Conditional neural holography: a distance-adaptive CGH generator

Abstract

A convolutional neural network (CNN) is useful for overcoming the trade-off between generation speed and accuracy in the process of synthesizing computer-generated holograms (CGHs). However, methods using a CNN have limited applicability as they cannot specify the propagation distance when synthesizing a hologram. We developed a distance-adaptive CGH generator that can generate CGHs by specifying the target image and propagation distance, which comprises a zone plate encoder stage and an augmented HoloNet stage. Our model is comparable to that of prior CNN methods, with a fixed distance, in terms of performance and achieves the generation accuracy and speed necessary for practical use.