Combining the Goodness of Euler Number and Cumulative Sum to Achieve Higher Accuracy for Iris Recognition Systems

Combining the Goodness of Euler Number and Cumulative Sum to Achieve Higher Accuracy for Iris Recognition Systems N.F. Shaikh, Department of Computer Engineering Modern Education Society‟s College of Engineering Pune, India Dr. D. D. Doye, Department of E&TC SGGS IET, Nanded, India _________________________________________________________________________________________ Abstract—Biometrics has dominated the areas of security for its ability to provide uniqueness, higher accuracy and minimum invasion instances. Using iris and its texture as a means of verification and validation evolved three decades ago and since then it has been one of the most reliable security methods with work being done to increase efficiency along with minimizing cost. The method proposed and implemented considers the two above stated requirements as its essence. Two algorithms have been proposed i.e. Euler number and Cumulative Sum for feature extraction. Both have simple mathematical computations which require less processor time thus reducing cost. Fusion of the two algorithms gives higher accuracy since cases of contradicting results can be cross verified with the individual algorithm score. Euler Number works on topographical features and gives scope for rescaling and resizing. Eight bit planes are obtained of which the four least significant bits are discarded since they give brightness and are random. Using the four most significant bits templates are formed. Cumulative Sum on the other hand explores the uniqueness of every iris template, calculating the cumulative sum where deviation from the mean is calculated and added to the previous value in the template. Each algorithm individually computes its own score which is then fused to obtain a final result. The fusion is carried out after template matching and normalization of the scores, ensuring a more accurate result. Different algorithms for template matching i.e. Vector Difference matching and modified version of Hamming Distance have been implemented. The focus therefore lies in the fusion model which has the underlying assumption that verifying an iris image with not just 1 but multiple algorithms would enhance the system, yielding more accurate results. Keywords-Iris Recognition, Segmentation, Feature Extraction, Fusion, Score, Template Matching. _________________________________________________________________________________________