The sclera is the opaque white tissue that covers the sphere of the human eye, and this causes a complex nonlinear deformation of the vessel pattern in the sclera as the eyes move. Over the last decade, although the blood vessels of the sclera have been investigated for use as a biometric modality, very little research has been undertaken on feature extraction and matching based on this anatomical structure. To address this problem, we propose a new local spherical structure (LSS) that incorporates a vessel shape feature (VSF) to represent the geometry and texture feature of the vessel pattern. The key ideas in this work are the designs of the VSF and the LSS, which are based on the fact that the vessels lie below a sphere-shaped sclera. To create the proposed VSF, a Harris corner detector is used to extract interest points, each of which is encoded into a normalized polar histogram that represents the shape of a vessel. These points are then projected onto the surface of a sphere, and we design an LSS to represent the topological relationships between each of the points, derived from the arc distance. In the matching step, we adopt a coarse-to-fine procedure that enables a fast similarity comparison between two point pairs using the VSF and the LSS. The final matching score is obtained by combining the number of matched pairs, the total number of interest points, and the similarity. Various experiments were conducted on the UBIRIS.v1 and UTIRIS databases, the latter of which was more challenging. The proposed method showed promising performance, as it yielded an error rate equal to those of other feature extraction and matching methods. Our findings suggest that the proposed approach not only outperforms the state-of-the-art methods of sclera recognition but also illustrates how to address the issue of nonlinear deformation of the vessel pattern based on the anatomical structure of the human eye.
Bibliographical noteFunding Information:
This research was supported by a multi-ministry collaborative R&D program (R&D Program for Complex Cognitive Technology) through the National Research Foundation of Korea (NRF), funded by MSIT, MOTIE, and KNPA (grant number NRF-2018M3E3A1057289 ).
© 2021 Elsevier Ltd
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Artificial Intelligence