Cancelable Iris Recognition with DPL

The security of users may be one of the challenges that need to be addressed in the practical applications of iris recognition. In order to prevent the theft of iris patterns, non-invertible transformations are desired to be adopted. Combining random projections (RP) with dictionary pair learning (DPL) mode, a cancelable iris recognition scheme is proposed in this paper. Using the framework of DPL, dictionary pair of synthesis dictionary and analysis dictionary can be learnt jointly for iris pattern classification, which greatly reduce the computation complexity without sacrificing match accuracy. And cancelable iris templates are created with random projections. It is very difficult to extract useful information from the transformed iris patterns, which can enhance the related security. The suggested algorithm of cancelable iris recognition with DPL is also robust with the sectored random projections. Experiment results on public datasets have shown that a robust and accurate cancelable iris recognition can be obtained.