How Computers Find Your Pupil Even With Glare

This patent describes a computer method to accurately find the outline of a person's eye pupil by using radial search lines and handling bright reflections differently from clear areas.

What it covers

The patent describes a method for a computer to precisely locate a user's pupil from an eye image. First, it identifies the general "eye area" and any "bright spot areas" within it, which are often reflections (Claim 1). It then sets a central "reference point" within the assumed pupil and draws many "first search lines" extending outwards like spokes on a wheel. For lines that hit a bright spot (a "second search line"), the system figures out how much the bright spot overlaps the pupil by checking brightness around the bright spot's circumference. This helps it set a specific "search range" to find a "first point" on the pupil's edge. For lines that don't hit a bright spot (a "third search line"), it directly finds a "second point" on the pupil's edge. Finally, it uses both the first and second points to map out the entire pupil. For example, if a camera needs to track where a user is looking, this method helps it find the pupil accurately even if there's a camera flash reflection.

What it doesn't cover

• —Pupil detection methods that do not use radial search lines extending from a central reference point.
• —Systems that do not differentiate between search lines passing through bright spots and those that do not.
• —Methods that do not determine a "degree of overlapping" for bright spots or set a "search range" based on it.
• —Pupil detection based solely on overall brightness thresholds without specific bright spot handling logic.
• —Methods that rely purely on machine learning models without the explicit radial search line and bright spot differentiation steps.

The clever bit

The novelty lies in intelligently combining two different strategies for finding the pupil contour: one for areas obscured by bright reflections (using "degree of overlapping" and a "search range") and another for clear areas, allowing for robust detection despite glare.

Why it matters

Accurate pupil detection is critical for eye-tracking technologies used in various fields. This patent offers a way to improve the reliability of pupil detection, especially in challenging lighting conditions where reflections might otherwise confuse a system. This can lead to more precise user interaction in virtual reality, augmented reality, and accessibility tools.

Real-world examples

• 1.Eye-tracking systems in VR/AR headsets
• 2.Driver monitoring systems in cars
• 3.Accessibility tools for computer control
• 4.Gaze estimation for user interfaces