The maximum a posteriori probability rule for atom column detection from HAADF STEM images.

Area of Science:

• Materials Science
• Electron Microscopy
• Data Analysis

Background:

• High-resolution HAADF-STEM is crucial for nanomaterial structure analysis.
• Beam-sensitive materials require low-dose imaging, leading to low contrast-to-noise ratio (CNR) and unreliable visual inspection.
• Objective and quantitative methods are needed for reliable atom detection in low-quality images.

Purpose of the Study:

• To describe the methodology, derivation, and implementation of the MAP probability rule for atom column detection.
• To compare the performance of the MAP rule against other model-order selection criteria (AIC, BIC).
• To introduce a new image quality measure, integrated CNR (ICNR), correlating better with atom detectability.

Main Methods:

• Utilizes a Bayesian framework for statistical parameter estimation and model-order selection.
• Applies the MAP probability rule to detect atom columns in HAADF-STEM images.
• Introduces and evaluates the integrated CNR (ICNR) as an image quality metric.

Main Results:

• The MAP probability rule demonstrates superior performance in detecting the correct number of atomic columns compared to AIC and BIC.
• The MAP rule provides an objective and quantitative approach for analyzing low-CNR HAADF-STEM images.
• Integrated CNR (ICNR) is proposed as a new image quality measure that shows a stronger correlation with atom detectability than SNR and CNR.

Conclusions:

• The MAP probability rule is an effective and objective method for atom column detection in low-quality HAADF-STEM images, particularly for beam-sensitive nanomaterials.
• The MAP rule surpasses traditional model-order selection criteria in accuracy.
• The newly introduced ICNR metric offers a more reliable assessment of image quality for atom detectability in STEM imaging.