Mosaicing videos to stream over multiple independent channels

Streaming high fidelity multimedia objects requires large amounts of network bandwidth resources. Sometimes these resources are achieved by aggregating a number of independent and lower capacity network channels. Network level aggregation schemes can stream the single video across all the network links. However, splitting multi-layer encoded video streams are not resilient to network failures on individual links because enhancement layers are not independent and depend on the availability of base layers. In this paper, we investigate several multiple description coding mechanisms that split the stream into multiple independent sub-streams. Our mechanisms attempt to retain the spatial and temporal redundancy inherent in the original stream in order to achieve good compression efficiency. We examine the impact of our approach on changes in peak transmission requirements, overall transmission size and stream quality. We show that the sub-streams are able to sustain substantial data loss while still providing a viewable stream. We also show the object size overhead for the various mechanisms.