Defect Inspection Image Alignment via Peak Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The rapid downscaling of semiconductor devices has made it challenging to accurately align inspection images with reference images for defect inspection, requiring improved methods to enhance alignment speed and precision.
Innovation Solution
A defect inspection method that filters inspection and reference images to identify peak samples, determines candidate correction constants, selects primary and secondary correction constants, and applies these to align the images, facilitating faster and more accurate defect detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full-image alignment methods are used, then alignment precision is improved, but processing time increases
Solution Approach 1:
The patent divides the alignment task into two stages: first aligning using a small number of peak points (key segments) to get rough alignment, then performing fine alignment on specific regions of interest. This segmentation reduces the overall processing time while maintaining precision where needed.
Solution Approach 2:
The patent extracts only the essential peak points from the full image to perform initial alignment, rather than processing all pixels. This extraction of key features significantly reduces processing time while preserving alignment accuracy.
2Measurement precision
If detailed image processing is performed, then alignment accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent extracts only peak points from the full images to perform alignment calculations. By taking out only the essential features (peaks) rather than processing all image data, computational complexity is reduced while maintaining alignment accuracy.
Solution Approach 2:
The patent performs preliminary alignment using peak points before conducting detailed fine alignment. This preliminary action establishes a good initial alignment state, reducing the complexity of subsequent detailed processing.
3Reliability
If complete image data is used for alignment, then alignment reliability is improved, but data processing volume increases
Solution Approach 1:
The patent extracts peak points from complete image data to perform alignment. This extraction reduces the volume of data to be processed from millions of pixels to just a few key points, while the reliability is maintained through subsequent verification steps.
Solution Approach 2:
The patent uses a partial approach by initially aligning with only peak points rather than all image data. This partial action is sufficient for achieving reliable alignment, and full image verification is performed only where necessary.
Data Source
AI summary
A defect inspection method includes: recognizing image peaks that are reference positions of image patterns included an inspection image; performing filtering on a reference image including reference patterns, recognizing reference peaks, and then selecting some of the reference peaks as peak samples; calculating candidate correction constants by overlapping the filtered inspection image and the filtered reference image, and then selecting a primary correction constant among the candidate correction constants; applying the first correction constant to the reference image and selecting a secondary correction constant by matching the image peaks to the reference peaks included in a primary corrected reference image, and then applying the secondary correction constant to the primary corrected reference image and forming a secondary corrected reference image aligned with the inspection image; and performing a defect inspection on the inspection image by matching the image patterns to reference patterns included in the secondary corrected reference image.