Human versus biometric detection of texture similarity in left and right irises

The iris codes for the left and right iris of a person have previously been reported to be uncorrelated. We replicate this result using images from 327 persons from the iris image dataset used in the Iris Challenge Evaluations. The same images are then used in an experiment in which subjects view a left and a right iris image, and judge whether they are correctly paired by similarity of iris texture. Subjects are able to distinguish between iris images belonging to the same person versus belonging to different persons with over 86% accuracy overall, and over 93% accuracy when they judge their decision as confident. Thus, there clearly is a similarity in the texture pattern of the left and right iris of the same person that can reliably be detected by novice observers with a brief viewing time. This is the first research that we are aware of that investigates the abilities of human observers in judging similarity or dissimilarity of iris texture.