Abstract: Based on self-similarity features and incremental support vector machine,a new iris recognition method is presented in this paper.According to the self-similarity of the rich texture variation details in irises,a new fractal dimension,or,variation fractal dimension,is proposed as the feature of irises based on the traditional fractal theory.Incremental support vector machines are implemented on different iris patterns acquired under different conditions.The resultant machines are used to recognize the iris patterns being tested.Experimental data shows that the method presented is amenable：the variation fractal dimension represents the variation details of the iris pattern properly,and the robustness of the system is enhanced by the introduction of learning mechanism into the iris recognition method.