Shape Signature for Retinal Biometrics

Abstract

The technique used in the only commercially available system for retinal biometrics is based on encoding the vessel structure surrounding the optic disc. However, it has been reported that this technique has low inter-personal variability, making it unsuitable for identifying an individual from a large database. In this paper we propose a new technique based on the shape signature to quantify the contours of the retina structure to increase the inter-personal distance and to lessen the effects of growth of new blood vessels around the optic disc. Correlations are computed as a measure of similarity of 1560 unique retina pairings between different eye images. The results are obtained from a modified shape signature technique of quantifying the contour of the retina structure particularly for the branch-like vessels. The distribution of the correlation, which range from -0.54 to 0.76, resembles a normal distribution with a mean of 0.09. Effects against translation, rotation, scaling and noise are also investigated.