Dual holographic and polarization encoding for high fidelity image transmission through multimode fibers

Multimode fibers (MMFs) enable high-resolution imaging due to their capacity to support numerous spatial modes within a compact and minimally invasive form factor. However, inherent intermodal coupling, modal interference, and reflections from the fiber sidewalls generate complex speckle like distortions, severely restricting image transmission fidelity. Here, we demonstrate a dual-encoding approach that simultaneously employs holographic and polarization multiplexing to significantly enhance the information capacity of MMFs. Images encoded holographically into distinct spatial and polarization channels were transmitted through a step-index MMF and reconstructed from far-field speckle patterns using a modified convolutional neural network (DeepLeakyU-Net). The acquired speckle patterns exhibit clear clustering based on their encoded polarization and holographic labels, facilitating robust multiplexed reconstruction. High-quality image reconstruction was achieved with structural similarity indices averaging 0.93 across all polarization states. Our method leverages angular correlations in the far-field regime of MMFs, providing a mechanism to exploit the fiber’s multimodal potential, thus paving the way toward scalable multimode fiber-based imaging systems with enhanced multiplexing capabilities for applications in biomedical endoscopy and optical communications.