Unsupervised Wafer Defect Detection via SVD Noise Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wafer or mask inspection systems face challenges in detecting small defects with low Signal-to-Noise-Ratio (SNR), as defects are often obscured by process variations and noise, making them difficult to identify and classify accurately.
Innovation Solution
Applying Singular Value Decomposition (SVD) to a difference image derived from comparing a current and reference image, where lower-valued singular values are removed to enhance the SNR of the defect image, and optionally emphasizing the defect by multiplying the reconstructed image with a factor, followed by merging it with the reference image to improve defect visibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image inspection methods are used, then inspection coverage is maintained, but defect detection accuracy deteriorates due to low SNR obscuring small defects
Solution Approach 1:
The patent extracts and removes noise components from the difference image through SVD decomposition. By identifying and eliminating the noise subspace (lower-valued singular values), the method isolates the defect signal from harmful noise and process variations, thereby improving defect detection accuracy without compromising inspection coverage
Solution Approach 2:
The patent introduces a difference image as an intermediary between the current and reference images. This difference image serves as a mediator that highlights potential defects while canceling out common process variations, enabling more accurate defect detection before final classification
2Productivity
If full image processing is performed, then comprehensive inspection is achieved, but computational load and processing time increase
Solution Approach 1:
The patent segments the image processing into distinct functional stages: difference image generation, SVD decomposition, noise subspace identification, and defect reconstruction. This segmentation allows each stage to be optimized independently and enables early termination when sufficient defect information is obtained, reducing overall computational load while maintaining inspection quality
Solution Approach 2:
The patent applies partial action by processing only the necessary portions of the image data through SVD decomposition. Instead of analyzing every pixel and component, the method identifies and processes only the significant singular values corresponding to defect signals, reducing computational complexity while maintaining detection accuracy
Data Source
AI summary
A method for increasing Signal-to-Noise-Ratio (SNR) of defect detection in inspection of wafers or masks, the method including receiving a current image, receiving a reference image, receiving an indication for existence of a defect in the current image, producing a difference image between the current image and the reference image, performing singular value decomposition (SVD) on the difference image, removing one or more lower-valued singular values from a diagonal middle matrix produced by the SVD, thereby producing a reduced middle matrix, and producing an improved-SNR difference image by reconstructing the difference image using the reduced middle matrix. Related apparatus and methods are also described.