Abstract
We have been investigating synthesizing arbitrary view images. For correct synthesis, it is necessary to use exact depth information. In this paper, we propose a method for depth estimation by comparing synthesized images to an actual image and for generation of more realistic arbitrary view images using the depth information.