A Bandwidth-Efficient Stereo Video Streaming System

Abstract

Stereo video streaming is an emerging topic ever since the popularity of stereo video in the recent years. In this paper, we consider the anaglyph video, which displays the red channel in left view and green/blue channel in right view, respectively. The traditional anaglyph streaming system transmits full RGB components of both views and those results in greater bandwidth consumption. In this paper, we present two novel algorithms to improve the traditional anaglyph system. For the first, we transmit the color channels that will be viewed by the users, and then we rearrange the data to adapt to the standard video compression format. For the second, we propose to perform video demosaicing in the post-processing stage instead of the pre-processing stage as in the traditional system. In addition, we develop a gradient-based stereo matching technique integrated with our demosaicing algorithm for better retrieval of observed data. Experimental results demonstrate that the proposed anaglyph video streaming system outperforms the state-of-the-art algorithms in both objective and subjective comparisons.