Binary gabor statistical features for palmprint template protection

The biometric template protection system requires a highquality biometric channel and a well-designed error correction code (ECC). Due to the intra-class variations of biometric data, an efficient fixed-length binary feature extractor is required to provide a high-quality biometric channel so that the system is robust and accurate, and to allow a secret key to be combined for security. In this paper we present a binary palmprint feature extraction method to achieve a robust biometric channel for template protection system. The real-valued texture statistical features are firstly extracted based on Gabor magnitude and phase responses. Then a bits quantization and selection algorithm is introduced. Experimental results on the HongKong PloyU Palmprint database verify the efficiency of our method which achieves low verification error rate by a robust palmprint binary representation of low bit error rate