NeuHolo: non-interferometric quantitative single-shot holographic imaging for 3D metrology using neural fields

Quantitative holographic imaging records and reconstructs the complex amplitude of a light field. Conventionally, it relies on an interferometric setup with a reference beam that is sensitive to external disturbances. This issue can be addressed by retrieving the complex field from diffracted intensity measurements using a non-interferometric system. However, existing techniques require multiple measurements or additional object support for complex amplitude reconstruction. This paper proposes NeuHolo, a non-interferometric holographic imaging framework based on the integration of neural field network and random phase modulation, which quantitatively estimates amplitude and phase from a single measurement by unsupervised deep learning without object support. Simultaneously, NeuHolo can automatically calibrate the physical parameters, further improving the robustness. We experimentally demonstrate that NeuHolo achieves high precision and large field of view in retrieving complex amplitude. Our study provides a powerful tool of 3D surface characterization for potential industrial applications.