High Quality Multi-path Multi-view Video Transmission considering Path Priority Control

Abstract

Multi-view video is one of promising techniques for future video services. In multi-view video delivery, transmitted video frames are highly vulnerable for losses during transmission. When a video frame in a certain camera is lost, the effect of the loss may propagate to the subsequent video frames in the same camera and video frames in the neighbor cameras. This is referred to as multi-view error propagation. To improve the loss resilience, some multi-path-based video transmission schemes have been proposed for multi-view video delivery. The existing transmission schemes assume that each transmission path has the same packet loss characteristic and thus do not account for the path diversity. However, each transmission path often has different loss characteristic, e.g., loss-free and loss-prone. To exploit the path diversity for high video quality, we propose two multi-view video transmission schemes. The first scheme utilizes features of multi-view video encoding. Specifically, reference, i.e., important, video frames are transmitted over lossfree transmission path while the other video frames are transmitted over loss-prone transmission paths. The second scheme utilizes decoding operation of multi-view video. In this case, a sender sends odd video frames over loss-free transmission path and exploits decoding operations for transmissions of even video frames. Evaluations show that our proposed scheme improves video quality (in terms of Peak Signal Noise Ratio) by 3 dB compared to existing transmission schemes when loss rate of transmission paths is 10%. In addition, our scheme achieves 1 MB lower video traffic in the same environment.