Dual-Edge Detection for Low S/N Ratio SEM Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge detection techniques struggle to accurately identify edges in images with low signal-to-noise ratios, particularly those with small luminance changes, due to noise interference, making it difficult to distinguish between edges and noise.
Innovation Solution
A dual-edge detection technique is employed, using a high-sensitivity peak detection method prone to noise and a noise-resistant method with low sensitivity, where peak positions from both techniques are compared to determine the true edge position, thereby stabilizing edge detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a high-sensitivity peak detection technique is used for edge detection, then the sensitivity to detect edges with small luminance changes is improved, but the influence of noise increases making it difficult to distinguish edges from noise
Solution Approach 1:
The patent combines two edge detection techniques with different characteristics: a high-sensitivity technique (original image differential method) and a noise-resistant technique (smoothing filter method). By merging the results of both techniques and selecting edges that satisfy both detection methods, the system achieves both high sensitivity and noise resistance, resolving the contradiction between detection precision and reliability.
2Reliability
If smoothing processing is applied to reduce noise, then the noise resistance is improved, but the ability to detect edges with small luminance changes deteriorates
Solution Approach 1:
The patent applies smoothing processing through a convolution filter to create a noise-resistant edge detection method, then combines this with a high-sensitivity differential method. The convolution filter uses coefficients that emphasize edge detection while suppressing noise, and by merging results from both the smoothed and original image methods, the system maintains noise resistance while preserving edge detection sensitivity.
3Reliability
If frame addition is performed to improve image quality, then the signal-to-noise ratio is improved, but image quality deterioration due to electrification and contamination occurs
Solution Approach 1:
Instead of performing full frame addition processing that would require multiple scans and expose the sample to repeated electron beam irradiation, the patent applies partial processing by using convolution filters and differential operations on single or fewer frames. This achieves sufficient noise suppression through mathematical processing without the harmful effects of repeated exposure, thereby preventing sample surface contamination and electrification.
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 effectively suppresses noise influence, allowing for robust edge detection in images with low S/N ratios, improving the accuracy of edge identification even in challenging conditions.
Implementation Method 1
detected secondary or reflected electrons, which are generated from the sample surface by the electron beam irradiation
Data Source
AI summary
An object of the present invention is to provide an edge detection technique and equipment which are capable of stably detecting an edge by suppressing the influence of noise even in the case where the image is obtained by charged particle radiation equipment, such as a scanning electron microscope and has a low S/N ratio. More specifically, the present invention is to propose a technique and equipment which are configured to determine a peak position (edge) on the basis of the following two edge extraction techniques. That is, the present invention is to propose a technique and equipment wherein at least two peaks are formed by using, as edge detection techniques, for example, one peak detection technique having a relatively high sensitivity and the other peak detection technique which is relatively less susceptible to the influence of noise than the one peak detection technique, and wherein a position where the peaks coincide with each other is determined as a true peak position (edge position).