Registration-based method for correcting nonlinear drift and random jitter in STEM imaging and spectroscopic mapping

Scanning transmission electron microscopy (STEM) images commonly suffer from spatial distortions caused by nonlinear drift and random jitter during acquisition, which compromise spatial accuracy and impede reliable atomic-scale analysis. We present a registration-based method to correct scanning drift in STEM images for two scenarios: (1) single images with periodic structures and (2) multiple images acquired with the same scan direction. For periodic structures, we exploit repeating lattice patterns to construct a high signal-to-noise ratio (SNR) reference image through lattice averaging. For multiple images, we employ multi-frame registration to generate the reference. Using these high-SNR references, we iteratively estimate and correct line-by-line offsets along the slow-scanning direction. This approach significantly improves image quality by reducing artifacts such as streaking in the FFT and atomic column misalignment. We further extend the method to atomic-resolution energy-dispersive X-ray spectroscopy (EDS) mapping, where spatial offsets derived from corresponding HAADF-STEM images correct drift-induced distortions in low-SNR elemental maps. Our results demonstrate that this technique effectively mitigates scanning drift without requiring multi-directional acquisitions and is broadly applicable to both structural and spectroscopic STEM data.