Video-based eye tracking for autostereoscopic displays

An autostereoscopic display system can provide users the great enjoyment of stereo visualization without the uncomfortable and inconvenient drawbacks of wearing stereo glasses or head-mounted displays. To render stereo video with respect to the user's viewpoints and to accurately project stereo video onto the user's eyes, the left and right eye positions of the user, who is allowed to move around freely, have to be obtained when the user is watching the autostereoscopic display. We present real-time tracking techniques that can efficiently provide the user's eye positions in images. These techniques comprise: 1. face detection by using multiple eigenspaces of various lighting conditions, 2. fast block matching for tracking four motion parameters (X and Y translation, scaling, and rotation) of the user's face, and 3. eye locating in the obtained face region. According to our implementation on a PC with a Pentium III 700 MHz CPU, the frame rate of the eye tracking process can achieve 30 Hz.