Department of Computer Science
Università degli Studi di Milano
Some biometric traits can be captured by using technologies that does not require the contact of the user with a sensor. Theoretically, it is possible to design biometric systems that capture these traits without subjects’ knowledge and in uncontrolled scenarios. This type of biometric systems should be useful in different applicative scenarios (e.g. police investigations, surveillance, etc.). In this context, iris is one of the most interesting biometric traits, since iris recognition systems are one of the most accurate biometric techniques. The behavior of most traditional iris segmentation algorithms is designed for controlled/low-noise environments and it is typically not satisfactory in noisy images, where specific methods are needed to successfully cope with this peculiar applicative contest. The work we propose deals with the segmentation of the iris patterns in this type of noisy images.
A finalist method of the interational competition NICE.I
Examples of the obtained results
Multi-agent pupil center localization
Multiple views iris boundaries refining