This paper presents a novel approach to generate computer-generated holograms (CGHs) with orthogonal scanning multi-view projections. An unified object three-dimensional Fourier spectrum sampling model is established, and based on its error analysis, the spectrum sampling circle of a circular scanning projection image at a specific angle θ can be simulated from a traditional orthogonal scanning projection image; then, the principle of "smaller θ (view angle of scanning) wins all" is adopted to determine the spectrum sampling information for overlapped positions so as to improve the spectrum utilization efficiency of orthogonal scanning projection image. Starting from the relationship between the ideal sampling point and its integration points in the consistency of spectrum information and the degree of spatial proximity, we set the weights of frequency-domain grid points where the ideal sampling points are located. Based on these weights, grid points with high consistency of spectrum information are selected adaptively as the actual sampling points, thereby achieving a balance between the sufficient sampling of projection spectrum and the introduction of background noise to obtain the optimal object three-dimensional Fourier spectrum. Experimental results of CGH with the orthogonal scanning multi-view projection for a virtual three-dimensional model show that the visual quality of the reconstructed hologram by the proposed method is notably improved. The signal-to-noise ratio of the reconstructed hologram is significantly better than that of traditional orthogonal and circular scanning methods under the same number of projected images; in addition, the signal-to-noise ratio of the reconstructed hologram implemented with a redundant sampling operation by the proposed method is also better than that of the traditional orthogonal method even when the number of projected images is reduced by one time. Therefore, the proposed method has great values both on theory and practice.