Robust pupil tracking algorithm based on ellipse fitting

Nowadays, eye tracking is an emerging research topic and it is entering its fourth era which is characterized by the utilization of video-oculography (VOG). In the VOG system, image processing algorithm is utilized to extract eye movement information. One main process in the VOG system is pupil tracking which is used to measure the movement of the eye. The accuracy of pupil tracking is important factor in the VOG system because the overall performance of VOG system depends on it. The accuracy of pupil tracking decreases significantly when the occlusion of eye is occurred. To increase the accuracy of pupil tracking, we propose a novel algorithm to track pupil in high occlusion condition by utilizing ellipse fitting, RANSAC outlier removal and moving average filtering. The proposed algorithm works well, shown by the significant increase in accuracy during the moment when the pupil is occluded less than 80%. The proposed algorithm can also be utilized in real-time system with an average processing time of 10ms. The high accuracy of the proposed method can increase overall VOG system performance.