Scalable Video Multicast over Diverse Networks

Most video and audio streams distributed over computer networks today are generated at a single bit rate, and receivers with different types of computing equipment and network connections have had to cope with this. For example, the same video stream is usually transmitted at an intermediate bit rate to both a high-end system such as a Unix workstation with a Gigabit Ethernet network connection, and a low-end system, such as handheld PC, with a wireless modem connection which can only handle video at low resolution and in monochrome. This intermediate bit rate deprives the high-end system of possible better quality and causes the low-end system to choke and stall. Alternatively, multiple versions of the same video can be encoded and stored in the video server and appropriate versions which match the capabilities of different video clients and the speeds of their network connections can be sent to them. This defeats, to some extent, the original purpose of video compression, which is to economize on storage and transmission bandwidth.

We have developed a multi-wavelet-based scalable video coder and decoder, and the required networking functionality residing in the source, receiver and intermediate nodes to support scalable video distribution. These components enable different users to experience the best possible multimedia, limited only by the capabilities of their respective equipment and the bandwidths of their network connections. The system addresses the current widespread demand for access to multimedia content using heterogeneous types of devices (e.g. desktop computer, notebook computer, personal digital assistant (PDA) and mobile telephone) over diverse network connections (e.g. broadband, LAN, 3G, GPRS) which will enable the benefits of multimedia communications to be spread beyond the relatively few users with high-end equipment and fast network connections.

Figure 1: The SeeWAVE scalable video delivery system.

A complete scalable video delivery system called SeeWAVE (Figure 1) comprising video server, multicast client, gateway and unicast clients has been developed. Users can select video playback at different resolutions, in colour or monochrome, and both the required network bandwidth and amount of processing at the video client will be scaled accordingly. Versions of the video client which run on desktop, handheld and pocket PCs have been developed. A wavelet-based scalable video encoder has been implemented in an optimized manner on a Digital Signal Processor (DSP) platform to offload the processing load of the server PC. 

An adaptive multicast congestion control algorithm has been designed, studied and implemented. This algorithm enables a video client to provide the best possible video quality that can be supported by the available network bandwidth, i.e. the video client adaptively subscribes to the highest number of video layers in multicast channels that can be supported by the available bandwidth. A second video codec which permits layered video encoding with motion estimation based on the 3D-CSPIHT (Colour Set Partitioning In Hierarchical Trees) encoding algorithm has been developed. This codec permits a higher compression ratio than the wavelet-based video codec mentioned above.

Figure 2: Different video quality received by a Pocket PC PDA over a GSM data connection.

An example of streaming video at two different quality levels over a low bandwidth GSM channel to a personal digital assistant (PDA) device is shown in Figure 2.