In this paper we presented a full-parallel Hopfield neural network based on a new model of the three-dimensional (3D) Gaussian point spread function (PSF) to restore the microscopic optical slices. As the result of the diffraction of the microscope's objective, the imaging model in three-dimensional optical- sectioning microscopy incorporates spherical aberration that worsens with increasing depth under the coverslip and changes in the PSF. Two- dimensional image restoration and three-dimensional serial images restoration are used to analyze the capability of the network, and the performance shows that the continuous Hopfield network can restore the blurred images of the depth-variant microscopic image.