Colorization of Automatic Hologram Generation from Photographs using Machine learning

Holography is the only imaging technology that can faithfully reproduce reflected light from objects. The hologram reproduced on a computer is called a computer-generated hologram (CGH), but the generation of CGH has previously been problematic due to the enormous amount of computation required.One research that has solved this problem is TensorHolography. This research will use techniques such as machine learning to estimate CGH at high speed. However, this research required specialized equipment such as a Tof camera and LiDAR to acquire depth information. This problem was solved by research called "From picture to 3D hologram". This research used machine learning to obtain depth information to create a monochromatic CGH. However, color CGH was not realized.In this research, we proposed a method to automatically generate color holograms from photographs by using machine learning to obtain depth information from photographs using a monocular depth estimation model and then to use machine learning to create color CGHs. And, the effectiveness of the proposed method was confirmed by playing back the CGH actually generated.