the individuals by using
some decision metrics. Hamming distance is one of the most known and used matching metric for identifying individuals.
Iris recognition has four stages; image acquisition, image preprocessing, image feature extraction, and image matching.
Proper segmentation of the iris region is an important step where researchers have to solve co-occurring effects such as
eyelid occlusion, eyelashes, or reflection of lights. After the segmentation, the second step is creating an iris template. Here
we modify the well-known rectangular iris template that is created by binary phasor information of individuals into a
different trapezoidal iris template. The tests were conducted employing traditional and recently developed EPK_IRIS
segmentation and different encoding schemes by new defined trapezoidal template under the different noise levels of
Gaussian noise and Salt-and-pepper noise. We report better accuracy and equal error rate (EER) results at some specific
noise levels. In this work, we conclude that the EPK_IRIS segmentation in both template types gives better results.
Index terms - Iris Recognition, Segmentation, Accuracy, EER, Hamming Distance, Iris Templates