Pupil Center Detection Approaches: A Comparative Analysis

In the last decade, development of technologies and tools for eye tracking has been a constantly growing area. Detecting center pupil, using image processing techniques, an essential step in this process. A large number techniques have proposed pupil detection both traditional machine learning-based methods. Despite methods proposed, no comparative work on their performance was found, same images metrics. work, we aim at comparing four most frequently cited terms accuracy, robustness, computational cost. These are based circular Hough transform, ellipse fitting, Daugman's integro-differential operator radial symmetry transform. The analysis performed with 800 infrared from CASIA-IrisV3 CASIA-IrisV4 databases containing various types disturbances. best obtained by method transform accuracy average robustness higher than 94%. shortest time, fitting method, 0.06 s.