Incoherent superresolution via diffraction-based Hermite–Gaussian imaging

The resolution of optical imaging is limited by diffraction. Hermite–Gaussian imaging (HGI) surpasses the diffraction limit via a quantum-inspired approach called spatial-mode demultiplexing. However, prior heterodyne-based implementations rely on a phase-locked local oscillator to project the spatial modes and are therefore limited to temporally coherent imaging. We present an implementation of diffraction-based Hermite–Gaussian imaging (D-HGI) that allows for HGI of temporally incoherent sources. D-HGI performs local-oscillator-free mode projection using a single phase-only element and estimates mode intensities by integrating the intensities in the ±1st diffraction orders, enabled by a point-spread-function-aware mask design. Experiments on point source localization verify the performance of the mode projection method. For HGI of full two-dimensional objects, we replace prior neural networks with a calibration-driven reconstructor, improving robustness and interpretability. Experimentally, D-HGI achieves an optical resolution of 0.31 λ /NA in an incoherent imaging system, demonstrating passive superresolution beyond the Rayleigh limit. The proposed approach paves a practical route to passive incoherent superresolution in microscopy and astronomy.