Synthesis of Arbitrary View Images Using Depth Estimation Based on Iterative Comparison

Abstract

In this paper, the method of synthesizing arbitrary view images that uses 16 cameras on a plane is examined. First of all, we propose the technique for estimating accurate depth information between the cameras and objects. Next, the real arbitrary view images are synthesized by using the 3D shape model estimated by the depth information and the actual images appropriately. The images at arbitrary view points can be generated without errors due to occlusion.