Improvement on Pupil Positioning Algorithm in Eye Tracking Technique

It is significant to accurately locate the pupil position in eye tracking system. And the commonly used method is the algorithm combining Starburst and ellipse fitting. The Starburst algorithm is used to detect the edge of the pupil while the ellipse fitting is supposed to find the optimum ellipse that finely fits the pupil. Then the center of the ellipse is regarded as the location of the pupil. This algorithm is acceptable except that the starting point and the definition of the pupil edge have a great influence on its efficiency. This paper focuses on the accuracy and the swiftness of the primary algorithm and makes some improvement to make up the shortcomings of it. Firstly, mainly localize the pupil position by means of gray projection. Secondly, enhance pupil edge using an edge detecting operator.