PRISM: Phase Retrieval via Iterative Spatial-frequency Masking for Ptychography

Ptychographic phase retrieval faces challenges of weak convergence, sensitivity to noise, and overlap ratio. Recognizing that the inherent structure of frequency information significantly impacts reconstruction, we propose phase retrieval via iterative spatial-frequency masking for ptychography (PRISM), a novel, to the best of our knowledge, reconstruction framework that decomposes reconstruction into a frequency-progressive optimization process. PRISM demonstrates improved convergence performance, overlap and noise robustness, and reconstruction quality. Both simulation and experimental results confirm PRISM's superior performance even under challenging imaging conditions with low overlap ratios and high noise levels.