Iris Recognition System Based on Texture Features

Nowadays iris recognition becomes one of the most common methods for identification like password, keys, etc. In this paper, a new iris recognition system based on texture has been proposed to recognize persons using low quality iris images. At first, the iris area is located, and then a new method for eyelash and eyelid detection is applied, the introduced method depends on making image statistical analysis, to isolate the unwanted eyelash/eyelid areas from the true iris area. For discrimination purpose a set of texture features determined from the Haar wavelet subbands is used. Here, the iris area is divided into overlapping blocks and the average energy of each Haar wavelet sub-band is determined to be used as a local feature indicator. Also, the new method depends on using a weighted overlap blocking algorithm to build the iris code. To evaluate the performance of proposed method, it was applied to identify a set of iris images taken from CASIA V4.0 (interval class) data set and CASIA V1 dataset. The test results indicated that the new method give good recognition rates (i.e., 100%) with small size of features vector for CASIA V4.0 and (99.2%) for CASIA V1 dataset.