Large field-of-view lensless holographic dynamic projection system with uniform illumination and U-net acceleration

Abstract

Compared to traditional projection systems, a diffraction-based lensless projection system offers the advantages of a compact size with high design freedom. However, the existing iterative algorithms for computer-generated holograms (CGHs) in large field-of-view diffraction systems are not mature, and are characterised by long processing times, making them unsuitable for application in dynamic displays. Furthermore, most large-scale diffraction algorithms neglect the intensity and phase nonuniformity of the illuminated spherical beam. Herein, we develop a large field-of-view laser projection system with U-net acceleration. Our system has a diffractive optical element (DOE) inserted between the digital micromirror device (DMD) and the screen to reduce the intensity and phase nonuniformity introduced by the spherical wave. Structural parameters of the system are also optimized, then amplitude CGHs are loaded on the DMD sequentially. A divergent laser beam illuminates the DMD and is converted to a uniform beam by the DOE before it reaches the screen. The Gerchberg–Saxton (GS) iterative algorithm based on the scaled angular spectrum method (SASM–GS) is used to generate the amplitude CGH. However, as this process of computation includes three forward and three backward diffractions, it is time-consuming. Therefore, CGHs generated by the SASM–GS algorithm were used as labels, and a U-net model was trained using the MNIST database. Simulations revealed that the trained U-net model can accurately generate amplitude CGHs and has a certain degree of versatility. Notably, the hologram generation time of the U-net model was one-tenth that of the SASM–GS algorithm. Finally, we performed experiments to verify the efficacy of the U-net model, demonstrating the feasibility of the proposed dynamic lensless projection system.