Patch Based Sclera and Periocular Biometrics Using Deep Learning

Biometric authentication has become an essential security aspect in today's digitized world. As limitations of the Unimodal biometric system increased, the need for multimodal biometric has become more popular. More research has been done on multimodal biometric systems for the past decade. sclera and periocular biometrics have gained more attention. The segmentation of sclera is a complex task as there is a chance of losing some of the features of sclera vessel patterns. In this paper we proposed a patch-based sclera and periocular segmentation. Experiments was conducted on sclera patches, periocular patches and sclera-periocular patches. These sclera and periocular patches are trained using deep learning neural networks. The deep learning network CNN is applied individually for sclera and periocular patches, on a combination of three Data set. The data set has images with occlusions and spectacles. The accuracy of the proposed sclera-periocular patches is 97.3%. The performance of the proposed patch-based system is better than the traditional segmentation methods.