Partial Transcaling for Wireless Packet Video

Scalable coding provides many advantages for video transmission over packet networks that lack Quality-of-Service (QoS) guarantees. Meanwhile, it is well known that the benefits of scalable coding come at the expense of degradation in video quality or reduction in coding efficiency when compared with non-scalable video compressed at a particular bitrate. Recently, the notion of TranScaling (TS) has been proposed [1] to improve the coding efficiency of scalable video coding. Under TS, which is a generalization of non-scalable transcoding, a scalable video stream, that covers a given bandwidth range, is mapped into one or more scalable video streams covering different bandwidth ranges. TS exploits the fact that the level of heterogeneity changes at different points of the video distribution tree over best-effort packet networks (e.g., the wireless Internet). This provides the opportunity to improve the video quality by performing the appropriate transcaling process. In this paper, we propose a new extension to the TranScaling framework by exploring Partial TranScaling, which provides a “Transcalar” the option of choosing a low complexity transcaling process. In particular, we propose partial-TS based approach for the MPEG-4 FGS Temporal (FGST) scalable video coding method. The proposed TS method for FGST significantly reduces the computational complexity of the original (full) TS scheme while achieving higher than 1 dB in PSNR improvements. We present some simulation results based on employing partial TS to a wide range of MPEG-4 FGST streams. In these experiments, we show 1.5 dB in PSNR improvements when compared with a non-transcaled MPEG-4 FGST video.