A machine vision approach to video based eye tracker calibration

Calibration of eye trackers to establish the relationship between the user's eye and the target environment is important to achieve accurate results. This paper concentrates on video based eye trackers with the gaze target being a computer monitor. Different calibration procedures have been proposed by many people (Sheena and Borah, 1981; Stampe, 1993; Baluja and Pomerleau, 1994) to relate the image of the eye from the tracker to a gaze position on the monitor. In this paper a machine vision approach to calibrating the eye tracker is used; specifically the Tsai camera calibration approach (Tsai, 1987; Lenz & Tsai, 1988). A comparison with two other eye tracker calibration methods is then presented.