Minimization of influence by the drooping eyelid in extracting pupil center for eye tracking systems

Precise pupil center detection is an important factor for gaze tracking in video-oculography (VOG) systems. Existing methods perform well to extract the features when the area of pupil in eye image is clear, whereas, interferences, such as eyelashes, corneal reflection etc., will lead to a low success rate. One main reason is the closure of eyelids. In this paper, a systemic 3D transformation algorithm is proposed to accurately ascertain the pupil center, in spite of the complicating factors mentioned above. Experiments show the good performance of our method. And the pupil center could be extracted accurately, even though only 25% of the pupil is visible.