Self-supervised denoising strategies for multimodality light-field microscopy

Light-field fluorescence microscopy (LFM) enables high-speed volumetric imaging by capturing 3D information in a single snapshot. However, its reconstruction quality is severely degraded by detection noise. We propose SSD-LFM, a self-supervised denoising framework that incorporates physical models, to substantially remove noise-induced artifacts in multimodality LFM. The SSD-LFM model is trained e…