Skip to Main Content
Video multicast over Wireless Wide Area Networks (WWAN) is difficult because of unavoidable packet losses and impracticality of retransmission on a per packet, per client basis, due to the known NAK implosion problem. Recent approach exploits clients' cooperation for packet recovery, so that a peer group's received WWAN packets are shared using a secondary network like Wireless Local Area Network (WLAN). For multiview video multicast, where a client can switch views interactively by subscribing to different WWAN multicast channels streaming different views, two new difficulties arise. First, system must provide timely view-switching mechanism, so that client can switch to a desired view quickly for correct decoding and display. Second, it is difficult for system to leverage neighboring peers for cooperative loss recovery, since neighbors are more likely to be subscribing to different views than a loss-stricken peer. In this paper, we use Distributed Source Coding (DSC), a new compression tool in video coding, to solve both problems. Each DSC frame is encoded with a set of predictor frames, and correct decoding only requires one of the predictors in the set to be available at decoder. Periodic insertion of DSC frames into video streams then enables a peer to switch from view v to v' at the DSC frame boundary, assuming DSC frame of view v' was encoded using a frame in view v as one predictor. For the same assumption, a neighbor watching view v can help a peer watching view v' evade error propagation resulting from earlier losses and resume decoding at the DSC boundary. Experiments show that optimized usage of DSC frames in a coding structure, where unequal error protection is enabled to decrease the probability of decoding failure earlier in a group of pictures, outperforms a structure using I-frames instead for view switching by up to 11 dB in video quality in typical WWAN network loss environment.