We consider a problem of eye-fundus image reconstruction from different-depth fundus sections based on an image stack, with the images obtained via imaging system's fast refocusing as superposition of 3D object sections and blurred images of adjacent-depth sections. An implicit iterative regularization method in the Fourier plane is used for 3D deconvolution. The results of mathematical modeling have demonstrated that the numerical sectioning shows promise when processing ophthalmological images distorted by various types of noise.