Improving Electron Micrograph Signal-to-Noise with an Atrous Convolutional Encoder-Decoder

TL;DR

This paper introduces a novel atrous convolutional encoder-decoder neural network that significantly improves denoising of electron micrographs, outperforming traditional methods in accuracy and consistency on a large dataset.

Contribution

The paper presents a new neural network architecture specifically designed for electron micrograph denoising, trained on a large dataset, and benchmarks it against existing restoration techniques.

Findings

Outperforms traditional denoising methods in MSE and SSIM.

Achieves 24.6% and 43.7% improvements in MSE for low and ordinary doses.

Lowest variance in MSE among tested methods.

Abstract

We present an atrous convolutional encoder-decoder trained to denoise 512$\times$512 crops from electron micrographs. It consists of a modified Xception backbone, atrous convoltional spatial pyramid pooling module and a multi-stage decoder. Our neural network was trained end-to-end to remove Poisson noise applied to low-dose ($\ll$ 300 counts ppx) micrographs created from a new dataset of 17267 2048$\times$2048 high-dose ($>$ 2500 counts ppx) micrographs and then fine-tuned for ordinary doses (200-2500 counts ppx). Its performance is benchmarked against bilateral, non-local means, total variation, wavelet, Wiener and other restoration methods with their default parameters. Our network outperforms their best mean squared error and structural similarity index performances by 24.6% and 9.6% for low doses and by 43.7% and 5.5% for ordinary doses. In both cases, our network's mean squared error has the lowest variance. Source code and links to our new high-quality dataset and trained network have been made publicly available at https://github.com/Jeffrey-Ede/Electron-Micrograph-Denoiser