Modeling of layered video for low-bit-rate video conferencing applications

Heterogeneous wireless networks present several challenges for real-time video applications. Traditional video codecs are not particularly robust and perform poorly when faced with packet losses from congestion and bit errors within the channel. A combination of deterministic traffic smoothing and layered video coding techniques improves error resilience. Smoothing over single frame intervals improves statistical multiplexing gains. To mitigate a portion of the delay introduced by smoothing, we propose a predictive smoother using the RLS algorithm that reduces the delay due to encoding. Layered video coding transmits video in scalable layers of increasing quality. We present a simple hybrid wavelet/DCT layered coder along with an associated optimized layered video traffic model. The traffic model employs a simplified multistate Markov chain representing a nonuniformly quantized pseudo-histogram of observed bit-rates for each layer. The model captures the effect of deterministic smoothing and aids in evaluating scheduling algorithms as well as providing a simple call acceptance scheme.