Feature Extraction in an Iris based Recognition System using Independent Component Analysis Algorithm

Abstract

The motivation of this research work is that the iris identification systems are the most authentic biometric system since iris patterns are unparallel to each individual and do not vary with time. After the development of first commercial system given by J. Daugman a variety of methods were innovated to handle eye data in biometric systems. In recent years, considerable progress has been made in the area of iris recognition. Although, the development of the numerous technologies for iris recognition has been made and they have surpassed many iris identification systems, the problem of iris recognition under gross variation remains predominately unresolved. Eye images utilized in iris recognition systems necessitate images to be carried under inflexible constraints. In order to obtain a good quality image, user’s cooperation is required like the user must look directly into camera with a proper illumination in the environment etc. In this thesis, we have focused on the area of iris localization or segmentation and blind source separation. Iris localization is the most significant part of iris recognition which requires extraction of iris boundaries in an iris image. We have developed the frame work for the iris segmentation in which iris boundary has been detected. The framework developed here is suitable only for the CASIA iris database and do not exclude the eyelashes parts inside the iris circular boundary. On the other hand a signal processing based algorithm i.e. independent component analysis (ICA) has been tested on the image which has given significant results. ICA has better performed on iris images in order to render satisfactory measures of accuracy. This dissertation discusses an approach towards less constraint iris recognition using occluded images. The ICA is implemented for a database of 200 images along with the iris localization for making a better iris recognition system.