Iris Recognition Accuracy via Local Density Confidence Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing iris recognition methods, such as Daugman's, suffer from high rates of positive and negative errors due to not considering the spatial distribution of incoherent bits between binary codes, leading to incorrect identification of iris images.
Innovation Solution
The proposed method utilizes local densities of coherent bits between binary codes to determine a confidence score, focusing on the concentration of coherent bits to improve identification accuracy by calculating a binary similarity code and weighting local densities to decide if two iris images originate from the same iris.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the Hamming distance method is used for iris code comparison, then the identification process is simple and fast, but the error rates (both positive and negative) are high
Solution Approach 1:
The patent divides the binary iris code into multiple local zones or sectors, and calculates local densities of coherent bits for each zone separately. This segmentation allows the system to evaluate similarity in different regions independently, improving reliability by identifying patterns that global Hamming distance might miss, while keeping computational complexity manageable through localized processing.
Solution Approach 2:
The patent applies different evaluation criteria to different local zones of the iris code. By calculating local density metrics for coherent bits in specific regions and comparing these against threshold values, the system adapts the comparison process to the local characteristics of each zone, thereby improving overall identification accuracy without requiring a complete redesign of the comparison methodology.
2Measurement precision
If the conventional binary code comparison is used, then the processing is fast, but negative errors increase (wrongly considering same iris as different)
Solution Approach 1:
The patent pre-calculates and stores threshold values for local densities of coherent bits for each zone, based on training data or statistical analysis. During actual identification, the system only needs to calculate local densities and compare them against pre-established thresholds, rather than performing complex statistical tests from scratch. This preliminary preparation significantly reduces processing time while maintaining high measurement precision.
3Reliability
If more bits are considered for comparison, then identification accuracy improves, but the system becomes more sensitive to non-visible zones and errors increase
Solution Approach 1:
The patent extracts and isolates the local density information for coherent bits in each zone, separating this meaningful information from the potentially harmful influence of non-visible zones. By focusing computation on local densities rather than global bit counts, the system takes out the essential identification features while leaving behind the noise from occluded or non-visible regions, thereby improving reliability without increasing sensitivity to errors.
Data Source
AI summary
The invention relates to a method for identification on the basis of biometric data of an iris of an eye to be identified, including the steps of: encoding an image of the iris to be identified and a second iris image so as to obtain binary codes that are representative of the images to be compared; determining a binary similarity code from the binary code of the image of the iris to be identified and the second binary code of the second iris image; determining a confidence score on the basis of the local densities of similarities between the two compared iris images, as well as on the basis of the binary similarity code, the local similarity densities being in turn determined on the basis of the binary similarity code; and deciding, depending on the value of the confidence score, whether or not the two iris images are from the same iris. The invention also relates to a system suitable for implementing the identification method.