Single-Shot High-Density Volumetric Particle Imaging Enabled by Differentiable Holography

Digital holography (DH) stands as a robust imaging method extensively utilized for 3-D imaging. Despite its versatility, when employed in 3-D particle imaging, DH encounters limitations pertaining to resolution and particle concentration. Conventional holographic particle imaging relies on a simplified linearized model and iterative optimization techniques to address pseudoinverse problems. However, this linear assumption inadequately mirrors the complexities of the actual imaging system, resulting in inaccurate particle reconstruction and constrained density. To address these challenges, we introduce an innovative technique named 3-D differentiable holography, which combines a precise theoretical holography model with real-world factors using differentiable programming. This allows us to solve particle imaging problems more effectively using automatic differentiation technology. The effectiveness of our approach is confirmed through numerical simulations and optical experiments, which clearly demonstrate improved imaging accuracy and particle density.