Image Analyzer Fourier Transform Principal Component Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for developing new materials lack effective visualization methods to identify which regions in material images are affected by changes in feature amounts, making it difficult to grasp important features impacting material performance.
Innovation Solution
An image analyzer comprising a transformer for performing Fourier transforms on material images to acquire power spectra, an analyzer for performing principal component analysis on these spectra, and a reconstructor for generating feature images by reconstructing images based on principal components and scores, and creating difference images to visualize changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If feature amounts are extracted from material images, then material performance analysis becomes possible, but visualization of which regions are affected by feature changes becomes difficult
Solution Approach 1:
The patent segments the image analysis process into distinct functional modules: a transformer module that performs Fourier transform to convert spatial domain images to frequency domain power spectra, an analyzer module that performs principal component analysis to extract feature amounts and their corresponding regions, and a reconstructor module that reconstructs images based on selected principal components. This segmentation allows each module to specialize in one aspect of the analysis, improving both the accuracy of feature extraction and the clarity of regional visualization.
Solution Approach 2:
The patent transforms the problem from spatial domain visualization to frequency domain analysis and back. By performing Fourier transform, the system converts 2D spatial images into power spectra in the frequency domain, where feature amounts can be analyzed through principal component analysis. The reconstructor then maps these frequency domain findings back to spatial domain images, enabling visualization of which specific regions contribute to each feature amount. This dimensional transformation resolves the contradiction by providing a mathematical bridge between feature extraction and regional localization.
2Measurement precision
If traditional image analysis methods are used, then material structure imaging is achieved, but identification of important features affecting performance is difficult
Solution Approach 1:
The patent introduces power spectra as an intermediary representation between the original material images and the final feature analysis results. The Fourier transform generates power spectra that serve as a bridge, allowing principal component analysis to operate on frequency domain data while maintaining the ability to trace results back to spatial regions. This intermediary step enables precise identification of important features through mathematical decomposition, while the modular architecture (transformer-analyzer-reconstructor) manages system complexity by breaking down the analysis into manageable, specialized components.
Data Source
AI summary
For easy visualization of features of an image representing a structure of a material, an image analyzer includes a transformer that performs a Fourier transform on each of a plurality of original images representing a structure of a material and acquires a plurality of power spectra; an analyzer that performs principal component analysis on the plurality of power spectra and acquires principal components and principal component scores; and a reconstructor that reconstructs an image based on the principal components and principal component scores and outputs a feature image representing features of the original image. The reconstructor generates a difference image based on a reconstructed changed image obtained by changing one or more of the principal component scores and a reference image, and outputs the generated difference image as the feature image. The principal component score to be changed may be determined using a trained regression/classification model constructed by machine learning.