Two-Step Iris Recognition Verification Using 2D Gabor Wavelet and Domain-Specific Binarized Statistical Image Features

The Iris is one of the most reliable biometric features due to its complex textural properties. However, using coloured contact lenses renders the iris unreliable in iris recognition systems. Colored contact lenses are one of the spoofing methods in biometrics that can conceal a person's identity. To prevent spoofing, a two-step verification process is needed in the iris recognition system. The first verification step is to detect colored contact lenses, while the second is to recognize or match a person's identity. The feature extraction methods used are Domain Specific Binarized Statistical Image Features (DSBSIF) and Gabor Wavelet. The method for detecting contact lenses is Support Vector Machine (SVM), and matching is performed using Hamming Distance (HD). This study conducted experiments using single features, feature fusion, and hybrid feature extraction methods combining DSBSIF and Gabor Wavelet for two-step iris recognition verification. The results indicate that the hybrid feature extraction method of DSBSIF and Gabor Wavelet achieved the highest accuracy of 99.95% for the first verification and 95.40% for the second verification. These results are 0.02 and 0.31 percentage points better, respectively than previous methods in the first and second verifications.