Nonlinear least squares regression for single image scanning electron microscope signal-to-noise ratio estimation.

Area of Science:

• Materials Science
• Image Analysis
• Metrology

Background:

• Accurate signal-to-noise ratio (SNR) estimation is crucial for quantitative analysis of Scanning Electron Microscope (SEM) images.
• Existing SNR estimation methods, including nearest neighborhood and interpolation techniques, have limitations in accuracy and applicability across diverse image textures.
• Developing robust and precise SNR estimation methods is essential for advancing SEM-based research and quality control.

Purpose of the Study:

• To introduce and validate a novel method for estimating the SNR of SEM images using nonlinear least squares regression (NLLSR).
• To compare the performance and accuracy of the NLLSR method against three established SNR estimation techniques.
• To evaluate the NLLSR method's effectiveness on SEM images with varying textural characteristics, contrasts, and edges.

Main Methods:

• Formulation of a new SNR estimation method based on nonlinear least squares regression (NLLSR).
• Comparative analysis of NLLSR against nearest neighborhood, first-order interpolation, and combined interpolation methods.
• Experimental testing using a diverse set of SEM images featuring different textures, contrasts, and edges.

Main Results:

• The NLLSR method demonstrated superior estimation accuracy for SEM image SNR compared to the three existing methods.
• The NLLSR method achieved an SNR error difference of less than 1% relative to the other tested methods.
• The NLLSR method proved effective across SEM images with diverse textural properties and image features.

Conclusions:

• The NLLSR method provides a more accurate and reliable approach for estimating SNR in SEM images.
• This enhanced accuracy has significant implications for quantitative SEM analysis and image quality assessment.
• The NLLSR method offers a valuable advancement for researchers and practitioners utilizing SEM technology.