Implicit feature compression for efficient cloud–edge holographic display

Abstract

Holographic displays, with their ability to vividly reconstruct object wavefronts, stand as promising candidates for future immersive display technologies. However, delivering such immersive experiences demands large volumes of holographic data. Compressing holographic data with high compression ratios remains challenging due to the substantial high-frequency content in holograms. To overcome this challenge, we propose an implicit feature compression-based cloud–edge system for efficient holographic display. The distinctive aspect of our approach lies in compressing the implicit features learned during hologram generation into an encoded stream, rather than compressing the hologram itself. This methodology integrates a joint design of a cloud-side encoder and edge-side decoder, with both components performing mixed hologram generation and data compression/decompression. Our results on 1,000 augmented DIV2K test images demonstrate that our approach remarkably reduces the original data volume by 99.8% on average, and the experiments validate our approach. This research establishes a technological foundation for the large-scale commercialization of holographic displays.