Eigenspace based accurate iris recognition system

Abstract

Biometric object based person identification is now gaining reliability and high acceptability. Iris as biometric object is used for such identification in many important applications. Almost all techniques, that use iris for recognition is not fully accurate. In this paper we present an accurate iris recognition system that is based on feature extraction in the eigen space and treating the iris in the lower dimensionality. Eigen irises are constructed from a set of iris images of a person, and a given iris image is recognized by passing it through different sets of eigen irises obtained from the sets of iris images of different persons in the database. We considered different numbers of eigenirises for each person, and compressed and compressed noisy iris images as well as only noisy iris images for recognition. Results show accurate person identification using this technique.