In this paper, a novel methodology to achieve the security of the biometric templates in biometric systems is proposed. An adaptive non-uniform quantization (ANUQ) algorithm is introduced to eliminate the contradiction between the fuzziness of the biometric information and the sensitive of the Hash function. The ANUQ algorithm maps different biometric feature vectors from the same individual to a unique quantized feature vector. Then the quantized feature vector can be saved and identified by its hash. A biometric template protection scheme is designed which avoids the storage of users? biometric templates, and further reduce the risk of losing of users? sensitive information. Simulations on an iris recognition algorithm show the feasibility of this method.