SEM Image Noise Filtering via GLCM Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional scanning electron microscope (SEM) inspection methods face challenges in aligning heterogeneous images, leading to time-consuming and expensive manual comparisons due to limitations in comparing and aligning images with shape or vector components that need to be similar.
Innovation Solution
A noise filtering method using a conversion model generated by AI algorithms, such as generative adversarial networks (GAN), to convert SEM images into gray level co-occurrence matrices (GLCM), extract statistical characteristics, and determine noise presence, facilitating accurate alignment and measurement error determination in SEM equipment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional inspection methods are used to compare and align images, then images with similar shapes or vector components can be aligned, but heterogeneous images cannot be compared and aligned accurately
Solution Approach 1:
The patent transforms SEM images into GLCM (Gray Level Co-occurrence Matrix) representations, changing the parameter space from raw image pixels to statistical texture features. This transformation enables heterogeneous images to be compared in a unified feature space, resolving the limitation of conventional methods that require similar shapes or vector components for alignment.
2Measurement precision
If manual comparison and alignment are performed when inspection fails, then accurate alignment can be achieved, but the process becomes time-consuming and expensive
Solution Approach 1:
The patent replaces manual mechanical comparison and alignment operations with an automated AI-based system. The system uses conversion models trained on SEM images to automatically transform and align images, substituting the manual process with computational algorithms that achieve comparable or superior accuracy without requiring operator intervention.
3Reliability
If noise is present in SEM images, then image quality deteriorates and alignment becomes difficult, but conventional noise filtering methods are insufficient for heterogeneous images
Solution Approach 1:
The patent converts SEM images into GLCM representations, transforming the image data from pixel intensity values to statistical texture characteristics. This parameter transformation enables effective noise filtering by analyzing texture statistics rather than raw pixel values, making the filtering process adaptable to heterogeneous images that would be difficult to process in their original form.
Data Source
AI summary
A noise filtering method includes converting a scanning electron microscope (SEM) image into a converted design image using a conversion model, converting the converted design image into a gray level co-occurrence matrix (GLCM), extracting statistical characteristics of the GLCM, and determining whether the converted design image includes noise or not based on the statistical characteristics.