Feature Selection Device for Image Identity Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing feature selection techniques for image signatures fail to optimize performance by not considering both discrimination capability and robustness, leading to suboptimal determination accuracy of image identity, especially when images undergo various alteration processes.
Innovation Solution
A feature selection device that extracts features from original and altered images, evaluating them based on discrimination capability and robustness to select a subset of features that maximizes both criteria, ensuring high accuracy in image identity determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If features are selected based on empirical knowledge or trial and error experiments, then the feature selection process is simple, but the determination accuracy of image identity cannot be optimized
Solution Approach 1:
The patent changes the evaluation parameters from empirical selection to quantitative assessment using discrimination capability and robustness metrics. By defining specific mathematical parameters (discrimination capability as the ability to distinguish different images, robustness as stability under alteration processes), the system objectively selects features that optimize determination accuracy without relying on trial and error.
Solution Approach 2:
The patent replaces the mechanical trial-and-error feature selection process with an automated computational system. The feature selection unit automatically evaluates multiple candidate features using discrimination capability and robustness calculations, substituting human empirical judgment with algorithmic optimization to achieve superior determination accuracy.
2Reliability
If existing feature selection techniques are used, then the selection process is straightforward, but both discrimination capability and robustness cannot be simultaneously optimized
Solution Approach 1:
The patent introduces a new evaluation dimension by simultaneously considering both discrimination capability and robustness as dual criteria. Instead of selecting features based on a single metric or empirical knowledge, the system evaluates features across two dimensions (discrimination and robustness) and selects features that optimize both, achieving comprehensive reliability.
Solution Approach 2:
The patent performs preliminary evaluation of candidate features by calculating their discrimination capability and robustness before final selection. The feature selection unit pre-assesses multiple features using alteration processes and comparison results, identifying optimal features in advance before they are used for actual image identity determination.
3Loss of information
If more features are extracted from local regions, then the image signature contains more information, but the determination accuracy does not improve without proper feature selection
Solution Approach 1:
The patent extracts and selects only the most valuable features from the set of all possible local region features. The feature selection unit identifies and extracts specific features (such as luminance distribution patterns, edge characteristics) that simultaneously provide high discrimination capability and robustness, removing redundant or less effective features to optimize determination accuracy.
Data Source
AI summary
The feature selection device includes a feature extraction unit that extracts M types of features from each of a plurality of original images and each of a plurality of altered images obtained by applying an alteration process to the plurality of original images; and a feature selection unit that handles an original image and an altered image of the original image as identical images and handles altered images of the same original image as identical images, while handles other images as different images, and with use of discrimination capability which is a degree of discriminating different images and robustness which is a degree that a value of a feature does not vary due to the alteration process applied to an image as evaluation criteria, evaluates the M types of features extracted from the respective images, and selects a collection of N types of features, the N types being smaller in number than that of the M types, from the M types of features as features for discriminating images.