Periocular recognition via effective textural descriptor

The region surrounding human eye, termed as periocular region, can act as a potential biometric trait when it comes to secure authentication of human subjects. Periocular region, being larger than the iris region and less vulnerable than the entire face region, has been an emerging biometric trait since last decade. In view of this, the present work is an attempt to develop more secure periocular recognition through the application of wavelets and a texture pattern with more robust thresholding mechanism. The wavelets are helping in the multiresolution analysis of periocular images. Whereas the texture pattern facilitates the representation of periocular region in form of histogram feature vectors. Experiments performed with a challenging publicly available database corroborate the premise of the suggested methodology. The performance measured using popular metrics like equal error rate, genuine acceptance rate and decidability index, validates the hypothesis of the suggested method. Comparison with existing approaches establishes the outperforming nature of the proposed approach.