Periocular authentication based on FEM using Laplace–Beltrami eigenvalues

Abstract

Periocular structure, a sub-region of face is characterized by spatial content with an ability to authenticate in non-intrusive civic E/M commerce surveillance applications. The current paper proposes a novel biometric authentication technique by utilizing geometric properties enduring a periocular region. Major portion of existing periocular pattern recognition methods is based on skin texture global information. The proposed approach quantifies periocular shape in ways that effectively find perceptually similar images in the database. The proposed work is an enhancement over state-of-the-art method which achieves low dimensional local periocular pattern data preserved on high dimensional non-linear plane. Periocular surface is discretized by triangular mesh cornered on natural edges. The Laplace–Beltrami eigen spectrum derived over the mesh using cotangent weighted Finite Element Method (FEM) implementation is implicitly compared to achieve periocular region shape based authentication.