FIB-SEM (Focused Ion Beam-Scanning Electron Microscopy) is an imaging technique that allows 3D ultrastructural analysis of cells and tissues at the nanoscale. The acquired FIB-SEM data are highly noisy, which makes denoising an essential step prior to volume interpretation. Gaussian filtering is a standard method in the field because it is fast and straightforward. However, it tends to blur the biological features due to its linear nature that ignores the rapid changes of the structures throughout the volume. To address this issue, we have developed a new approach to structure-preserving noise reduction for FIB-SEM. It has abilities to locally adapt the filtering to the biological structures while taking advantage of the simplicity of Gaussian filtering. It uses the Optical Flow (OF) to estimate the variations of the structural features across the volume, so that they are compensated before the subsequent filtering with a Gaussian function. As demonstrated qualitatively and objectively with datasets from different samples and acquired under different conditions, our denoising approach outperforms the standard Gaussian filtering and is competitive with state-of-the-art methods in terms of noise reduction and preservation of the sharpness of the structures.
Keywords: Denoising; FIB-SEM tomography; Gaussian filtering; Optical flow; Structure-preserving noise reduction.
Copyright © 2022 The Author(s). Published by Elsevier B.V. All rights reserved.