Two-tier image features clustering for iris recognition on mobile


Nowadays, many smartphones are provided with built-in sensors for the acquisition and the recognition of specific biometric traits of the user. This policy has been adopted since the massive use of such devices brought the user to store sensible data in them as well as effectuate sensitive transactions on-the-move. As a consequence, many biometric systems have been migrated from stand alone to mobile environments. The methodology proposed in the following presents an approach to the iris recognition in visible spectrum. Iris images are first enhanced by a fuzzy color/contrast preserving technique and then passed to a two-tier clustering: the first is based on the linear decomposition of the iris into superpixels; the second one exploits an unsupervised learning network model to built a feature vector of the iris. According to the performance obtained in terms of time and recognition rate, the method is compliant with the needs of real-time and in-movement environments.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) Volume 10147 LNAI, 2017, Pages 260-269
Silvio Barra
Silvio Barra
Assistant Professor

Assistant Professor @ University of Naples