DH-GAN: A Physics-driven Untrained Generative Adversarial Network for 3D Microscopic Imaging using Digital Holography

Digital holography is a 3D imaging technique by emitting a laser beam with a plane wavefront to an object and measuring the intensity of the diffracted waveform, called holograms. The object’s 3D shape can be obtained by numerical analysis of the captured holograms and recovering the incurred phase. Recently, deep learning (DL) methods have been used for more accurate holographic processing. However, most supervised methods require large datasets to train the model, which is rarely available in most DH applications due to the scarcity of samples or privacy concerns. A few one-shot DL-based recovery methods exist with no reliance on large datasets of paired images. Still, most of these methods often neglect the underlying physics law that governs wave propagation. These methods offer a black-box operation, which is not explainable, generalizable, and transferrable to other samples and applications. In this work, we propose a new DL architecture based on generative adversarial networks that uses a discriminative network for realizing a semantic measure for reconstruction quality while using a generative network as a function approximator to model the inverse of hologram formation. We impose smoothness on the background part of the recovered image using a progressive masking module powered by simulated annealing to enhance the reconstruction quality. The proposed method is one of its kind that exhibits high transferability to similar samples, which facilitates its fast deployment in time-sensitive applications without the need for retraining the network. The results show a considerable improvement to competitor methods in reconstruction quality (about 5 dB PSNR gain) and robustness to noise (about 50% reduction in PSNR vs noise increase rate). An additional 3 dB gain is observed for activating the adaptive masking module. Moreover, our model is sufficiently robust against noise and tolerates AWGN noise up to σ = 10 . It shows only about 0.4 dB decay per unit noise variance increase, lower than similar methods. Our method elevates the DL-based digital holography to higher levels with a subtle computation increment. Furthermore, we explored transfer learning to enable fast utilization of the proposed method in time-constrained applications. Our experiments show that using a model trained for a similar sample can offer a reasonable reconstruction quality. Using transfer learning by borrowing network weights trained for a similar sample and performing additional 500 iterations for the new sample brings a considerable gain of about 12 dB compared to independent training with 500 iterations. This observation suggests that the developed model is highly transferrable between samples of the same type, but transferability across different sample types needs further investigation.