The invention discloses a high-resolution single-photon imaging method and system combined with a physical noise model. The method comprises the steps of building the physical noise model of a single-photon detection array, calibrating the parameters of the single-photon detection array, and simulating the photon detection efficiency of the single-photon detection array, the Poisson process of photons, the crosstalk of the detection array, the post-pulse and other noise factors, downsampling the high-resolution picture into a plurality of low-resolution pictures and adding the low-resolution pictures into the noise model to generate a low-resolution noisy data set, constructing a deep learning neural network, and training the network by taking the low-resolution noisy picture as an input and the high-resolution picture as a true value to obtain optimized network model parameters, and finally, recovering a high-resolution noise-free picture from the low-resolution noisy picture by usingthe trained network. The method is combined with an actual physical model, the measurement noise of the single-photon detection array can be effectively reduced, and the imaging resolution of the single-photon detection array can be effectively improved.