Related Experiment Videos
A multiscale attention network for mixed artifact suppression in AFM images
Mingzhen Bu1, Yanyan Wang1, Junlong Huang1
1Tianjin Key Laboratory of Information Sensing and Intelligent Control, Tianjin University of Technology and Education, Tianjin, China.
Micron (Oxford, England : 1993)
|June 4, 2026
Summary
We developed MS-HINet-CBAM, a deep learning model that effectively removes noise from Atomic Force Microscopy (AFM) images. This method preserves crucial surface details, unlike previous approaches that either left artifacts or smoothed the data.
Area of Science:
- Nanotechnology
- Materials Science
- Life Sciences
Background:
- Atomic Force Microscopy (AFM) is vital for nanoscale imaging but suffers from complex mixed noise.
- Existing deep learning models struggle to balance artifact removal with preserving fine topographic details.
Purpose of the Study:
- To introduce MS-HINet-CBAM, a novel multiscale attention network for robust AFM image denoising.
- To overcome the limitations of current methods in handling directional line artifacts and stochastic scars.
Main Methods:
- Developed a multi-stage denoising framework integrating a Multiscale (MS) module for long-range noise correlations.
- Incorporated a Convolutional Block Attention Module (CBAM) to adaptively focus on genuine topographic features.
Main Results:
- MS-HINet-CBAM significantly outperforms state-of-the-art methods on simulated and real AFM data.
- The model accurately preserves Root Mean Square (RMS) roughness, unlike standard models that underestimate it.
- Validated on real-world AFM images of organic molecules and bacterial cells.
Conclusions:
- MS-HINet-CBAM offers a robust solution for denoising AFM images, preserving critical surface information.
- The method enhances the reliability of quantitative analysis in materials and life sciences using AFM data.
