Drift Correction of Scan Images by Snapshot Referencing

TL;DR

This paper introduces a snapshot-referencing (SSR) drift correction method for scanning electron microscopy that uses high-signal snapshots to correct spatial distortions in spectral mapping, enhancing data accuracy without extra hardware.

Contribution

The authors present a novel software-based SSR approach employing basis functions to model and correct various drift types in spectral imaging data.

Findings

Effectively restores spatial integrity in cathodoluminescence datasets.

Does not require specialized hardware for drift correction.

Applicable to any probe-based analytical technique with fast imaging signals.

Abstract

Reliable quantitative analysis in scanning (transmission) electron microscopy (S(T)EM) is often hindered by image drift during long-duration spectral mapping for elemental analysis or for various material functions. We here present snapshot-referencing (SSR) drift correction, a retrospective approach to eliminate spatial distortions based on the temporal nature of the scanning process; A continuous drift vector for every pixel is calculated for a normalized time-field of the scan pattern (e.g., serpentine or raster) utilizing a high-signal, fast-scan "snapshot" as a drift-free reference to guide the correction of simultaneously acquired analytical maps. To describe the drift, we employed Bezier basis functions to model smooth thermal or mechanical drifts and piece-wise linear basis for high-frequency "spiky" shifts such as those caused by charging. We demonstrate the efficacy of this approach on experimental cathodoluminescence (CL) datasets, showing that it effectively restores spatial integrity to hyperspectral data cubes without the need for specialized hardware. This flexible, software-based solution is broadly applicable to any probe-based analytical technique where a fast imaging signal can be recorded alongside slow spectroscopic data.