pipette
ESEspañol

Deep-Unfolded Penalized MLEM for Rapid Poisson Image Reconstruction

Zohar Davidov, Alon Osovizky, Max Ghelman, Nir Shlezinger

Preprint

In the authors' words

Maximum likelihood expectation maximization (MLEM) is a common approach for Poisson image reconstruction, but accurate recovery requires many iterations and is sensitive to mismatch in the assumed acquisition model. We propose a deep-unfolded penalized MLEM framework that maps a prescribed small number of iterations into trainable layers while retaining the analytical forward/backward projections and the multiplicative structure of MLEM. The unfolded architecture learns layer-dependent regularization and Poisson-model parameters, together with a data-dependent correction of the sensitivity normalization to mitigate model mismatch. We numerically show that our method improves reconstruction resolution at a fixed iteration budget and attains reconstruction quality comparable to long MLEM runs and direct data-driven models.

Main resultThe abstract does not state a limitation.

Appeared: Friday, October 9. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: 5 pages, 4 figures. Submitted to IEEE ICASSP 2027 (under review)