Accurate eye center location through invariant isocentric patterns.

Area of Science:

• Computer Vision
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Accurate eye center location is crucial for numerous applications.
• Existing eye-gaze trackers require expensive hardware and have limitations with low-resolution images.
• Appearance-based methods lack the necessary accuracy for low-resolution eye center detection.

Purpose of the Study:

• To develop a robust and accurate method for locating the eye center within the pupil area in low-resolution images.
• To overcome the limitations of current eye-gaze tracking systems for standard webcam imagery.
• To provide a cost-effective solution for eye center localization.

Main Methods:

• Utilized isophote properties for invariance to linear lighting changes (contrast, brightness) and in-plane rotation.
• Implemented a scale-space pyramid to achieve scale invariance.
• Tested robustness against variations in illumination, head pose, scale, occlusion, and eye rotation.

Main Results:

• The proposed method demonstrates significant improvements in accuracy compared to state-of-the-art techniques.
• Achieved reliable eye center localization even under challenging conditions like varying illumination and head poses.
• Maintained low computational costs, making it suitable for real-time applications.

Conclusions:

• The developed method effectively bridges the gap in accurate eye center localization for low-resolution imagery.
• The approach offers a viable and accurate alternative to expensive commercial eye-gaze trackers.
• This technique enhances the potential for widespread use of eye-tracking in various applications using standard imaging devices.