Charged Particle Beam Image Noise Reduction via Extreme Value Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Defect inspection systems for semiconductor fabrication face challenges in achieving high image quality with reduced scan times, as traditional methods like median filtering often ignore defects and increase image distortion, leading to increased costs and false defect determination rates.
Innovation Solution
A method and system that record grayscale sequences from charged particle beam scans, remove extreme values, and calculate a nominated grayscale value using averaging or median methods to enhance image quality, thereby reducing noise and improving defect detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multi-scans are performed to reduce noise and improve image quality, then image quality is improved, but inspection time and cost increase
Solution Approach 1:
The patent extracts and removes extreme values (outliers) from the grayscale sequences obtained through multi-scans. By identifying and eliminating these extreme values that represent noise, the method preserves the beneficial noise-reduction effect of multi-scanning while reducing the number of scans needed, thus decreasing inspection time while maintaining image quality.
Solution Approach 2:
The patent changes the parameter processing method from traditional averaging or median filtering to a method that removes extreme values based on statistical analysis. This parameter change in the signal processing approach allows for more efficient noise reduction with fewer scans, resolving the contradiction between image quality and inspection time.
2Productivity
If scan times are reduced to lower cost and save time, then inspection cost and time are reduced, but image quality deteriorates
Solution Approach 1:
The patent extracts extreme values from grayscale sequences even when scan times are limited. This extraction of noise components allows the method to achieve better image quality with fewer scans compared to traditional methods, thus improving inspection efficiency without sacrificing image quality.
Solution Approach 2:
The patent applies partial action by performing limited multi-scanning (fewer scans than traditional methods) combined with extreme value removal. This partial scanning approach, when coupled with the extreme value extraction technique, achieves sufficient image quality improvement without requiring excessive scan times, thus resolving the contradiction between productivity and measurement precision.
3Object-affected harmful factors
If traditional median filter is used to remove noise, then noise is reduced, but defects may be ignored and image distortion increases
Solution Approach 1:
The patent applies local quality by treating different grayscale values differently based on their statistical characteristics. Instead of uniformly filtering all values as the median filter does, the method identifies and removes only extreme values (outliers) while preserving normal variations that may represent defects. This localized approach to noise removal maintains defect detection accuracy while reducing noise.
Solution Approach 2:
The patent inverts the traditional filtering approach by not filtering based on median value proximity, but rather by identifying and removing values that deviate significantly from the norm (extreme values). This inverted approach prevents the loss of defect information that occurs with traditional median filtering, as defects are not treated as noise to be removed but as valid data points to be preserved.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances image quality, reduces inspection time and cost, and improves defect detection sensitivity while minimizing false rates, ensuring accurate alignment and defect identification.
Implementation Method 1
The interaction of the electron beam with the inspected surface generates secondary electron signals, which is collected as grayscale images
Data Source
AI summary
This invention relates to methods and systems for enhance the signal-to-noise ratio of an image scanned by a charged particle beam. In an embodiment, a sequence of grayscales of a pixel is recorded first, extreme values of the sequence of grayscales are then identified and removed, and the remained grayscales are used to determine a nominated grayscale of the pixel.