RDOES: A rate-distortion optimized encoding and streaming framework for live video transmission

摘要

In this paper, we jointly consider the work of rate distortion optimization (RDO) coding and rate distortion optimization (RaDiO) streaming. A statistical mean-squared error (MSE) prediction system considering the changing macroblock refreshing rate and the error propagation effect is introduced to make the expected distortion expression used in RaDiO more accurate. According to this real time distortion prediction method, we propose an integrated rate distortion optimized encoding and streaming (RDOES) framework which jointly optimize the coding and streaming process by combining control on the selection of macroblock refreshing rate, the scheduler on transmission and retransmission of the video packet. Using this framework, we demonstrate that each frame should be encoded as I frame or none-intra-macroblock-refreshing P frame to achieve the best transmission performance whenever RaDiO or RDOES is used. Simulation results show that our RDOES framework provides flexibility and demonstrates substantial performance gains of at least 1.1dB on average over conventional RaDiO.