Ocular Recognition via Multi-Scale Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low-quality iris images lead to loss of iris information, resulting in ineffective matching, as existing technologies struggle to accurately identify subjects from degraded ocular images.
Innovation Solution
A system and method that decompose the ocular region into filtered images of different orientations and scales, using Log Gabor or Gabor filters, and combine these images with adaptively weighted scores to enhance matching across varying image qualities, incorporating holistic ocular appearance recognition and iris recognition for improved identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional iris matching methods are used on low-quality images, then the process is simple, but the matching accuracy deteriorates due to loss of iris information
Solution Approach 1:
The ocular region is decomposed into multiple scale components using Laplacian pyramids, where each scale captures different levels of detail. This segmentation allows the system to preserve both coarse structural information and fine iris patterns across different resolutions, preventing information loss that would occur in traditional single-scale matching approaches.
Solution Approach 2:
The patent extends the matching process from a single 2D image plane to a multi-scale 3D feature space by creating Laplacian pyramid representations. This dimensional extension allows the system to extract and match iris features at multiple scales simultaneously, improving robustness against image quality variations while maintaining comprehensive iris information.
2Measurement precision
If multi-scale decomposition is applied to preserve iris information, then matching accuracy improves, but computational complexity increases
Solution Approach 1:
By segmenting the ocular region into multiple scale bands through Laplacian pyramid decomposition, the system processes information at different levels of detail separately. This allows efficient computation where coarse scales provide robust matching under poor conditions and fine scales enhance accuracy when available, reducing overall computational burden compared to processing full-resolution images exclusively.
Solution Approach 2:
Different scale components are assigned different importance weights based on local image quality assessment. Regions with high quality retain fine-scale details while degraded regions rely on coarse-scale information. This local quality adaptation optimizes computational resources by focusing processing effort where it provides maximum benefit, balancing accuracy improvement with computational efficiency.
3Reliability
If adaptive weighting is used to combine scale scores, then robustness to image variations improves, but processing time increases
Solution Approach 1:
The system dynamically changes the weighting parameters of different scale components based on image quality metrics and scale-specific matching scores. This adaptive parameter adjustment allows the system to prioritize reliable scale information under varying conditions, improving robustness while avoiding unnecessary computation on low-quality scale representations, thus managing processing time efficiently.
Data Source
AI summary
A system and method include decomposing via a computer an ocular region into several filtered images of different orientation and scale, using the computer to combine the decomposed images for each scale, using a computer executed classifier for each scale, matching across different quality images, and using a computer, constructing a matching score by combining the scale scores using adaptively weighted sum for each scale.