Self-Supervised Single-Image Deconvolution with Siamese Neural Networks

TL;DR

This paper introduces a fast, self-supervised 3D image deconvolution method using Siamese neural networks and Fourier transforms, outperforming previous approaches in microscopy applications.

Contribution

It proposes a novel Siamese invariance loss and FFT-based convolutions to enhance 3D deconvolution efficiency and accuracy in a self-supervised framework.

Findings

Outperforms previous state-of-the-art deconvolution methods

Uses FFT convolutions for faster training in 3D microscopy

Identifies optimal network architecture for invariance loss placement

Abstract

Inverse problems in image reconstruction are fundamentally complicated by unknown noise properties. Classical iterative deconvolution approaches amplify noise and require careful parameter selection for an optimal trade-off between sharpness and grain. Deep learning methods allow for flexible parametrization of the noise and learning its properties directly from the data. Recently, self-supervised blind-spot neural networks were successfully adopted for image deconvolution by including a known point-spread function in the end-to-end training. However, their practical application has been limited to 2D images in the biomedical domain because it implies large kernels that are poorly optimized. We tackle this problem with Fast Fourier Transform convolutions that provide training speed-up in 3D microscopy deconvolution tasks. Further, we propose to adopt a Siamese invariance loss for deconvolution and empirically identify its optimal position in the neural network between blind-spot and full image branches. The experimental results show that our improved framework outperforms the previous state-of-the-art deconvolution methods with a known point spread function.