Virtual View Synthesis Using Moving Multi-Camera Array

Considerable improvements of information technologies (IT) allow consumers to enjoy various forms of multimedia services. Recently, 3D video has been established as a new format that provides more realistic and natural experiences to users by free-viewpoint TV (FTV). One of the main challenges is about rendering continuous viewpoint images using color and depth information. Thus, image interpolation, which generates virtual view-point images, is a key part in 3D display systems. In this paper, we propose a virtual view synthesis method for a moving multi-camera array using stereo images. Our method synthesizes virtual viewpoint images using both spatial and temporal information. The proposed method consists of two parts: depth estimation and virtual viewpoint image generation. In order to evaluate the view synthesis method, the peak signal-to-noise ratio (PSNR) of depth images are estimated. Our results show that the PSNR results from the proposed method outperforms the conventional method. Keywords— FTV; view synthesis; depth map upsampling;