Abstract: A cancellable face recognition system needs a method to restore templates from facial images for matching even if the templates have been stolen. In this study, we proposed a face recognition scheme using bit-string representation of facial features, so that, templates can be reconstructed. Bit-string representation enables rapid matching and also is useful for applications that require binary data such as biometric cryptosystems. First, we selected several images of the same person and made a 'standard face' of that person by mixing his/her images. Then, from the standard face of several people we made a standard face for that group. Finally, comparing a certain person with the standard face of the group, we extracted bit-string representation for the person. Moreover, when the bit-string was exposed by accident, it could be restored by changing the standard face of the group. In an experiment with the Yale Face database we showed that, using our proposed method, our cancellable face recognition scheme based on bit-string representation worked appropriately.