Scalable MPEG-4 video transmission. using video quality metrics

Video streaming applications place high bandwidth requirements on the network, even for efficiently compressed MPEG-4 and H.264 videos. There is a need for these applications to reduce the use of bandwidth, especially when operated on bandwidth-constrained access networks involving cable, DSL, and wireless. However, a direct reduction in allocated bandwidth to a video may have an adverse effect on its visual quality. In addition, the high bit rate variability associated with MPEG-4 compressed video makes it difficult to determine the appropriate level of bandwidth reduction for such videos. MPEG-4 encodes videos as a sequence of I, P, and B-frames, with B-frames containing the least amount of information. Furthermore, the encoded sequence contains forward and backward references for the predictive coding of P and B-frames, which can be used to predict these frames when there is a frame loss. We exploit this feature to selectively drop B-frames to reduce the bandwidth requirements and bit rate variability. We also employ the frame-size (content-independent metric) and the relative PSNR (content-dependent metric) to achieve scalable video transmission. It will be shown via an actual network setup that substantial bandwidth savings can be achieved using the proposed methods and. in the majority of the cases, this can be done without sacrificing video quality during playback.