88 IRIS SEGMENTATION AND NORMALIZATION APPROACH Mahboubeh Shamsi, Puteh Bt Saad, Abdolreza Rasouli Faculty of Computer Science & Information System University Technology Malaysia, lohor, Malaysia mahboubehshamsi@yahoo.com, drputeh@gmail.com, rs.reza@gmail.com Abstract: Iris is a desirable biometric representation of an individual for security-related applications. However the iris segmentation and normalization process is challenging due to the presence of eye lashes that occlude the iris, the dilation of pupils due to different light illumination and several other uncontrolled factors. In this work, we enhanced Daugman method to locate the iris and normalized it from polar to Cartesian coordinate. Iris is located by using a variable parameter binning approach. The algorithm is tested using iris images from CASIA database and MMU database. The percentage detection on MMU iris database is 99% and that of CASIA is 98%. Our approach is feasible to produce an iris template for identity identification and biometric watermarking application. Keywords: iris recognition, biometric identification, recognition, normalization, automatic segmentation. 1. INTRODUCTION A biometric system provides automatic recognition of an individual based on a unique feature possessed by an individual. Biometric systems have been developed based on fingerprints, facial features, voice, hand geometry, handwriting, the retina (1], and the one presented in this paper, the iris. The first phase of iris Biometric systems is capturing the sample of the iris. Then iris samples are preprocessed and segmented to locate the iris. Once the iris is located, it is then normalized from polar coordinate to Cartesian. Finally a template representing a set of features from the iris is generated. The iris template can then be objectively compared with other templates in order to determine an individual's identity. Most biometric systems allow two modes of operation. An enrolment mode for adding templates to a database, and an identification mode, where a template is created for an individual and then a match is searched from a database of pre-enrolled templates. Iris biometric has the following desirable properties, firstly an iris image is unique, the statistical probability that two irises would be exactly the same is estimated at 1 in 10 72 [20]. Jilid 20. Bi1.3 (Disemoer 2008) Jumal Teknologi Maklumat