Advanced EPISM approach for holographic stereogram generation utilizing neural radiance fields

This paper presents a synthetic holographic stereogram printing approach that integrates neural radiance fields (NeRF) with the effective perspective images segmentation and mosaicking (EPISM) method. Sparse perspectives of a 3D scene are captured through random sampling and used to train a NeRF model with multi-resolution hash encoding, enabling rapid construction of an implicit scene representation. The EPISM method calculates the camera pose parameters needed for parallax images, which are rendered through the trained neural network. These rendered images are then encoded using the EPISM process to generate a sequence of synthetic effective perspective images for direct exposure printing. Experimental results demonstrate that the integration of EPISM and NeRF effectively addresses challenges such as inefficient light field acquisition, low computational efficiency, and the limitations of traditional techniques. This approach ensures rapid computation, precise scene reconstruction, and high-quality hologram printing, offering an end-to-end solution for acquiring and printing high-fidelity synthetic holographic stereograms of real-world scenes.