A Message-Passing Perspective on Ptychographic Phase Retrieval
Published in IEEE Transactions on Computational Imaging, 2025
We introduce a probabilistic approach to ptychographic reconstruction in computational imaging. This paper formulates ptychography as a Bayesian inverse problem and derives an inference algorithm, Ptycho-EP, based on belief propagation and Vector Approximate Message Passing (VAMP). The method integrates prior knowledge of the object into the model and provides uncertainty quantification for the reconstructed image. Experiments demonstrate that the proposed algorithm achieves reconstruction accuracy near the information-theoretic limit and performs robustly even when the probe is unknown, through an Expectation–Maximization extension.
Keywords: Ptychography, Phase Retrieval, Belief Propagation, Bayesian Inference, Message Passing
Recommended citation: H. Ueda, S. Katakami, and M. Okada, “A Message-Passing Perspective on Ptychographic Phase Retrieval,” *IEEE Transactions on Computational Imaging*, vol. 11, pp. 1–15, 2025. doi:10.1109/TCI.2025.3622992
Download Paper