Streaming media applications are becoming increasingly popular on the Internet. For streaming applications the media data must be encoded without the benefit of knowing the state of the channel during transmission. For this reason, in streaming media systems, the encoding must be flexible and the server must adaptively select and transmit the appropriate data to the client, as a function of the state of the network. Since many media streaming applications use pre-roll buffer, there is large flexibility in packet scheduling algorithms, significantly affecting rate-distortion performance [1,2].
In our study, we analyze the expected global distortion in video streaming assuming error concealment at the receiver. We also propose a simple and real-time applicable rate-distortion optimization method based on the effective bandwidth approach. We evaluate the effective bandwidth of VBR video data by computing the autocovariance of data size in a group of picture (GOP), taking into account the variance of channel bandwidth fluctuation. Assigning a different priority to motion vector and texture, we obtain SNR scalability effectively . We also adopt temporal scalability, resulting in graceful degradation in channels with heavy loss. Experimental results show that our scheduling algorithm, motion-texture discrimination, and temporal scalability yield a performance improvement of several dBs in PSNR compared to conventional sequential transmission of video data without scalability.