Publication

Abstract : The proposed work presents the technique for designing an M-band statistically matched wavelet for human iris recognition. To analyze the applicability of designed wavelet, Indian Institute of Technology, Delhi (IITD) iris image database is used. It contains iris images of 224 male and female in the age group of 14-55 years. Using automatic cropping, 2-D iris image is obtained. The transformation ring projection method is used to derive 1-D pattern from obtained 2-D iris image. The normalized 1-D statistical pattern is used for designing M-band statistically matched wavelet. To extract features for iris recognition, input iris images are decomposed with their designed statistically matched wavelet up to first level. The statistical features that are predominant for iris recognition are obtained from LL sub-band obtained through first level decomposition. The features similarity is matched on the basis of Sum of Squared Differences (SSD) below a tolerance value with an accuracy of 99.54%.