Robust Pupil Localization Algorithm under Off-axial Pupil Occlusion

The off-axial pupil is one of several challenges in pupil localization during real-time eye tracking. The off-axial occlusion is a laying condition of a head-mounted eye tracker that is not in a straight line towards the eye. Thus, this occlusion causes the pupil to be recorded in an elliptical shape instead of a perfect circular shape. In the case of off-axial occlusion, pupil localization failure is not only caused by a shifting angle that is too large but also a pupil that is occluded by eyelashes and glints. To solve this problem, this study proposes an ellipse fitting and a fine tuning-based algorithm for robust pupil localization under off-axial condition. The proposed algorithm was tested on 600 labelled images taken from the Swirski dataset. The performance of the proposed algorithm was compared with the modified Circular Hough Transform (CHT) algorithm. The result shows that the proposed algorithm produces a significant increase of 0.83 accuracy with a Z-score of 41. S5(p<; 0.05). In future, this study may be used as a reference for research in the field of pupil localization and eye tracking.