Detection of feature points in video-based eye-gaze detection

Proposes an image processing algorithm for detecting the two feature points in the eye image, the centers of the pupil and glint, which are used to determine eye-gaze. The algorithm was considered to correctly detect the center of the pupil image altered in shape by eyelashes, eyelids, and so on, and to shorten the computation time. This algorithm is executed after the pupil windowing process, which has been already proposed. The results show that, under remote and normal room lighting conditions, the total algorithm can detect the two feature points in real-time on the altered pupil image including the glass reflection light which occurs when subjects wear corrective eyeglasses.