Optical superresolution assisted by multi-mode fiber and neural network

We demonstrate a novel approach for surpassing the diffraction limit in passive optical imaging using a standard step-index multi-mode fiber (MMF) combined with a simple neural network. Unlike previous techniques based on spatial mode demultiplexing and multi-plane light converters, our method relies on the complex speckle pattern generated by mode interference in the MMF. This speckle pattern is highly sensitive to small changes in the input field and is analyzed using a perceptron-type neural network trained to extract parameters such as the separation and intensity ratio of two incoherent point sources. Our experimental results show that the system can resolve beam separations well beyond the classical diffraction limit. The method is flexible and cost-effective, enabling high-resolution and multi-parameter measurements using standard optical components. This work opens new possibilities for passive super-resolution imaging in diverse applications where structured illumination or active modulation is not feasible.