# Spatial incoherence-driven optical reconstruction of holograms with observer shift-invariance

## TL;DR

This paper introduces a new holographic method using incoherent light to create high-quality 3D images with a much larger viewing area.

## Contribution

The novel approach uses incoherent light to enable shift-invariant optical reconstruction and a significantly expanded eyebox in holograms.

## Key findings

- Incoherent light can reconstruct 3D scenes with high image quality and a 1000-fold larger eyebox.
- The reconstructed intensity remains shift-invariant under pupil displacement.
- A neural network enables real-time synthesis of incoherent holograms with reduced computational costs.

## Abstract

Coherence preserves phase consistency between wavefields, enabling accurate recording and reconstruction in holography. Although recent advances in computational optics have realized holographic data acquisition using incoherent light by computationally retrieving information, optical reconstruction still requires partially coherent light sources. We demonstrate a hologram that reconstructs 3-dimensional distribution utilizing incoherence. By decomposing incoherent light into infinitesimal coherent lights and calculating their propagations, the incoherent sum is optimized to resemble the desired 3-dimensional scene, whereas individual coherent lights reconstruct completely different intensities. Incoherence provides high image quality and a wide eyebox, with the reconstructed intensity remaining shift-invariant under pupil displacement, allowing a 1000-fold expansion of the eyebox. We confirm the shift-invariance through a proof-of-concept experiment and demonstrate real-time synthesis of incoherent holograms using a neural network, significantly reducing computational costs. Our method could inspire new approaches in photonics using incoherent light and be practically adopted in holographic displays.

The expanded eyebox of holographic displays has been realized using incoherent light by optimizing the incoherent sum of infinitesimal coherent components, enabling high-quality, shift-invariant reconstruction and real-time synthesis via neural networks.