Based on the principle of computer-generated holograms reproduction imaging, this paper used the Gerchberg-Saxton (GS) algorithm to iteratively solve the phase distribution of the original simulation images under different characteristic parameters (line width, ring diameter) and different calculated sampling intervals by performing direct and inverse Fourier transforms on the optical field distributions of the input and output planes, and the corresponding reproduced images were obtained by simulation calculation. The optical path of the holographic reproduction experiment was constructed by using the liquid crystal spatial light modulator, and the reproduction experiment was carried out by loading the phase distribution maps of different original simulation images, the holographic reproduction images of far-field diffraction were taken by the camera, and the actual feature size of the reproduced images was obtained by image processing. The experimental results show that the feature size of the reproduced images is basically linear with the characteristic size of the original simulation images. Furthermore, the reproduction image size shows a non-linear change relationship with the sampling intervals of the simulation calculation, which is consistent with the derived theoretical calculation relationship curve. In order to further verify the correctness of the conclusion, when the size of the expected reproduced image is designed as the ring diameter of 0.943 mm and the line width of the central cross of 0.015 mm. The characteristic size and sampling interval of the original simulation image of the expected target are obtained by the simulation calculation as the line width of 3 pixel, the ring diameter of 594 pixel and the sampling interval of 25 μm, respectively. The ring diameter and line width of the holographic reproduction image, as measured by the reproduction experiment, are 0.93 mm and 0.017 mm, respectively. The error accuracy is within 0.02 mm. The findings of this study provide an effective reference for application scenarios such as holographic display and AR/VR display to improve the authenticity of virtual display image size.