Frontal Iris Reconstruction via Precomputed Transforms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional iris recognition technologies face difficulties in accurately identifying individuals from images taken at off-angle perspectives, leading to reduced recognition performance due to issues like corneal refraction and limbus occlusion.
Innovation Solution
The implementation of an anatomically accurate eye model combined with ray tracing techniques to reconstruct a frontal view of the iris from off-angle images, accounting for corneal refraction and limbus effects, and using precomputed transforms to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional iris recognition is used on off-angle images, then the system is simple to operate, but recognition accuracy deteriorates due to corneal refraction and limbus occlusion
Solution Approach 1:
The system performs preliminary actions by precomputing transformation matrices for various gaze angles before recognition. These precomputed transforms are stored and applied during operation to correct off-angle images, eliminating the need for complex real-time calculations while maintaining high recognition accuracy
Solution Approach 2:
An intermediary computational layer is introduced that transforms off-angle iris images into frontal-view equivalents using precomputed transformation matrices. This intermediary step corrects for corneal refraction and limbus occlusion effects, allowing conventional recognizers to process corrected images with high accuracy without modifying their core algorithms
2Measurement precision
If precomputed transforms are used to correct off-angle images, then recognition accuracy improves, but computational overhead increases
Solution Approach 1:
Transformation matrices for correcting various gaze angles are precomputed and stored in advance. During recognition, these precomputed transforms are simply applied to off-angle images rather than calculating transforms in real-time, significantly reducing processing time while maintaining improved match score distributions
Solution Approach 2:
The system dynamically selects and applies the appropriate precomputed transformation matrix based on the detected gaze angle of the input image. This dynamic selection allows the system to optimize processing by using the most suitable precomputed transform for each specific case, balancing accuracy improvement with computational efficiency
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
Significantly enhances iris recognition performance even at large off-angle gaze positions, improving match score distributions and reducing false reject rates, with up to 146% improvement in match score distribution mean for 50-degree off-angle images.
Implementation Method 1
Corneal refraction can be modeled
Data Source
AI summary
Iris recognition can be accomplished for a wide variety of eye images by correcting input images with an off-angle gaze. A variety of techniques, from limbus modeling, corneal refraction modeling, optical flows, and genetic algorithms can be used. A variety of techniques, including aspherical eye modeling, corneal refraction modeling, ray tracing, and the like can be employed. Precomputed transforms can enhance performance for use in commercial applications. With application of the technologies, images with significantly unfavorable gaze angles can be successfully recognized.