ROI Pattern Denoising with Gaussian Scanline Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional denoising methods for critical dimension scanning electron microscopy result in poor sensitivity and signal-to-noise ratio due to equal weighting of neighboring scan lines, leading to inaccurate measurements of pattern edges.
Innovation Solution
Applying a gaussian weighted distribution to neighboring scan lines based on their positional proximity, using the equation Weight(x) = e(-1/2) * (x - μ/σ)^2, to accurately average scan lines and reduce noise in electron images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simple averaging is used to denoise scan lines, then the denoising process is simple and fast, but the sensitivity and signal-to-noise ratio are poor
Solution Approach 1:
The patent applies different weighting factors to different scan lines based on their local position relative to the center scan line. Neighboring scan lines receive higher weights while distant scan lines receive lower weights, creating a local quality variation that improves denoising accuracy while maintaining computational efficiency through the localized gaussian weighting scheme.
2Device complexity
If uniform weighting is applied to all scan lines, then the calculation is straightforward, but the positional impact of neighboring scan lines is not accurately accounted for
Solution Approach 1:
The patent changes the weighting parameter from a uniform value to a position-dependent gaussian distribution. The weight of each scan line is determined by its distance from the center scan line, with the weighting factor varying continuously according to the gaussian function, thereby accurately reflecting the positional impact while maintaining a relatively simple calculation approach.
Data Source
AI summary
A system and method for denoising a grey scale image of a pattern on a substrate, including: scanning a gaussian beam consecutively across scanlines extending along an x direction, wherein the scan lines are disposed adjacent to each other in a y direction within a predetermined region of the pattern, obtaining waveforms from secondary electrons and backscattered electrons reflected from the scanned pattern; converting the waveforms into a grey scale image, and applying a gaussian weighted distribution for each of the scan lines of the grey scale image and the neighboring scan lines of the grey scale image to account for the impact of the neighboring scan lines to each scan line.