Biometrics: Iris Recognition

Iris recognition is one of commonly employed biometric for personal recognition. In this paper, Single Value Decomposition (SVD), Principal Component Analysis (PCA), Automatic Feature Extraction (AFE) and Independent Component Analysis (ICA) are employed to extract the iris feature from a pattern named IrisPattern based on the iris image. The IrisPatterns are classified using a Feedforward Backpropagation Neural Network (BPNN) and Support Vector Machines (SVM) with Radial Basis Function (RBF) kernel with different dimensions and a comparative study is carried out. From the experimental result, it is observed that ICA is the most appropriate feature extraction method for both BPNN and SVM with Gaussian RBF and SVM with Gaussian RBF can classify faster than BPNN.

A biometric recognition system can be used with a number of physiological characteristics (e.g. fingerprint, palmprint, hand geometry, face, iris, ear shape, and retina vein) and behavioral characteristics (e.g. gait, voice, signature and keystroke dynamics) to provide automatic identification of individuals based on their inherent physical and /or behavioral characteristics. Among these biometrics, iris recognition is one of the most accurate and reliable biometric for identification because of following characteristics (i) Iris pattern has complex and distinctive pattern such as arching ligaments, crypts, corona, freckles, furrows, ridges, rings and a zigzag collarette [1]. (ii) possess 266 degrees-of-freedom in variability and uniqueness in the order of one in 1072 [2].

However, iris recognition also has disadvantages. Some parts of the iris are generally occluded by the eyelid and eyelash. The pupil and iris boundaries are not always circles and their centres are not c...

... middle of paper ...

...l intensity (black or almost black). To find the pupil, a linear threshold (of value 70 in present work) is applied to the image as

(1)

where f is the original iris image and g is the thresholded image.

By applying this, pixels with intensity greater than empirical value of 70 are converted to 1 (black) and others are assigned to 0 (white). Some parts of eyelashes satisfy (1), but have a much smaller area than the pupil area. We can remove all small regions other than pupil by applying code segment (2) for each region R if AREA(R) < 2500 (2) set all pixels of R to 0

Thus, pupil region is obtained. Two imaginary orthogonal lines are drawn passing through the centroid of the pupil region and the first pixel with intensity zero, from the center to the extremities is the boundaries of the binarized pupil. The output of this process is illustrated in Figure 2.