Gaze Detection via Pupil Ellipse Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional gaze detection methods require light sources to illuminate the eye for detecting glints, which may not be efficient or effective in all scenarios, particularly in determining the gaze direction based solely on images of the eye.
Innovation Solution
The method involves capturing an image of the eye and determining a best-fit ellipse for the pupil using edge detection or pixel intensity gradient methods, allowing for the estimation of gaze direction and pupil dimensions without the need for external illumination, by fitting ellipses to the pupil's edge pixels or intensity gradients and using projective geometry to determine a corresponding circle in 3D space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional gaze detection methods using light sources and glint detection are employed, then gaze direction can be determined, but the system requires external illumination which reduces reliability in scenarios without sufficient light
Solution Approach 1:
The system uses the eye's own structural features (pupil shape, iris patterns, corneal reflections) to determine gaze direction, eliminating dependence on external light sources. The pupil is modeled as an ellipse and tracked through image sequences, allowing gaze detection to function autonomously based on the eye's intrinsic properties rather than requiring external illumination.
Solution Approach 2:
The system transitions from detecting glints (reflection-based parameter) to directly tracking pupil geometry parameters (ellipse fit, center position, dimensions). By changing the detection parameter from reflected light patterns to direct pupil structural features, the system achieves reliable gaze detection across varying lighting conditions.
2Measurement precision
If multiple ellipse fitting methods are applied to determine the best-fit ellipse, then measurement precision of pupil position is improved, but computational complexity increases
Solution Approach 1:
The system applies multiple ellipse fitting approaches (edge-based, intensity-based, active contour) but selectively combines them only where necessary. The best-fit ellipse is determined by evaluating multiple methods and selecting the most reliable one for each specific image region, rather than uniformly applying all methods everywhere, thus balancing precision with computational efficiency.
Solution Approach 2:
The system uses confidence metrics to evaluate the quality of each ellipse fit and iteratively refines the pupil detection. By feedback-based evaluation of fit quality and selective application of refinement steps, the system achieves high precision without unnecessarily increasing computational complexity through exhaustive processing.
Data Source
AI summary
Described herein are systems and methods for detecting a gaze direction and/or pupil dimension of an eye based on an image of the eye. Example systems and methods include: (i) capturing an image of an eye; (ii) based on the image, determining a set of edge pixels corresponding with an edge of a pupil of the eye; (iii) selecting at least one subset of pixels from the set of edge pixels; (iv) fitting an ellipse to each subset of pixels; (v) determining a respective confidence value for each fitted ellipse based on a distance between each pixel of the set of pixels and an edge of the respective fitted ellipse; (vi) determining a best-fit ellipse based on the respective confidence value(s); and (vii) determining, based on the best-fit ellipse, at least one of a gaze direction of the eye or a pupil dimension of the eye.