Total: 1
Ptychography is a computational imaging technique that achieves high spatial resolution over large fields of view. It involves scanning a coherent beam across overlapping regions and recording diffraction patterns. Conventional reconstruction algorithms require substantial overlap, increasing data volume and experimental time. We propose a reconstruction method employing a physics-guided score-based diffusion model. Our approach trains a diffusion model on representative object images to learn an object distribution prior. During reconstruction, we modify the reverse diffusion process to enforce data consistency, guiding reverse diffusion toward a physically plausible solution. This method requires a single pretraining phase, allowing it to generalize across varying scan overlap ratios and positions. Our results demonstrate that the proposed method achieves high-fidelity reconstructions with only a 20% overlap, while the widely employed rPIE method requires a 62% overlap to achieve similar accuracy. This represents a significant reduction in data requirements, offering an alternative to conventional techniques.