Neural fusion and rendering of multidimensional light field for holographic stereogram

The synthesis of holographic stereograms through spatiotemporal light field fusion enables immersive 3D visualization, yet faces critical challenges in cross-dimensional data rendering and encoding, including computational redundancy, inefficiency, and spatiotemporal inconsistency. To address these limitations, we propose a direct holographic primitive encoding method based on neural radiance fields. Compressible-composable neural radiance field fuses sparse spatial light fields into Synthesized Effective Perspective Images (SEPIs) through direct holographic primitive encoding. Dynamic neural radiance field achieves precise temporal light field extraction and encoding via discrete sequence quantization. This unified rendering pipeline integrates spatial-temporal radiance decomposition with a self-supervised consistency, and is implemented through the holographic exposure via Effective Perspective Images’ Segmentation and Mosaicking (EPISM) method. Experimental validation demonstrates that the direct holographic primitive encoding method enables the rapid generation of high-quality exposure images. The synthesized holograms exhibit significant advantages in terms of 3D reconstruction clarity, detail preservation, and visual immersion. This research establishes a new paradigm for complex-scene holographic imaging and provides an efficient computational framework for augmented reality and 3D display systems.