High Frame Rate Eye Tracking Algorithm Using Relative Intensity, Geometric Relations, and Depth Information

In a real-time 3D rendering system, eye tracking controls the virtual camera and determines how and what to render on the monitor. To consider the persistence of vision, the refreshing rate of the monitor nowadays is usually beyond 30 frames per second (fps) for each eye. Therefore, a real-time eye-tracking algorithm is needed to control a virtual camera. Without it, the quality of the system decreases. In this study, an efficient eye-tracking algorithm was developed using depth, relative intensity, and geometric relations. With the algorithm with the features extracted from the connected components, the eye was tracked in real time. Experiment results showed that the proposed algorithm executed efficiently with a frame rate of higher than 60 fps.