Abstract: A new method for iris feature extraction and recognition was proposed based on the global and local texture information of iris. The normalized iris was divided first. Then the multi-orientation Gabor filters were used to encode the whole iris and sub-block iris, and the feature vector was constructed. Furthermore, the Adaboost algorithm was adopted to choose some features whose recognition performances are good. The results of the experiments carried out on CASIA-IrisV3-Lamp database indicate that the accuracy of recognition reaches up to 99.85%. Moreover, the algorithm has good recognition performance on the NICE: II database which contains a large amount of poor quality images. It shows that the method not only can take fully advantage of the texture information of iris, but also can reduce efficiently the effect of noises.