Adaptive Iris Image Comparison Using Local Quality Region Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing iris comparison methods, such as the Daugman method, face challenges in adapting to individual iris specifics and handling regions with varying textures, leading to potential noise inclusion and skewed comparison results due to predefined application point positions and binary processing.
Innovation Solution
A method that determines quality measurements for regions in iris images, merges these measurements, and selects the highest quality regions for encoding into binary codes, allowing for adaptive point placement and robust comparison, especially in images with little texture, without using masks to conceal poor quality areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If predefined positions for points of application are used in the Daugman method, then artifacts such as eyelashes and eyelids are excluded from comparison, but the method cannot adapt to individual iris specifics and processes all regions uniformly
Solution Approach 1:
The patent applies local quality by computing a quality measure for each region of the iris image individually, allowing different regions to be weighted differently based on their specific characteristics. This enables the system to identify and prioritize high-quality textured regions while downweighting or excluding low-quality regions containing artifacts, thereby achieving both reliability through artifact exclusion and adaptability through individualized region assessment.
2Reliability
If a segmentation mask is used to cover artifacts, then unmasked regions can be processed, but all unmasked regions are processed uniformly regardless of texture quality
Solution Approach 1:
The patent extends the segmentation mask approach by introducing local quality measures that evaluate each region's texture characteristics. Instead of uniformly processing all unmasked regions, the system computes quality scores for each region and uses these scores to weight or select regions for comparison, thereby adapting the processing to individual region qualities while maintaining artifact exclusion.
3Measurement precision
If quality measurements are used to weight regions, then textured regions can be prioritized, but regions with little texture receive very low quality values reducing available information
Solution Approach 1:
The patent addresses this issue by modifying the quality measurement parameters and their application. Instead of using quality values as direct weights that could eliminate low-texture regions, the system adjusts the threshold for region selection and uses quality measures to rank and select a sufficient number of regions. This ensures that even regions with moderate texture quality are included if they meet the selection criteria, thereby preserving information from all parts of the iris while still prioritizing high-quality regions.
Data Source
AI summary
A method for comparing at least two iris images comprises determining M measurements each, of quality level associated with M regions each making up the first and second image. Said measurements are centered on M measurement points, the M measurements of the second image corresponding to the M measurements of the first image by the fact that the M measurement points of the second image correspond to the M measurement points of the first image. The method comprises merging the quality measurements, being obtained by the combination of two corresponding measurements belonging to the two images. The method also comprises selecting N regions exhibiting the N highest quality levels. The method also comprises encoding the two images by using the N selected regions to obtain a binary code for each image. Furthermore, the method comprises comparing the two binary codes to quantify the level of similarity between the two images.