ABSTRACT

Iris recognition is accepted as one of the most efficient biometric method. Implementing this method to the practical system requires the special image preprocessing where the iris feature extraction plays a crucial role. In this paper we have presented a new approach for iris feature extraction based on Gaussian-Hermite Moments. In the implemented algorithm, iris image is initially located by using circular contour method. Furthermore, intensity normalized flat bed iris image is generated by using Dougman’s rubber sheet model, which is decomposed into a set of 1D intensity signals which retain most local variations of the iris, and then important and meaningful features have been extracted from such signals using Gaussian-Hermite Moments. Euclidian distance is used to measure the degree of dissimilarity between the iris feature vector sets. The recognition performance of the implemented algorithm has been observed. Experimental results show that the algorithm is efficient to describe local information. A CASIA iris database of iris images has been used for implementation.