Periocular Image Correction for Biometric Recognition with Glasses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biometric face recognition systems face challenges in identifying individuals wearing glasses, especially when automated, due to distortion caused by spectacle lenses, and also struggle with obscured facial areas, leading to increased computational effort and decreased accuracy.
Innovation Solution
A device and method that use a camera and image processing to detect the periocular region, account for distortion caused by glasses, and generate a spectacle-free image of the eyes, allowing for reliable identification without removing glasses, by detecting pixels away from the spectacle edge and applying a rule to correct optical effects, and optionally using additional cameras for improved image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If glasses are worn during identification, then the person can maintain natural appearance and comfort, but biometric recognition accuracy deteriorates due to lens distortion
Solution Approach 1:
The patent segments the face image into multiple regions, specifically identifying and isolating the periocular region (area around the eyes) from the rest of the face. By focusing only on this specific segment and excluding the spectacle lenses and frames from the biometric analysis, the system maintains recognition accuracy while allowing the person to wear glasses comfortably.
Solution Approach 2:
The patent extracts and removes the distorted regions caused by spectacle lenses from the biometric analysis. By taking out the affected areas (lenses and frames) and relying solely on the undistorted periocular region for identification, the system eliminates the negative impact of glass wear on recognition accuracy.
2Reliability
If the entire face is analyzed for biometric features, then comprehensive identification is achieved, but computational effort increases
Solution Approach 1:
The patent extracts only the essential periocular region for biometric analysis, removing unnecessary portions of the face that would increase computational load. This extraction approach maintains reliable identification by focusing on stable, distinctive features while significantly reducing the amount of data to be processed.
Solution Approach 2:
The patent segments the full face image to isolate only the periocular region for analysis. By dividing the face into relevant and irrelevant portions and processing only the relevant segment, the system achieves reliable identification with reduced computational effort compared to analyzing the entire face.
3Object-affected harmful factors
If facial areas are obscured by masks or other coverings, then protection or privacy is maintained, but complete face detection becomes difficult
Solution Approach 1:
The patent extracts the detectable periocular region from the partially obscured face image. By focusing on the visible area around the eyes that remains accessible even when other facial regions are covered by masks or protectors, the system maintains identification capability while allowing protective coverings to remain in place.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate identification of individuals wearing glasses and those with obscured facial areas, reducing computational effort and ensuring reliable access control without requiring users to remove their glasses.
Implementation Method 1
a camera (101) for detecting the periocular region (104) of the person
Implementation Method 2
detect those pixels in the course of the facial train which correspond to a distortion of the facial train caused by the spectacle lens
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a device (100) for generating a spectacle-free image of a periocular region (104) of a person wearing spectacles, comprising a camera (101) for capturing the periocular region (104) of the person, and an evaluation unit (102) connected to the camera (101), which is configured, by means of the image supplied by the camera (101), to identify pixels of a gradient (115a, 115b, 115c, 115d) of at least one facial feature (115) located in the periocular region (104) of the person wearing spectacles away from the rim of the spectacles (116) and away from a spectacle lens (114), and to capture those pixels in the gradient (115a, 115b, 115c, 115d) of the facial feature (115) that are subject to a distortion of the facial feature caused by the spectacle lens (114). (115) correspond to deriving or applying at least one rule from the sequence (115a, 115b, 115c, 115d) of the recorded pixels,and subsequently to shift all pixels contained in the area of the spectacle lens (114) according to the determined or applied rule such that the spectacle-free image of the periocular region (104) is created. The invention also relates to a corresponding method, an access control, and a method for granting or denying access.