Pitch-flexible implicit neural hologram representation

We propose a method for computer-generated holography (CGH) that enables flexible pixel-pitch control using implicit neural representations (INRs). Conventional hologram synthesis methods require recalculating individual holograms for each fixed pixel pitch. In contrast, the proposed INR-based hologram—modeled as a continuous function mapping spatial coordinates to phase—allows pitch-flexible hologram generation with a single training procedure. We validate the proposed method through numerical and optical experiments and demonstrate clear advantages over representative conventional approaches. Additionally, we confirm that our model can be naturally extended to three-dimensional representation, whereby a single trained model can synthesize high-fidelity holograms across a wide range of pixel pitches and reproduction distances. This study realizes a continuous hologram representation, offering a promising direction for computationally efficient hologram synthesis and compact hologram storage.