Tracking and modeling human iris surface deformation

Iris biometric system is widely used in various authentication applications. Further improvements in iris recognition can be achieved by diminishing error rate caused by deformation of iris patterns during changes in size of the pupil. This paper investigates deformation of iris patterns by tracking iris features using Local Binary Pattern and Elastic Graph Matching techniques. Texture of 31 irises from two databases have been tracked and analyzed. Our experiments indicate that iris surface deformation can be modeled by linear equations. The linear prediction offers the maximum summation of squared error, equals to 0.11. However, the iris surface deforms non-uniformly. The iris near to pupillary margins deforms significantly in comparison to one close the iris root. Therefore, compensation of the iris surface deformation should be performed locally.