4D Wavefront Sensing Using Dynamic Neural Image Sharpening
Journal of the Optical Society of America A (JOSA A) — 2026
Sensing and correcting atmospheric phase errors is essential for long-range imaging, communication, and astronomy. Building on neural image sharpening (NIS)—which models atmospheric phase errors with implicit neural representations from a single digital-holographic measurement—we introduce dynamic neural image sharpening (DNIS), which learns a 4D functional representation of the phase errors, φ(x, y, z, t), from a temporal sequence of digital-holographic data. By exploiting temporal correlation, DNIS better constrains phase-error estimates, can interpolate between measurement times to increase the effective sampling rate of the sensor, and can extrapolate several frames into the future. In simulation across weak to deep turbulence conditions, DNIS consistently outperforms per-frame NIS, with the largest gains under deep turbulence—suggesting DNIS may ultimately help reduce latency in closed-loop adaptive optics systems.
@article{lloyd20264d,
title={4D wavefront sensing using dynamic neural image sharpening},
author={Lloyd, Robert L and Hardy, Tyler J and Metzler, Christopher A and Spencer, Mark F and Pellizzari, Casey J},
journal={Journal of the Optical Society of America A},
volume={43},
number={7},
pages={C52--C60},
year={2026},
publisher={Optica Publishing Group}
}