Neural Image Normalization and Feature Extraction for Iris Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods, such as those described in Patent Document 1, face challenges in improving the performance of recognition using subject images, particularly in iris recognition, due to variations in iris image sizes and the need for enhanced multi-scale feature processing.
Innovation Solution
An information processing system that includes a first neural network for initial processing, a normalization unit for generating a normalized image, a second neural network for feature extraction, and a correction unit for adjusting parameters based on normalized image information, utilizing a common loss function to optimize parameter correction across both networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If normalization processing is performed to standardize iris image sizes, then recognition accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent performs normalization processing as a preliminary step before feature extraction and recognition. By pre-normalizing iris images to standard sizes, the system prepares data in advance for subsequent processing stages, improving recognition accuracy while managing computational load through staged processing.
Solution Approach 2:
The patent divides the processing into distinct segments: first normalization processing to standardize image sizes, then feature extraction, and finally recognition. This segmentation allows each stage to be optimized independently, balancing accuracy requirements with processing efficiency.
2Reliability
If multi-scale feature processing is applied to enhance feature performance, then recognition performance is improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent applies multi-scale feature processing by analyzing iris images at multiple resolution levels or scales. This dimensional approach to feature extraction captures both fine details and broader patterns, enhancing recognition performance while organizing complexity through structured multi-level processing.
Solution Approach 2:
The normalization processing unit serves multiple functions: it standardizes image sizes for consistent processing, prepares data for multi-scale feature extraction, and optimizes input for subsequent recognition stages. This multi-functionality reduces overall system complexity by consolidating preprocessing tasks.
3Adaptability or versatility
If separate neural networks are used for different processing stages, then functional flexibility is improved, but parameter optimization difficulty increases
Solution Approach 1:
The patent implements feedback mechanisms where recognition results and performance metrics are used to adjust and optimize parameters in both the normalization processing unit and feature extraction network. This feedback loop enables coordinated optimization across multiple networks, managing complexity through iterative improvement.
Solution Approach 2:
The system employs dynamic parameter adjustment where normalization parameters and network weights are continuously optimized based on performance feedback. This dynamic approach allows separate networks to be trained and tuned independently while maintaining overall system coherence through iterative optimization processes.
Data Source
AI summary
An information processing system (100) includes a first processing unit (102), a normalization unit (103), a second processing unit (104), and a correction unit (105). The first processing unit (102) performs first processing using a first neural network with a subject image as an input. The normalization unit (103) performs normalization processing using first output information being a result of the first processing, and generates a normalized image relating to the subject image. The second processing unit (104) performs second processing using a second neural network with the normalized image as an input, and extracts an image feature relating to the normalized image. The correction unit 105 corrects, based on information relating to the normalized image, a first parameter being a parameter used in the first neural network.