Advanced depth-layering using a synthetic-depth map and deep neural network for real-world object-based holographic displays

We propose an advanced layering method for real-world object-based holographic displays, leveraging what we believe to be a novel synthetic-depth map and deep neural network. The proposed system aims to implement a holographic display that displays natural-like three-dimensional visualizations of real objects by enhancing data quantity and ensuring accurate depth layers. A simplified light-field image acquisition system combined with a deep neural network is employed to efficiently gather organized omnidirectional three-dimensional information from the object, achieving high quality while minimizing processing time. Subsequently, a novel high-accuracy synthetic-depth map containing data from both initial depth and position maps is estimated. Finally, the sub-holograms for each depth layer are generated and integrated as a single main hologram by encompassing comprehensive object information, which is displayed on the spatial light modulator of a holographic display system and illuminated by a coherent light source. Experimental results confirm the superiority of the proposed system, particularly demonstrating its effectiveness for objects with a wide depth range or multiple objects separated by considerable distances.