The multi-modes Bessel-Gaussian OAM hologram encoding based on convolutional neural networks
Multi-mode vortex light is a superposition of different orbital angular momentum (OAM) lights. However, as the number of OAM modes increases, the sampling constant changes. Using the traditional sparsely sampling will lead to severe loss of detail, reduced image resolution. To achieve high capacity and resolution of the OAM hologram, this paper prepares a multi-mode Bessel-Gaussian (MBG) selected hologram by stacking different mode combinations of BG phases on a MBG saved hologram in stages. Using a MBG beam with opposite combination parameters to illuminate the MBG OAM hologram, the target image can be reconstructed after the Fourier transform, and the sampling constant is flexible and controllable. The holograms encode MBG mode combination parameters. The additional degree of freedom provided by combining with MBG OAM beam offers more multiplexing channels and a higher security hologram. To further improve the quality of holograms, we first save the holograms and the corresponding MBG mode combination parameters when the quality of the obtained hologram is high based on Actor-Critic neural networks. Secondly, we gradually adjust the MBG mode combination parameters. Finally, we confirm the reasonable range of the MBG mode combination parameters.