Deep-learning-assisted Hologram Calculation via Low-Sampling Holograms

Digital holograms can be calculated by simulating light wave propagation on a computer. Hologram calculations are used for three-dimensional displays. However, the calculations take a long time, and the data size of the calculated holograms becomes large. This study presents a deep-learning-assisted hologram calculation using low-sampling holograms. We calculate holograms with low-sampling rates, resulting in the acceleration of the hologram calculation and the decrease of the hologram size. However, the low-sampling holograms decrease the quality of the reconstructed images and will occur the aliasing errors when not satisfying the Nyquist rate. The proposed method uses a deep neural network to retrieve the full-sampling holograms from the low-sampling holograms. We show elementary results of the proposed method in numerical simulation.