Eye Segmentation Using Directional Filters for Iris Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial image processing technologies fail to accurately detect eye locations and colors, especially when users wear eyeglasses, have partially closed eyes, or images are of low quality, due to issues like stochastic noise and poor illumination.
Innovation Solution
A method and system using directional filters to detect the center point and radius of the iris in facial images, reducing the impact of noise and improving accuracy by segmenting the eye area effectively, even in challenging conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pattern recognition algorithms are used to detect eye locations, then eye detection can be performed, but the detection becomes unreliable when users wear eyeglasses, have partially closed eyes, or in low-quality images
Solution Approach 1:
The patent segments the eye detection process into multiple stages: first detecting candidate regions using pattern recognition, then filtering and validating these candidates using geometric constraints and statistical models. This segmentation allows the system to handle challenging cases by processing information in discrete, manageable steps, improving reliability despite the presence of eyeglasses or partial eye closure.
Solution Approach 2:
The patent employs statistical parameters and probabilistic models to characterize eye regions, using parameters such as region size, shape, and pixel intensity distributions. By transforming the detection problem into a statistical parameter estimation problem, the system can robustly identify eye locations even in low-quality images or when eyes are partially closed, as long as the statistical characteristics remain consistent.
2Measurement precision
If existing image recognition algorithms are used to find eye positions, then eye location can be determined, but accuracy deteriorates in images with imperfect data, random noise, and artificial edges
Solution Approach 1:
The patent converts the harmful effect of noise and artificial edges into a beneficial filtering mechanism. By using statistical models and probabilistic approaches, the system learns to distinguish between true eye features and noise-induced artifacts. The random noise and artificial edges, which would normally degrade detection accuracy, are instead utilized to train the statistical models to be more robust, ultimately improving the system's ability to detect eye positions accurately despite the presence of imperfections.
3Measurement precision
If directional filters are used to scan the first area, then the center point and radius of the iris can be determined, but the processing time increases
Solution Approach 1:
The patent performs preliminary actions by first detecting candidate eye regions and establishing initial estimates of the iris center and radius before applying the directional filter scanning. This preliminary processing narrows down the search space and provides initial parameters that guide the subsequent directional scanning, significantly reducing the computational time required while maintaining high detection accuracy.
Solution Approach 2:
The patent applies directional filters in specific directions rather than scanning all possible directions uniformly. By focusing the scanning action on the most likely directions based on preliminary analysis and statistical models, the system achieves accurate iris center and radius detection with reduced processing time, avoiding the excessive computation that would result from exhaustive directional scanning.
Data Source
AI summary
An eye segmentation system for determining a center point of an eye, a radius of an iris of the eye, and an area of interest from a white area (sclera) of the eye. The system includes at least one input interface that receives a facial image of a user, wherein the facial image is captured using an image capturing device and a data processing arrangement that receives the facial image of the user as input data and processes the facial image to segment an eye region of the facial image. The system determines an eye that is suitable for segmentation from a left eye and a right eye using at least one quality metric, determines a center point of the eye using at least one directional filter, determines, using the at least one directional filter, a radius of an iris of the eye from the center point of the eye.