Texture Analysis Based Iris Recognition

Abstract

In this paper, we present a new method for personal identification, based on iris patterns. The method composed of iris image acquisition, image preprocessing, feature extraction and finally decision stages. Normalized iris images are vertically log-sampled and filtered by circular symmetric Gabor filters. The output of filters are windowed and mean absolute deviation of pixels in the window are calculated as the feature vectors. The proposed method has the desired properties of an iris recognition system such as scale and rotation invariance. The system is realized in Windows platform using Delphi compiler.