Live Denoising for High Spatiotemporal-resolution STEM Imaging Using Markov Random Field Model

scite shows how a scientific paper has been cited by providing the context of the citation, a classification describing whether it supports, mentions, or contrasts the cited claim, and a label indicating in which section the citation was made.

Header Abstract Content Author and article information References Comments

Record: found
Abstract: found
Conference Proceedings: found

Is Open Access

Live Denoising for High Spatiotemporal-resolution STEM Imaging Using Markov Random Field Model

Published

conference-abstract

Author(s): Taichi Kusumi ¹ , Shun Katakami ¹ , Ryo Ishikawa ² , Kazuaki Kawahara ² , Naoya Shibata ² ^, ³ , Masato Okada ¹ ^,

Publication date (Electronic): 21 January 2025

Conference name: 13th Asia Pacific Microscopy Congress 2025 (APMC13)

Conference date: 2-7 Febuary 2025

Keywords: Scanning transmission electron microscopy, High spatiotemporal-resolution imaging, Bayesian statistics, Live denoising, Markov random field model

Bookmark

Abstract

Annular Dark-field Scanning Transmission Electron Microscopy (ADF-STEM) imaging has been utilized to determine local atomic structures with sub-angstrom spatial resolution and has a unique ability to directly visualize individual atoms. Additionally, with the recent development of a new high-speed scanning probe system, temporal resolution in atomic-resolution STEM imaging has reached 25 frames per second (fps), potentially enabling atomic-resolution in situ observations of material dynamics [ 1, 2]. However, in high-speed STEM imaging, signal-to-noise ratio is significantly reduced, making it difficult to perform atomic-resolution STEM imaging. Therefore, a high-performance live denoising method is essential to achieve high spatiotemporal-resolution observations. Recently, Kawahara et al. applied a total variation (TV) algorithm to STEM imaging and demonstrated its high noise-removal performance [ 3]. However, as their algorithm requires numerous iterative computations, it takes a few minutes to perform denoising on an image of 512 × 512 pixels. We have recently enhanced the TV algorithm on the basis of Bayesian statistics and proposed an extremely fast denoising algorithm. We found that the denoising process is equivalent to the maximum a posteriori (MAP) estimation in the Markov Random Field (MRF) model, a probabilistic model based on Bayesian statistics. By deriving the MAP solution analytically and eliminating iterative computations from the entire process [ 4– 6], we achieved significantly faster denoising [ 7, 8].

To evaluate the noise-removal performance, we applied our algorithm to low-dose STEM imaging. Figure 1(a) and (b) show the experimental and denoised ADF-STEM image of CaF ₂ viewed along the [001] direction, respectively. The Ca and F columns are clearly visualized in Fig. 1(b), where the computation time was 135 ms, underscoring the effectiveness of our algorithm for atomic-resolution STEM imaging [ 7].

Fig. 1

(a) Atomic-resolution experimental ADF-STEM image of CaF ₂ viewed along the [001] direction with the illumination angle of 22 mrad at 200 kV. The image size is 512 × 512 pixels. The scale bar is 1 nm. (b) Denoised image from 20 frames of experimental images with the computational time of 135 ms.

Content

Author and article information

Conference

Publication date (Electronic): 21 January 2025

Electronic Location Identifier: e203

Affiliations

[1 ]Graduate School of Frontier Sciences, The University of Tokyo, Kashiwa, Chiba, Japan

[2 ]Institute of Engineering Innovation, The University of Tokyo, Bunkyo, Tokyo, Japan

[3 ]Japan Fine Ceramics Center, Nagoya, Aichi, Japan

Author notes

*Corresponding author: okada@ 123456edu.k.u-tokyo.ac.jp

Article

DOI: 10.14293/APMC13-2025-0203

SO-VID: b98b5005-0730-44cb-9a38-b659fcc46c92

License:

Published under Creative Commons Attribution 4.0 International ( CC BY 4.0). Users are allowed to share (copy and redistribute the material in any medium or format) and adapt (remix, transform, and build upon the material for any purpose, even commercially), as long as the authors and the publisher are explicitly identified and properly acknowledged as the original source.

Conference name: 13th Asia Pacific Microscopy Congress 2025

Conference acronym: APMC13

Conference number: 13

Conference location: Brisbane, Australia

Conference date: 2-7 Febuary 2025

History

Funding

Funded by: MEXT: Data Creation and Utilization-Type Material Research and Development Project, Japan

Award ID: JPMXP1122712807

Funded by: JST ERATO

Award ID: JPMJER2202

Funded by: JST FOREST, Japan

Award ID: JPMJPR2033

Funded by: Advanced Research Infrastructure for Materials and Nanotechnology Japan

Award ID: JPMXP1222UT0244

Funded by: JSPS KAKENHI Grant-in-Aid for Scientific Research (A), Japan

Award ID: 23H00486

The authors acknowledge funding from MEXT: Data Creation and Utilization-Type Material Research and Development Project, Japan (JPMXP1122712807), JST ERATO (JPMJER2202), JST FOREST, Japan (JPMJPR2033), and the Advanced Research Infrastructure for Materials and Nanotechnology Japan (JPMXP1222UT0244) by MEXT. Additionally, this work was supported by JSPS KAKENHI Grant-in-Aid for Scientific Research (A), Japan (No. 23H00486).

References

R Ishikawa et al., Microscopy 69 (2020), p. 240. https://doi.org/10.1093/jmicro/dfaa017
R Ishikawa et al., Sci. Adv. 10 (2024), p. eadk6501. https://doi.org/10.1126/sciadv.adk6501
K Kawahara et al., Microscopy 71 (2022), p. 302. https://doi.org/10.1093/jmicro/dfac032
Y Nakanishi–Ohno et al., J. Phys. A: Math. Theor. 47 (2014), p. 045001. https://doi.org/10.1088/1751-8113/47/4/045001
H Sakamoto et al., J. Phys. Soc. Jpn. 85 (2016), p. 063801. https://doi.org/10.7566/JPSJ.85.063801
S Katakami et al., J. Phys. Soc. Jpn. 86 (2017), p. 064801. https://doi.org/10.7566/JPSJ.86.064801
T Kusumi, et al., Ultramicroscopy 253 (2023), p. 113811. https://doi.org/10.1016/j.ultramic.2023.113811
T Kusumi, et al., Ultramicroscopy 264 (2024), p. 113996. https://doi.org/10.1016/j.ultramic.2024.113996

Comments

Comment on this article

[1] R Ishikawa et al., Microscopy 69 (2020), p. 240. https://doi.org/10.1093/jmicro/dfaa017

[2] R Ishikawa et al., Sci. Adv. 10 (2024), p. eadk6501. https://doi.org/10.1126/sciadv.adk6501

[3] K Kawahara et al., Microscopy 71 (2022), p. 302. https://doi.org/10.1093/jmicro/dfac032

[4] Y Nakanishi–Ohno et al., J. Phys. A: Math. Theor. 47 (2014), p. 045001. https://doi.org/10.1088/1751-8113/47/4/045001

[5] H Sakamoto et al., J. Phys. Soc. Jpn. 85 (2016), p. 063801. https://doi.org/10.7566/JPSJ.85.063801

[6] S Katakami et al., J. Phys. Soc. Jpn. 86 (2017), p. 064801. https://doi.org/10.7566/JPSJ.86.064801

[7] T Kusumi, et al., Ultramicroscopy 253 (2023), p. 113811. https://doi.org/10.1016/j.ultramic.2023.113811

[8] T Kusumi, et al., Ultramicroscopy 264 (2024), p. 113996. https://doi.org/10.1016/j.ultramic.2024.113996

Live Denoising for High Spatiotemporal-resolution STEM Imaging Using Markov Random Field Model

Conference

Affiliations

Author notes

Article

History

Funding

Categories

Comment on this article