Three-dimensional Deconvolution for Large-angle Illumination ADF-STEM Depth Sectioning
Published in Microscopy and Microanalysis (M&M) 2025, 2025
We propose a three-dimensional deconvolution algorithm for large-angle illumination ADF-STEM depth sectioning, enabling improved depth resolution in atomic-scale imaging.
The method is based on Vector Approximate Message Passing (VAMP), a Bayesian inference algorithm that incorporates prior knowledge of the reconstructed image.
Applied to simulated datasets of Ce-doped w-AlN, the approach achieves sharper localization of dopant atoms along the depth axis compared to conventional methods, demonstrating its effectiveness for 3D defect characterization in crystalline materials.
Keywords: ADF-STEM, Deconvolution, Depth Sectioning, Bayesian Inference, Approximate Message Passing, 3D Reconstruction
Recommended citation: T. Kusumi, H. Ueda, T. Futazuka, M. Hanai, S. Katakami, K. Kawahara, R. Ishikawa, N. Shibata, and M. Okada, “Three-dimensional Deconvolution for Large-angle Illumination ADF-STEM Depth Sectioning,” *Microscopy and Microanalysis*, vol. 31, suppl. 1, ozaf048.712, July 2025. doi:10.1093/mam/ozaf048.712
Download Paper