Photomask Inspection Using Normalized Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing photomask inspection methods fail to accurately identify foreign particles, especially in regions without pattern information, due to noise and relative position shifts between reflected and transmitted images, leading to erroneous identification.
Innovation Solution
The method involves acquiring image data of photomasks with different layer structures, creating inverted and offset image data, computing normalized correlation with a Gaussian distribution-type kernel, and comparing the results to a threshold to identify foreign particles, while distinguishing between true foreign particles and noise-induced distortions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional inspection methods (reflected-transmitted image comparison) are used, then inspection coverage includes regions without pattern information, but measurement precision deteriorates due to noise and relative position shifts causing erroneous identification
Solution Approach 1:
The patent transforms the inspection approach by changing the mathematical parameters used for defect detection. Instead of directly comparing reflected and transmitted images, it applies normalized correlation with Gaussian distribution-type kernels to the inverted images, transforming the detection parameter from simple pixel difference to statistical correlation measure, thereby improving precision while maintaining coverage
Solution Approach 2:
The patent introduces an intermediary processing step by creating inverted images and applying normalized correlation computation as a mediator between the raw reflected/transmitted images and the final defect identification. This intermediary transformation reduces the direct influence of noise and position shifts on the detection result
2Reliability
If reflected and transmitted images are directly compared to identify defects, then foreign particles can be detected, but reliability deteriorates due to relative position shifts between images causing erroneous identification
Solution Approach 1:
The patent performs preliminary actions by pre-processing the reflected and transmitted images through inversion and normalization before comparison. By computing inverted images and applying offset corrections in advance, the method eliminates position shift issues before the actual defect detection, thereby improving reliability without requiring complex real-time alignment mechanisms
3Measurement precision
If simple threshold comparison is used for defect identification, then processing speed is maintained, but measurement precision deteriorates due to noise causing erroneous identification near pattern edges
Solution Approach 1:
The patent changes the detection parameter from simple threshold-based pixel value comparison to normalized correlation computation. By using correlation coefficients that measure the similarity between image patterns and reference profiles, the method achieves better noise discrimination while maintaining computational efficiency through the use of optimized correlation algorithms
Data Source
AI summary
A photomask inspection method that identifies a foreign particle such as dirt on a photomask with high sensitivity by suppressing erroneous identification due to an influence of noise is provided. The photomask inspection method includes acquiring image data of a photomask having regions with different layer structures on a surface thereof, creating inverted image data by subtracting the image data from pixel value data of the regions, creating offset inverted image data by raising pixel values of the inverted image data by a fixed amount, creating normalized correlation image data by computing a normalized correlation of the offset inverted image data and an offset Gaussian distribution-type kernel, and identifying foreign particles by comparing the normalized correlation image data and a predetermined threshold.