Adaptive joint H.263-channel coding for memoryless binary channels

Our main purpose is to show that when video source coding is dependent on the channel characteristics, better reconstructed video quality can be achieved. Our study considers a fixed overall bit rate transmission of 64 kbit/s over error prone memoryless binary channels and the source coder is the ITU-T H.263 video standard. The variable H.263 source coding rate adapts to a particular channel bit error rate (BER) at any instant of time. We first investigate the error sensitivities of H.263 syntactical elements and group them into several classes of different significance. By identifying the most sensitive elements, we develop a data grouping (DG) technique which, by itself, exhibits improved error resilience. We then design an unequal error protection (UEP) scheme and compare its performance with an equivalent (same source coding bit rate) equal error protection (EEP) scheme, using rate compatible punctured convolutional (RCPC) codes of different rates for forward error protection. Both forward error correction schemes adopted have resulted in PSNR improvements of over 20 dB for BERs higher than of 4/spl times/10/sup -3/. Finally, if during the real time encoding process, the encoder is aware of the channel BER and there is an unlimited number of code rates, the paper shows that UEP and EEP strategies provide quite similar performances.