Generalizable Holographic Reconstruction via Amplitude-Only Diffusion Priors

The introduction of a new method for phase retrieval in holography marks a significant advancement in computational imaging. By employing a diffusion model trained exclusively on object amplitude, researchers have developed a technique that successfully reconstructs both amplitude and phase from diffraction intensities, eliminating the need for ground-truth phase data. This innovative approach has been rigorously validated through extensive simulations and experiments, showcasing its robust generalization across diverse object shapes and imaging configurations, including complex biological tissue structures. The adaptability of this method, demonstrated through its application to various imaging modalities, highlights its potential as a cost-effective solution for tackling nonlinear inverse problems in the field. This development not only enhances the capabilities of holographic imaging but also lays the groundwork for broader applications in coherent imaging technologies.