Joint iris boundary detection and fit: a real-time method for accurate pupil tracking

Area of Science:

• Ophthalmology
• Computer Vision
• Biomedical Engineering

Background:

• Accurate pupil tracking is crucial for visual science applications.
• Existing methods often focus on independent contour detection, overlooking joint approaches for pupil boundary fitting.
• A unified method for simultaneous pupil contour detection and fitting is needed.

Purpose of the Study:

• To present a novel global method for simultaneously detecting and fitting circular or elliptic pupil contours.
• To develop an algorithm that performs accurately under non-ideal recording conditions.
• To enhance the speed and robustness of pupil tracking algorithms.

Main Methods:

• A global optimization approach is proposed for simultaneous pupil contour detection and fitting.
• The method assumes a predefined geometric shape (circle or ellipse) for the pupil.
• Analytic formulae for the gradient and Hessian of the objective function were derived, enabling vectorized computation.

Main Results:

• The proposed method achieves consistently accurate pupil contour fitting, even with challenging factors like reflections, droopy eyelids, and tears.
• The algorithm demonstrates significantly improved speed compared to existing related methods due to vectorized calculations.
• Both analytical and numerical comparisons confirm the robustness and accuracy of the new method using real and idealized iris images.

Conclusions:

• The developed global method offers a robust and efficient solution for simultaneous pupil contour detection and fitting.
• This approach is particularly effective under challenging real-world recording conditions.
• The findings advance the field of automated pupil tracking for visual science research.