A Zero-Shot Deep Image Prior Framework for Denoising and Deconvolution in Fluorescence Microscopy

2026-06-25

AI Summary

Fluorescence microscopy images are often degraded by noise and diffraction-induced blur, hindering quantitative analysis, while existing supervised deep learning approaches rely on scarce paired training data. This work proposes the SDIP framework, which performs denoising and deconvolution sequentially without requiring external training data. It first employs an aSeqDIP module that introduces sequential autoencoding regularization to suppress noise while preserving fine structural details. Subsequently, it integrates wavelet-based background correction with an RLG-DIP module, leveraging Richardson–Lucy deconvolution results as a physically consistent guiding prior to fuse the imaging model with deep image priors, thereby stabilizing the ill-posed deconvolution problem. To our knowledge, this is the first application of physics-guided, zero-shot deep image priors to fluorescence microscopy image restoration, achieving significant improvements in signal-to-noise ratio and resolution on the BioSR dataset, with both visual quality and quantitative metrics surpassing current state-of-the-art methods.

Abstract

Fluorescence microscopy images are degraded by noise and diffraction-induced blur, which compromise structural fidelity and limit quantitative analysis. Supervised deep learning methods achieve impressive restoration performance but require large-scale paired datasets that are difficult to obtain in practice. To address this issue, we propose SDIP, a zero-shot deep image prior (DIP) framework that sequentially performs denoising and deconvolution without external training data. An aSeqDIP-based module first suppresses noise while preserving fine structures through sequential autoencoding regularization. In the deconvolution stage, a wavelet-based background correction step is incorporated before the proposed RLG-DIP module performs artifact-reduced deconvolution. RLG-DIP uses the Richardson-Lucy deconvolution result as a physically consistent guidance prior, integrating the imaging model with the implicit prior of DIP to stabilize the ill-posed deconvolution process. Experiments on the BioSR dataset across multiple cellular structures demonstrate that SDIP improves both signal-to-noise ratio and resolution, achieving superior visual quality and improved quantitative performance on most evaluated structures. The proposed framework may also provide useful insights for designing physically guided DIP methods for other inverse problems.

Problem

• fluorescence microscopy
• image denoising
• image deconvolution
• deep image prior
• zero-shot learning

Innovation

• Zero-shot
• Deep Image Prior
• Deconvolution
• Fluorescence Microscopy
• Physically Guided