Evaluation of Rate - Control Strategies for Real - Time Video Communication

Abstract

In real time communication such as interactive video conference we have limits on the maximum delay allowed. Since the signal to be transmitted can be lossy coded, we have a trade off between the bit rate used in the coder and the quality of the representation of the signal. If the communication takes place over a channel with a time varying bit rate, we can compensate for the varying rate of the channel by using buffering. We can not buffer too much data since then the delay will be too large. This master thesis project will consist in simulating some strategies for controlling the rate of the coder when using a limited buffer. Between the bit rate and the corresponding distortion a simple analytical model can be used. This model should take into account that when coding video we have two different kinds of frames. One kind is the I-frame when the entire picture is coded as a still image. The other kind is the P-frame where the frame to be coded first is predicted from previous frames. P-frames will require less bit rate for the same distortion compared to I-frames. The P-frames could be inserted periodically or occur at different time instants that are chosen from some random distribution. The objective is to evaluate some different strategies in terms of what average distortion they will give for different buffer sizes and what the corresponding delays are.