Video Traffic Prediction based on the Source Information and Its Application to RCBR

摘要

We propose a source traffic prediction method which is based on the rate-distortion relation of video. With this prediction, we address the transmission rate decision problem for real-time VBR video. The transmission rate bound is derived using the underflow and the overflow constraint of the encoder and the decoder buffers. Based on the predicted bit-rate of future video frames, transmission rate is calculated with the MVS(Minimum Variance Smoothing) and some delay constraining algorithms. From simulation results, the proposed traffic prediction method is shown efficient in detecting scene changes and estimating changed traffic levels. Also, the proposed transmission control method is shown to have low renegotiation cost and high channel utilization without delay violating frames.