Effectiveness of an image generation method that expresses the visual impression of space for two-point perspective landscapes

Abstract

Photographs and CG images are generally rendered using a perspective projection. However, because a perspective projection image does not always represent the visual impression of a real space, we occasionally feel a sense of incompatibility when viewing such images. We believe that clarifying the human visual characteristics will enable the creation of more realistic images. A previous study investigated the human perception of size in a real space and proposed a magnification-rate function that shows the relationship between the subjective visual size of an object to be drawn and the observation distance. Images applying the magnification-rate function provide an impression closer to that in a real space than do perspective projection images. However, these results have only been verified for single-point landscapes. In this study, to investigate the impression of an image, we applied the magnification-rate function to images portraying landscapes with a two-point perspective. The results show that the magnification-transformed images were evaluated as being closer to the impressions in a real space than those of perspective projected images. This is similar to the evaluation of transformed images for one-point perspective landscapes, suggesting that an image transformation using the magnification-rate function is effective for two-point perspective landscapes.