Pupil Region Identification Using Directional Edge Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing technologies fail to accurately detect the pupil region in face images due to noise from eyelid edges and individual variations, especially when the face is not oriented frontally, leading to false detection of eyelid outlines as pupil positions.
Innovation Solution
An image processing device that differentiates the eye region in both horizontal and vertical directions to extract edge points and curves, using a Hough transform with weighted voting for ellipse detection, and identifies the pupil region based on a circle representing the pupil outline and B-spline or Bezier curves for eyelid outlines, enhancing accuracy by considering the positional relationship and luminance gradients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the Hough transform is used to detect circular pupil outlines, then the detection process is simplified, but false detection of eyelid outlines as pupil positions occurs due to noise from long eyelid edges
Solution Approach 1:
The patent segments the edge detection process by direction. It differentiates edges in the vertical direction (yielding first edge points for pupil detection) and horizontal direction (yielding second edge points for eyelid detection). This segmentation allows the system to use different processing strategies for pupil and eyelid edges, resolving the contradiction between simple circular detection and accurate pupil identification.
Solution Approach 2:
The patent applies local quality by treating vertical and horizontal edges differently. Vertical edges (from pupil boundaries) are processed with circle-fitting algorithms, while horizontal edges (from eyelids) are processed with curve-fitting algorithms. This localized differentiation improves measurement precision without unnecessarily complicating the overall detection process.
2Device complexity
If quadratic curves (ellipse or parabola) are used to approximate eyelid outlines, then the processing is simplified, but accurate detection fails when the face is not oriented frontally due to individual variations in eyelid shape
Solution Approach 1:
The patent changes the parameters used to describe eyelid curves from fixed quadratic forms (ellipse, parabola) to flexible B-spline curves defined by multiple control points. This parameter change allows the system to adapt to various eyelid shapes and face orientations while maintaining reasonable processing complexity through automated control point determination.
Solution Approach 2:
The patent introduces dynamics by making the eyelid curve representation adaptable rather than static. B-spline curves with adjustable control points can dynamically conform to different eyelid geometries and face orientations, enabling the system to handle individual variations and non-frontal orientations effectively.
3Productivity
If only pupil outline detection is performed, then the processing is faster and simpler, but accurate pupil region identification fails when part of the pupil is hidden behind eyelids
Solution Approach 1:
The patent merges pupil outline detection with eyelid outline detection into a unified processing framework. By simultaneously detecting both vertical edges (pupil) and horizontal edges (eyelids) and combining their results, the system accurately identifies the pupil region even when partially obscured, without significantly compromising processing efficiency.
Data Source
AI summary
An image processing device for identifying a characteristic of an eye from a face image comprising: a first differentiation unit configured to differentiate an eye region in a crosswise direction of the eye to obtain a first luminance gradient; a first edge extraction unit configured to extract a first and a second edge points; a voting unit configured to vote for an ellipse; a pupil outline identification unit configured to identify an ellipse expressing a pupil outline; a second differentiation unit configured to obtain a second luminance gradient by differentiating in a vertical direction; a second edge extraction unit configured to extract a third edge point; a curve identification unit configured to identify a curve that fits to the third edge point as a curve expressing an eyelid outline; and a pupil region identification unit configured to identify a pupil region based on the identified ellipse and the identified curve.