Open-Eye Facial Image Selection With Person-Specific Eyelid Ranges
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional facial image capture methods fail to consistently ensure that eyes are open, leading to quality issues in images used for authentication or security documents.
Innovation Solution
A method involving two video streams is employed to determine a person's face with open eyes, utilizing a first stream for determining a person-specific eyelid lift range and a second stream for selecting images where eyelids are above a threshold within this range, combined with image parameter optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional trained algorithms are used to select facial images, then the selection process is automated and efficient, but the accuracy of determining whether eyes are open is poor due to individual variations in eyelid characteristics
Solution Approach 1:
The system performs preliminary action by recording a first video stream before the actual facial image capture to collect eyelid movement data. This preliminary data collection enables the determination of person-specific eyelid lift ranges that are then used to accurately assess eye opening status in subsequent images, resolving the inaccuracy of conventional algorithms.
Solution Approach 2:
The invention changes the assessment parameter from fixed algorithmic thresholds to dynamic, person-specific eyelid lift ranges. By determining individual characteristics through preliminary video analysis and using these customized parameters for evaluation, the system achieves both automation and high precision in detecting whether eyes are open.
2Measurement precision
If multiple video streams are recorded to determine person-specific eyelid characteristics, then the accuracy of eye opening detection is improved, but the recording time and processing complexity increase
Solution Approach 1:
The first video stream serves as a preliminary action phase where eyelid characteristics are collected and analyzed. This preliminary data gathering, though requiring additional time, enables accurate person-specific parameter determination that prevents re-recording and ensures quality facial images are captured on the first attempt.
Solution Approach 2:
The system performs self-service by automatically analyzing the first video stream to determine person-specific eyelid lift ranges without requiring manual intervention. This automated characteristic extraction and parameter determination minimizes processing complexity and reduces the time burden on users.
3Productivity
If the threshold for eye opening is based on conventional algorithms, then the selection process is simple and fast, but the selected images often have closed or partially closed eyes resulting in poor quality
Solution Approach 1:
The system changes from fixed conventional thresholds to dynamic person-specific thresholds based on individual eyelid lift ranges. By using customized parameters derived from preliminary analysis, the system maintains fast automated selection while significantly improving the reliability of eye opening detection and overall image quality.
Solution Approach 2:
The system implements feedback by using the determined person-specific eyelid lift ranges to continuously improve the accuracy of eye opening detection. The threshold adaptation based on individual characteristics creates a feedback loop that ensures both speed and reliability in image selection without compromising quality.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for determining an image of a person's face with open eyes, comprising receiving a first video stream having a plurality of first images of the person's face; for each of the first images, determining the eyes of the face in the respective first image, and determining, for each eye, an eyelid position in the respective first image; determining a person-specific eyelid lift range based on the eyelid positions of the eyes determined for the first images; receiving a second video stream having a plurality of second images of the person's face; for each of the second images, determining the eyes of the face in the respective second image, and determining, for each eye, an eyelid position in the respective second image; and selecting at least one second image in which at least one eye has an eyelid position above a threshold value in the person-specific eyelid lift range.