Transmitting low-latency video over best-effort networks where packets can be lost (that is, the Internet) and using protocols that do not provide reliability for the sake of low latency (that is, the user datagram protocol (UDP)) is still a topic of research despite decades of endeavor. Every technique that addresses one aspect of low-latency transmission comes at a cost. Added resiliency typically adds a bit rate overhead, while the retransmission of lost packets adds latency and can cause congestion. To add to the complexity of the problem, there is the fact of heterogeneous receivers of video, each with its own, often varying, capabilities and network conditions. Hence the need for scalable coding where a video is encoded in layers in such a way that allows each receiver to receive videos of varying quality based on the layers it receives. An ideal video transmission encompasses resiliency to packet loss, low bit rate overhead, and low latency while providing scalability.
While signal-to-noise ratio (SNR) scalable video coding (SVC) provides scalability of coded video by providing a base layer and an enhancement layer, multiple description coding (MDC) provides resiliency in case of packet loss by providing multiple layers, each independently decodable. The combination of these two (SNR-SVC plus MDC) is the focus of this paper, as well as providing a good overview of the state of the art in MDC techniques. More specifically, a video coded stream is split into two (or multiples of two) descriptions, and then each goes through SNR-SVC such that the base layer is repeated in both descriptions for added redundancy at the cost of a bit rate overhead of about 30 percent. Additionally, care has been taken to intelligently adapt the base and enhancement layers to the transmitted value.
Anyone in the research field of video coding, and more specifically the MDC technique, will be able to understand the concept presented; the presentation and organization is clear and concise. The performance evaluation is also sound, providing the advantages, disadvantages, and limitations of the technique presented. For those outside of video coding and unfamiliar with MDC or SVC, it will be tough to grasp the main idea. To the practitioner of video streaming over best-effort networks, the paper shows the possibility of combining two separate techniques for added resiliency and scalability of video.