Temporally dense holographic video generation using reconstruction‐guided complex hologram optimization

Abstract Holographic video enables realistic 3D visualization by encoding the amplitude and phase information of light. However, generating high‐framerate holographic video is computationally demanding, and conventional frame interpolation techniques are not directly applicable owing to the complex nature of holographic wave fronts. This paper proposes a novel interpolation framework combining deep‐learning‐based 2D frame prediction with sequential phase and amplitude optimization via stochastic gradient descent. Intermediate reconstructions are predicted using a pre‐trained interpolation model and then used as targets to refine complex holograms. The proposed method is evaluated on single‐ and multi‐depth holographic video sequences, including cases with up to five‐depth layers. The results demonstrate that the proposed method significantly improves temporal resolution (from 3.2 to 24.2 frames per second) while preserving image fidelity and depth consistency. The interpolated holograms maintain structural sharpness and exhibit no noticeable artifacts, even under speckle noise conditions. These results confirm the proposed method's scalability and effectiveness for high‐framerate holographic video generation.