Texture analysis of iris biometrics based on adaptive size neighborhood entropy and linear discriminant analysis

Abstract

Novel method for personal identification, based on adaptive size neighborhood entropy of iris images, was proposed. Through the process of segmentation, iris was extracted from other regions of the human eye, geometrically transformed and normalized. Entropy calculations performed for different neighborhood sizes allows simultaneous distinguishing of fine and global iris texture. Described method also allows recognition of images which contain artifacts and their removal from further analysis after application of principal component analysis (PCA). In the last analytical step, linear discriminant analysis (LDA) with training vector set was applied, allowing rigorous classification. Described procedure is suitable for the application in security systems with small number of authorized persons and a high degree of safety.