Segmentation-Less and Non-Holistic Deep-Learning Frameworks for Iris Recognition

Hugo Proenca, Joao C. Neves; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 0-0


Driven by the pioneer iris biometrics approach, the most relevant recognition methods published over the years are "phase-based", and segment/normalize the iris to obtain dimensionless representations of the data that attenuate the differences in scale, translation, rotation and pupillary dilation. In this paper we present a recognition method that dispenses the iris segmentation, noise detection and normalization phases, and is agnostic to the levels of pupillary dilation, while maintaining state-of-the-art performance. Based on deep-learning classification models, we analyze the displacements between biologically corresponding patches in pairs of iris images, to discriminate between genuine and impostor comparisons. Such corresponding patches are firstly learned in the normalized representations of the irises - the domain where they are optimally distinguishable - but are remapped into a segmentation-less polar coordinate system that uniquely requires iris detection. In recognition time, samples are only converted into this segmentation-less coordinate system, where matching is performed. In the experiments, we considered the challenging open-world setting, and used three well known data sets (CASIA-4-Lamp, CASIA-4-Thousand and WVU), concluding positively about the effectiveness of the proposed algorithm, particularly in cases where accurately segmenting the iris is a challenge.

Related Material

author = {Proenca, Hugo and Neves, Joao C.},
title = {Segmentation-Less and Non-Holistic Deep-Learning Frameworks for Iris Recognition},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}