An Efficient Packet Loss Recovery Methodology for Video Streaming Over IP Networks

Abstract

Large amounts of data and limited bandwidth are always at odds for digital video streaming over internet protocol (IP) networks. Packets could possibly be delayed or lost. Sender-based recovery techniques for packet loss generate redundant information that causes lower compression ratio and consumes additional bandwidth. In order to address this issue, a novel information hiding based recovery methodology has been developed. The basic idea is that the redundant information used for error recovery is embedded within the frames of the original video contents, by means of high bitrate information hiding techniques. In the receiver end, the hidden information is extracted for lost-packet recovery. Experimental results show that the damaged macroblocks can be recovered with a higher quality using the proposed methodology, as compared to spatial extrapolation used in H.264/AVC reference software. The main contributions of the proposed methodology are: (1) the delivery of redundant information does not increase bandwidth requirement significantly under the same media quality, bringing out advantages in practical applications; (2) the structures of encoder and decoder do not need to be changed in the case of intra-only coding mode.