Template Matching Using Similarity Distribution for Semiconductor Inspection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional template matching techniques for scanning electron microscopes often result in erroneous position determination due to pattern distortion, uneven image luminance, or noise, leading to incorrect coordinate decisions for similar patterns in search images.
Innovation Solution
An inspection apparatus that performs template matching using marginal similarity calculation, search image similarity calculation, and similarity distribution-to-similarity distribution similarity calculation to determine accurate matching positions, even in the presence of noise and distortions, by generating and comparing similarity distribution information across different magnifications and handling rotations or expansions/contractions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional template matching is performed on search images, then matching positions can be determined, but erroneous position determination occurs due to pattern distortion, uneven image luminance, or noise
Solution Approach 1:
The patent divides the template into multiple sub-templates and divides the search image into corresponding sub-search images. By performing template matching on these segmented smaller regions rather than the entire image at once, the method achieves more robust matching that is less sensitive to global distortions and noise, thereby improving reliability while maintaining position determination accuracy
Solution Approach 2:
The patent applies different processing and comparison strategies to different local regions of the template and search image. By analyzing local similarity distributions and considering the specific characteristics of each region, the method adapts to local variations in luminance and distortion, improving both measurement precision and reliability in challenging conditions
2Measurement precision
If template matching is performed to correct stage alignment deviation, then measurement position accuracy improves, but the process becomes complex requiring multiple calculation steps
Solution Approach 1:
By segmenting the template and search image into sub-regions, the patent simplifies the overall matching process into manageable local comparisons. This segmentation reduces the computational burden of comparing entire images while maintaining the ability to correct stage alignment deviation accurately, thus improving measurement precision without excessively increasing process complexity
Solution Approach 2:
The patent transforms the template matching problem by changing parameters such as dividing the images into sub-regions and calculating similarity distributions. This parameter transformation converts a complex global matching problem into simpler local comparison tasks, reducing process complexity while preserving measurement position accuracy
Data Source
AI summary
An inspection apparatus performing template matching of a search image capable of outputting a correct matching position even if a pattern similar to a template exists in the search image is provided. The inspection apparatus includes a template cutout means for cutting out a template from a template selection image, a marginal similarity calculation means for calculating marginal similarity distribution information, which is a similarity distribution of the template selection image to the template, a search image similarity calculation part for calculating search image similarity distribution information, which is a similarity distribution of the search image to the template, a similarity distribution-to-similarity distribution similarity calculation means for calculating similarity distribution-to-similarity distribution similarity information between the marginal similarity distribution information and the search image similarity distribution information, and a matching position determination part for determining a matching position based on the similarity distribution-to-similarity distribution similarity.