David Young, Hilary Tunley, Richard Samuels
It is useful to find the elliptical outlines of the pupil and iris on images of a human eye, obtained from a head-mounted camera, in order to automate one type of eye tracker. This task becomes highly constrained when a simple model of the eye is used. After calibration, the model has only two degrees of freedom corresponding to pan and tilt movements of the eye. We show how the model's constraints can be built into Hough transform and active contour methods at the lowest level, allowing high performance in speed, reliability and accuracy.
Download compressed postscript file