Total third-degree variation for noise reduction in atomic-resolution STEM images

Area of Science:

• Materials Science
• Microscopy
• Image Processing

Background:

• Scanning Transmission Electron Microscopy (STEM) is crucial for atomic-level material analysis.
• Electron beam irradiation can damage beam-sensitive materials, necessitating low-dose imaging.
• Existing noise removal methods like Total Square Variation (TSV) can cause image blurring and intensity loss.

Purpose of the Study:

• To develop an improved noise reduction technique for atomic-resolution STEM imaging.
• To overcome the limitations of TSV regularization in preserving image quality for beam-sensitive materials.

Main Methods:

• Proposed a novel denoising approach using L2 norm regularization based on higher-order total variation.
• Utilized Total Third-Degree Variation (TTDV) as a regularization term, suitable for quadratic function approximations of STEM images.
• Applied the TTDV method to denoise atomic-resolution STEM images of CaF2.

Main Results:

• Successfully removed noise from atomic-resolution STEM images.
• Preserved image quality without significant blurring or intensity reduction.
• Clearly visualized Ca and F atomic columns in CaF2.

Conclusions:

• The proposed TTDV-based denoising method effectively enhances atomic-resolution STEM imaging of electron-beam-sensitive materials.
• This technique allows for precise structural analysis without compromising image quality.
• Offers a valuable tool for studying materials like battery components at the atomic scale.