Enhanced 2D image reconstruction for 3D structured illumination microscopy using the plug-and-play priors framework

We present PnP-SIM, a reconstruction method that integrates a pretrained denoising neural network as a plug-and-play (PnP) prior within a physics-based Proximal ADMM framework for three-dimensional structured illumination microscopy. The proposed method combines thick-slice two-dimensional processing with a DnCNN prior, enabling efficient reconstruction of selected single focal planes within the sample while incorporating multiple axial slices of the 3D point spread function, thereby suppressing out-of-focus light without requiring full 3D computation. Results from both simulated and experimental data demonstrate that our method outperforms 2D-FairSIM in terms of PSNR and SSIM, achieving a performance that approaches computationally intensive 3D reconstruction methods.