Computer-generated hologram (CGH) is a technique that enables recording optical systems for holography by computer simulation, and can reconstruct ideal 3D images. Since CGH requires the model data of objects, it is necessary for creating 3D images of real existing objects to model them precisely. However, it is often difficult to obtain this data. To solve this problem, there has been much research into multi-view projection image based methods that do not require 3D model data. CGH using integral photography and CGH using multi-view projection obtained by lens array are examples of this research. However, these methods have some problems. First, their optical recording systems are complex and expensive, and therefore, realizing these systems is difficult. Another problem is that enormous projections are required to generate a hologram with sufficient pixel size. In the past, a method to generate holograms from multi-view images taken with digital cameras was proposed. Additionally, to generate CGHs using multi-view images from randomly arranged cameras, we proposed a method that uses distance transformations, rotations of light waves, and an algorithm that sets up hologram parameters automatically by using camera arrangement information. In this paper, the results of numerical reconstructions are compared, and both the motion parallax and resolution of the reconstructed images are discussed. The proper number of multi-view projections needed will be derived from this discussion.© (2011) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.