Improving perceptual quality for Internet video streaming

This paper investigates the efficacy of including redundant essential video data in methods for transforming burst packet losses to isolated. Critical video information is separated from non-critical video signals, assuming the availability of an efficient layered coder. Packets are interleaved and scrambled so as to spread out the loss of critical video information. In order to provide higher degree of key video signal protection, redundant packets of essential video data are included in the interleaved/scrambled packet sequence. Experiments were conducted on packet traces collected from the MBone. Results indicate that interleaving/scrambling significantly reduces burst losses. Furthermore, inclusion of essential redundant video data dramatically reduces the number of disjoint losses, as well as reducing the number of consecutive packet losses.