Enhancing signal-to-noise in high-resolution in-situ videos with a new AI denoising technique

Gatan instrument used

The K3^® IS camera delivers simultaneous low-dose imaging via real-time electron counting, fast continuous data capture, and a large field of view.

Background

Recent advances in direct detection cameras have vastly improved the efficiency of signal collection in electron microscopy. However, trade-offs between signal-to-noise ratio, spatial/temporal resolution, and dose rate cannot be eliminated by better detection. Even if every electron could be precisely captured, increasing the signal-to-noise ratio while maintaining fixed spatial and temporal resolutions would require a higher dose rate. AI-based denoising techniques can increase the signal-to-noise ratio by taking advantage of correlations between spatiotemporally neighboring pixels, provided the noise is much less strongly correlated. This is the case for very high magnification videos captured with high-quality counting cameras like the K3 IS.