Iris Region Extraction Using Neural Network Classification Maps

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current iris recognition technologies face challenges in accurately and efficiently extracting the iris region from images, particularly in degraded or noisy conditions, which affects the overall accuracy and speed of the authentication process.

Innovation Solution

The method employs a trained neural network model with convolutional hidden layers, using backpropagation for training, to extract the iris region by estimating geometric parameters such as circles, ellipses, and curves, and generates a classification map to refine the segmentation, improving accuracy and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional iris segmentation methods are used, then the extraction process can be completed, but the accuracy is reduced in degraded or noisy image conditions

Engineering Contradiction:
Improveiris region extraction accuracyVSAvoidimage degradation and noise
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent divides the iris extraction task into multiple stages: initial segmentation to identify candidate regions, followed by refinement stages that progressively improve boundary accuracy. This multi-stage segmentation approach allows the system to handle degraded images by breaking down the complex extraction task into manageable steps, each addressing specific challenges in noisy conditions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary processing steps including noise filtering, contrast enhancement, and edge detection before the main segmentation process. These preliminary actions prepare the degraded image data by reducing noise and enhancing relevant features, making the subsequent segmentation more accurate despite poor image quality.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If complex segmentation algorithms are applied to improve accuracy, then extraction precision increases, but computational time and resource requirements increase

Engineering Contradiction:
Improveiris region extraction accuracyVSAvoidauthentication processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the computation into hierarchical stages: a fast initial segmentation pass that provides coarse accuracy, followed by targeted refinement only in critical regions. This avoids applying computationally intensive algorithms to the entire image, reducing overall processing time while maintaining high accuracy where needed.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different levels of processing complexity to different regions of the image. High-accuracy computationally intensive methods are applied only to critical boundary regions, while less complex methods are used in stable regions. This local differentiation maintains overall accuracy while significantly reducing total computational time.

Inventive Principle:
Principle #3Local quality

3Measurement precision

If multiple processing stages are used to handle degraded images, then extraction accuracy improves, but device complexity increases

Engineering Contradiction:
Improveiris region extraction accuracyVSAvoidsegmentation system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent combines multiple processing functions into integrated modules. For example, noise reduction and edge detection are merged into a single preprocessing stage, and segmentation and boundary refinement are combined in unified algorithms. This merging reduces the number of separate components while maintaining the multi-stage processing benefits for handling degraded images.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3321850B1Method and apparatus with iris region extraction
Publication Date: 2021.08.18 SAMSUNG ELECTRONICS CO LTD
  • EP3321850B1 patent drawingFigure 1
  • EP3321850B1 patent drawingFigure 2
  • EP3321850B1 patent drawingFigure 3

AI summary

A method and apparatus for extracting an iris region is disclosed. The apparatus may generate a classification map associated with an iris region from an eye image using a trained neural network model, estimate a geometric parameter associated with the iris region using the generated classification map, and extract the iris region from the eye image based on the estimated geometric parameter.