Pattern Inspection Reference Image Generation via Low-Pass Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pattern inspection apparatuses face challenges in generating highly precise reference images due to the loss of high-frequency components when estimating coefficients using sample images with reduced resolution, leading to pseudo defect detection issues.
Innovation Solution
The apparatus includes a magnification conversion unit, low-pass filter, optical filter, coefficient acquisition unit, and comparison unit to generate reference image data by convolving design image data with appropriate filtering and model functions, ensuring accurate comparison with actual optical image data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sample optical image data is converted to higher resolution (N times) to match design image data resolution, then the reference image generation accuracy is improved, but high-frequency components are lost causing pseudo defect detection
Solution Approach 1:
The patent applies preliminary action by performing low-pass filtering on the design image data before convolving with the optical model function. This preprocessing step removes high-frequency components from the design image data in advance, ensuring that when the optical model is applied, both the design image data and sample optical image data operate at compatible frequency ranges, preventing pseudo defect detection while maintaining reference image accuracy
Solution Approach 2:
The patent changes the frequency domain parameters of the design image data by applying low-pass filtering. This transforms the design image data from containing high-frequency components to having filtered high-frequency components removed, matching the frequency characteristics of the sample optical image data and enabling accurate coefficient estimation without information loss
2Productivity
If coefficient estimation is performed using sample images with reduced resolution, then processing speed is improved, but measurement precision deteriorates due to high-frequency component loss
Solution Approach 1:
The patent changes the frequency parameters of the design image data by applying low-pass filtering to remove high-frequency components. This parameter transformation allows the system to work with filtered design image data that matches the frequency characteristics of sample optical image data, enabling accurate coefficient estimation without requiring high-resolution sample images, thus maintaining processing speed while improving inspection precision
3Measurement precision
If design image data is convolved with optical model function without filtering, then the theoretical accuracy is improved, but pseudo defect detection occurs due to high-frequency component mismatch
Solution Approach 1:
The patent applies preliminary anti-action by removing high-frequency components from the design image data through low-pass filtering before convolving with the optical model function. This preemptive removal of problematic high-frequency components prevents the generation of pseudo defects that would otherwise occur during the convolution process, ensuring that only genuine defects are detected
Solution Approach 2:
The patent converts the potentially harmful high-frequency components into a benefit by deliberately removing them through low-pass filtering. This transformation eliminates the source of pseudo defect detection while preserving the essential information needed for accurate defect inspection, turning what would be a harmful artifact into a controlled preprocessing step
Data Source
AI summary
A pattern inspection apparatus includes a magnification conversion unit to convert first sample optical image data to higher resolution second sample optical image data, a low-pass filter configured to filter first design image data which has a resolution N times that of the first sample optical image data, an optical filter which calculates third design image data by convolving the second design image data with an optical model function, a coefficient acquisition unit configured to acquire a coefficient of the predetermined optical model function using the second sample optical image data and the third design image data, an optical image acquisition unit configured to acquire actual optical image data of an inspection target workpiece, a reference image data generation unit configured to generate reference image data corresponding to the actual optical image data, and a comparison unit configured to compare the actual optical image data with the reference image data.


