Abstract: This paper presents an iris recognition algorithm based on fractal dimension in feature extraction. Considering that the iris of Asian people has low contrast texture, a system is devised to get high resolution and good quality images. First, the iris is located by a circle detector according to mathematical morphology, and mapped to polar coordinates space. Then a new definition of fractal dimension -- extreme value weighted mean fractal dimension is given, and with it the feature matrix is extracted. At last, a correlation classifier is used in matching stage. The experimental results show that extreme value weighted mean fractal dimension can extract the features that represent the iris, and the recognition algorithm is precise and rapid.