UEP of progressive images in wireless channels

We present a general optimization method of rate allocation for embedded image sources over fading channels under unequal error protection (UEP). 2-state Markov chain or Gilbert-Elliott (G-E) channel models are exploited to approximate slowly fading channels in wireless communications in which each state corresponds to a binary symmetric channel (BSC). Under the assumption that exact knowledge of channel state information (CSI) is not available, the partition of the bit stream, the allocation of source/channel coding rate, and the number of protection levels for UEP schemes are all optimized by using the statistics of the G-E models and maximization of the expected error-free source bit stream length. Experiment results lead to different rate allocation strategies from previous ones in some channel conditions and the performance improvement is up to 0.4 dB.