A deformable model of the human iris driven by non-linear least-squares minimisation

This paper describes a deformable model of the human iris which forms part of a system for accurate off-line measurement of binocular eye movements, particularly cyclotorsion (torsion), from video image sequences. At least two existing systems measure torsion from infrared video images by pupil tracking followed by cross-correlation of band-pass filtered iris sectors. Unfortunately, pupil expansion and contraction reduces the accuracy of this method. In addition, infrared iris images typically contain very little texture, so correlation can be unreliable. A five-parameter deformable model of the iris was therefore developed for analysing images taken in visible light. This model can translate (horizontal and vertical eye motion), rotate (torsion) and scale both uniformly and radially (pupil changes). Torsion measurements obtained with the model are repeatable and accurate to within 0.1°; this performance is illustrated by analysing binocular torsion during constant fixation on a stationary target