Comparative Analysis of Biomechanical Normalization Schemes for Iris Recognition
Team: Jiayi Li
- Program: Applied Mathematics and Statistics, Computer Science
- Course:
Project Description:
Iris recognition algorithms are prevalent in today’s high-tech market. However, variations in pupil dilation still present major challenges to this technology. Recent advances showed that integrating biologically-based iris deformation models into the iris normalization scheme, otherwise known as iris biomechanical normalization, is effective for handling these difficulties. Here, we provide a trade-off analysis of two families of iris biomechanical normalization schemes: infinitesimal and approximate large scale deformation where empirical studies, conducted on the West Virginia University Pupil Light Reflex Ramp (WVU-PLR Ramp) database, demonstrate identification accuracies between 98% and 99.79%. These results show that using infinitesimal (small) deformation approximations is a viable alternative while also reducing the additional computational overhead to handle these nonidealities.
