Performance Analysis of Feature Extraction Techniques for Iris Pattern Recognition System

Iris patterns are very complex and the combination of complexity with randomness confers mathematical uniqueness to a given iris pattern. Once the image is captured, the iris elastic connective tissue is analyzed, processed into an optical “fingerprint,” and translated into a digital form. The fundamental computing concepts at the core of modern biometrics include image processing, pattern recognition, statistics, basic signalling, and some machine learning models such as knowledge based systems and neural nets. In this paper, methods employed for segmentation as Hough transform with methods employ for iris feature extraction are Hough transform, discrete cosine transform and discrete fractional transforms. In order to extract iris features a normalized iris image is divided into patches. The method is effective compared to existing methods. Performance analyses of different feature extraction methods are proposed. For verification, a variable threshold is applied to the matcher and the False Accept Rate (FAR) and False Reject Rate (FRR) are recorded. Experimental results show that the proposed method can be used for personal identification in an effective manner.