Iris Recognition Algorithm Based On Wavelet Transform

With the development of information technology, Information security becomes an important and urgent problem gradually. Therefore, biological characteristic recognition, which can be used to protect the information, is attracting more and more attention. Iris recognition is a new kind of biological characteristic recognition. Compared with the other biological characteristic recognitions (fingerprint recognition, facial recognition voice recognition, etc.), Iris recognition has the following characteristics: stability, high reliability, non-contact.Iris recognition mainly has three stages: Iris image preprocessing, feature extraction, feature matching. This thesis proposes a new iris recognition algorithm based on 2-D wavelet transform. The algorithm has made certain improvement and innovation in each stage of the iris recognition. The identification experiments are done in the CASIA iris database (version 1.0), and good recognition results are achieved.Firstly, in iris localization of the Iris image preprocessing stage, the localization of the interior edge of the iris is more important than the localization of the exterior edge. To localize the interior edge, the thesis first converts the iris image into bi-value image. Then, with the Roberts operator, the interior edge of the iris will be drawn in the bi-value image. Finally, the precise interior edge will be found based on Hough transform.In normalization of the Iris image preprocessing stage, the iris image is normalized into a 64×1024 grey pixel image. After normalization, the normalized image is divided into three parts and then these parts are divided into eighteen bands.Secondly, in the feature extraction stage, Haar 2-D wavelet transform is used in each band of the normalized image. The mean values and variances of the coefficients of 7 main wavelet channels are extracted as features. 252 features will be extracted in each iris image.Finally, in the feature matching stage, with the variance reciprocal used as the weighting coefficients, the features from each parts of the iris image (the first iris image division) are matched separately and then 3 matching results are got. Different weighting coefficients are used to multiply the matching result of each part.Good recognition results are achieved after 40000 recognition experiments have been done in the CASIA iris database (version 1.0).