Orbital Angular Momentum Holography Using Neural Network and Camera in the Loop

Orbital Angular Momentum (OAM), a spatial mode of light, is employed as an information carrier for encryption and multiplexing in optical communication and holography. Conventional methods for designing OAM-multiplexed holograms exhibit suboptimal reconstruction accuracy as the number of multiplexed channels exceeds the nominal multiplexing capacity. Reconstruction quality further degrades in practical systems due to optical aberrations and imperfections in holographic displays. In this work, two methods are proposed, a Gradient Descent (GD) optimization and a deep learning approach, for designing OAM-multiplexed phase-only holograms. Both methods are integrated with a Camera-In-The-Loop (CITL) calibration technique that learns a realistic parameterized propagation model to compensate for system imperfections. The experimental results show that when operating at twice the nominal OAM multiplexing capacity, the proposed GD and neural network methods combined with CITL calibration, reduce cross-correlation errors in reconstructed images by up to 82% and 58% compared to the conventional method, respectively. These methods enable accurate, high capacity, and real-time OAM-multiplexed holography in practical optical systems.