FlowDenoising: Structure-preserving denoising in 3D electron microscopy (3DEM)

FlowDenoising is a software tool that implements an adaptive Gaussian denoising filter that preserves visually appreciable structures in volumes of 3D electron microscopy (3DEM). It proceeds by nonrigidly aligning the 2D slices in each dimension, using an optical flow estimator, prior to applying a standard separable (1D) Gaussian filter. FlowDenoising has been developed in Python leveraging well-known public domain libraries, such as OpenCV and NumPy. Furthermore, the software tool exploits data -level parallelism to significantly reduce processing times. Its abilities to denoise huge volumes in just minutes on standard multicore computers makes it a useful tool in 3DEM to explore the interior of cells and tissues at the nanoscale.& COPY; 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Automated summary

FlowDenoising is a software tool that utilizes an adaptive Gaussian denoising filter to preserve visually appreciable structures in 3D electron microscopy volumes. It achieves this by nonrigidly aligning 2D slices in each dimension using an optical flow estimator, followed by applying a standard separable Gaussian filter. Developed in Python, FlowDenoising makes use of well-known public domain libraries like OpenCV and NumPy. It also utilizes data-level parallelism to greatly reduce processing times, allowing for efficient denoising of large volumes on standard multicore computers. It proves to be a valuable tool in 3DEM for exploring the interior of cells and tissues at the nanoscale.