Merged CNN Architecture for Eye Segmentation and Quality Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional iris code-based biometric identification systems are sensitive to variations such as image cropping, blurring, lighting conditions, and occlusions, and require accurate segmentation of the iris from the pupil and sclera, which is challenging.
Innovation Solution
A convolutional neural network (CNN) with a merged architecture for both eye image segmentation and quality estimation, comprising a segmentation tower and a quality estimation tower, which shares layers to improve segmentation accuracy and filter out poor-quality images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional iris code-based biometric identification systems are used, then personal biometric identification can be achieved, but the system becomes sensitive to variations such as image cropping, blurring, lighting conditions, and occlusions
Solution Approach 1:
The system performs preliminary quality estimation and segmentation before iris code extraction. By evaluating image quality and segmenting the iris region in advance, the system identifies and handles poor-quality images or occluded regions before they affect the biometric identification process, thereby reducing sensitivity to image variations.
Solution Approach 2:
The patent introduces an intermediary quality estimation module and segmentation module between image acquisition and iris code extraction. These intermediary components assess image quality and isolate the iris region, acting as mediators that protect the identification system from harmful variations in lighting, occlusion, and image quality.
2Measurement precision
If accurate segmentation of the iris from the pupil and sclera is performed, then biometric identification accuracy is improved, but the segmentation process becomes challenging due to image quality issues
Solution Approach 1:
The system performs preliminary quality estimation before segmentation. By assessing image quality in advance, the system can adjust segmentation parameters or reject poor-quality images before attempting segmentation, thereby improving segmentation accuracy while managing the complexity of the process.
Solution Approach 2:
The patent replaces traditional manual or rule-based segmentation methods with a neural network-based automated segmentation system. This substitution of mechanical/manual processes with intelligent algorithms improves segmentation precision while reducing the perceived difficulty through automation.
3Productivity
If a merged CNN architecture with shared layers is used for both segmentation and quality estimation, then computational efficiency is improved, but the model complexity increases
Solution Approach 1:
The patent merges the quality estimation network and segmentation network into a single unified CNN architecture with shared layers. By combining these two functions into one model, the system improves processing efficiency through a single forward pass while using the shared layers to reduce redundant computations, thereby managing the complexity through functional integration.
Solution Approach 2:
The shared layers in the merged CNN architecture serve multiple functions simultaneously - they extract features for both quality estimation and segmentation tasks. This multi-functionality of the shared components improves overall processing efficiency while the modular tower structure manages complexity by separating task-specific operations.
Data Source
AI summary
Systems and methods for eye image segmentation and image quality estimation are disclosed. In one aspect, after receiving an eye image, a device such as an augmented reality device can process the eye image using a convolutional neural network with a merged architecture to generate both a segmented eye image and a quality estimation of the eye image. The segmented eye image can include a background region, a sclera region, an iris region, or a pupil region. In another aspect, a convolutional neural network with a merged architecture can be trained for eye image segmentation and image quality estimation. In yet another aspect, the device can use the segmented eye image to determine eye contours such as a pupil contour and an iris contour. The device can use the eye contours to create a polar image of the iris region for computing an iris code or biometric authentication.