Poisson wavefront imaging in photon-starved scenarios

Low-photon phase imaging is essential in applications where the signal is limited by short exposure times, faint targets, or the need to protect delicate samples. We address this challenge with Poisson Wavefront Imaging (PWI), an optimization-based method that incorporates Poisson photon statistics and a smoothness prior to improve wavefront reconstruction. By using multiple spatial light modulator’s phase patterns, PWI enhances Fisher information, boosting theoretical accuracy and regularizing the retrieval process effectively. In simulations, PWI approaches the theoretical phase error limit, and in experiments it reduces phase error by up to 1.6x compared to the Gerchberg-Saxton algorithm, achieving 1.8x higher resolution wavefront imaging in low photon regime. This method advances photon-limited imaging with applications in astronomy, semiconductor metrology, and biological systems.

💡 Research Summary

This paper introduces “Poisson Wavefront Imaging (PWI),” a comprehensive framework designed to achieve high-fidelity phase imaging under extreme photon-starved conditions. Applications such as astronomy, biological imaging with light-sensitive samples, and EUV/X-ray imaging often suffer from severely limited photon counts due to short exposure times, faint targets, or sample damage concerns. PWI addresses this fundamental challenge by synergistically combining a novel hardware architecture with a tailored optimization algorithm that rigorously respects the physics of low-light detection.

The core innovation of PWI lies in its holistic approach. On the hardware side, it employs a coded-detection setup. Light from a complex phase target propagates to a Spatial Light Modulator (SLM), which applies a sequence of distinct phase diversity patterns. These patterned beams then propagate to a camera, producing a set of intensity measurements. Crucially, these multiple patterns ({Φ_i}) enrich the system’s sensitivity to phase variations, acting as strong, user-defined physical regularizers.

The corresponding reconstruction algorithm is meticulously formulated to match the experimental conditions. Instead of the Gaussian noise model used in conventional methods like the Gerchberg-Saxton (GS) algorithm, PWI’s objective function incorporates a negative log-likelihood term based on Poisson statistics, which accurately models shot noise in the low-photon regime. This is combined with a Total Variation (TV) regularization term that promotes piecewise smoothness in the recovered phase, effectively suppressing noise while preserving edges. The resulting constrained optimization problem is solved efficiently using the Alternating Direction Method of Multipliers (ADMM), leading to a computationally tractable iterative solver that alternates between forward propagation and closed-form updates.

The authors provide a rigorous theoretical foundation for PWI by analyzing its Fisher Information and comparing the Cramér-Rao Lower Bound (CRLB) to that of a standard Shack-Hartmann wavefront sensor (SHWFS). Their analysis proves that using multiple random phase patterns in PWI provides significantly higher theoretical phase sensitivity (lower CRLB) than using a single flat pattern or an SHWFS under the same photon budget.

Simulation and experimental results robustly validate the theory. In simulations, PWI with TV regularization achieves a phase reconstruction error that approaches or even dips below its theoretical CRLB (possible for a biased MAP estimator). Experiments using a USAF phase target demonstrate PWI’s dramatic advantage in low-light conditions. While the GS algorithm fails completely at very low flux (e.g., ~2.0 mean photons per pixel), PWI maintains recognizable reconstructions. At a moderate low-light level (66.0 mean photons/pixel), PWI with TV resolves features up to Group 7, Element 1 of the USAF target, corresponding to a spatial frequency of ~128 lp/mm. This represents an approximately 1.8x improvement in resolvable spatial resolution compared to the GS result at the same photon budget, which only resolved up to Group 6, Element 2 (~71.8 lp/mm). In a separate quantitative experiment with a known phase profile imposed by the SLM, PWI reduced the phase RMSE by up to 1.6x compared to GS.

In conclusion, PWI represents a significant advance in computational imaging by demonstrating that through the co-design of measurement physics (multiple phase patterns) and inversion algorithms (Poisson likelihood + TV prior), it is possible to push phase imaging performance dramatically closer to its fundamental limits in photon-starved scenarios. It offers a robust, data-prior-free alternative to deep learning methods, with direct implications for faster, gentler, and more sensitive imaging across multiple scientific and industrial fields.