Austrian team presents new take on iris ageing
19 January 2017 16:51 GMT

A team of researchers led by Heinz Hofbauer from the University of Salzburg has studied the controversial topic of iris ageing from a different angle, this time isolating template ageing effects in a semi-controlled environment.

The team, which also comprised Andreas Uhl from Salzburg University and Inmaculada Tomeo-Reyes from the Queensland University of Technology, said their study aimed to control as many factors as possible in the biometric tool-chain. This included manual segmentation and the use of a non-linear biomechanical model to alleviate the influence of pupillary dilation.

The subject of Iris ageing centres on the effect of the passage of time on the recognition of individuals based on the analysis of their iris characteristics - the subject is a hotly contested debate in the biometric sector.

While an IREX VI report by NIST Computer Scientist Patrick Grother stated that there is no evidence of an iris template ageing effect, another paper by Kevin Bowyer and Estefan Ortiz  of the University of Notre Dame critically evaluated these findings.

One of the points of disagreement was the definition of iris ageing.

In his paper Hofbauer writes that while the ISO/IEC 19795–1 standard on biometric performance testing and reporting defines template ageing as the increase in error rates caused by time-related changes, that in contrast the IREX VI report excludes effects which are not related to “irreversible changes to the anatomy, primarily the iris texture”.

The IREX VI report states that “dilation should not be considered part of ageing because it varies stochastically and can be mitigated”.

As noted in the new report by Hofbauer, Notre Dame’s Bowyer and Ortiz criticize this view and the fact that, of the assumed reversible effects, only dilation is considered in the report.

The new approach by Hofbauer’s team included the use of an iris database with a four-year time lapse between captures, because this can exclude acquisition factors which could potentially result in apparent iris template ageing.

“We also provide a manual segmentation of the database to minimize the influence of these factors”, write the authors.

In terms of findings, the team writes that despite this controlled approach, a decrease in the pupillary radius with time is still detectable.

“This is consistent with recent medical research results which show that the pupillary radius decreases with age.

“We tried to compensate the effect of the pupillary change by using a biomechanical model for nonlinear iris normalization.

“Despite our efforts, iris template ageing effects are still apparent, with worse iris recognition performance obtained for inter-year comparisons compared to intra-year comparisons. this controlled approach, a decrease in the pupillary radius with time is still detectable."

The full report can be read here.