| The classical results in multiterminal source encoding, namely Slepian-Wolf theory, side-information theory and Wyner-Ziv theory, all are based on the following configuration. Each of two sources is connected to a dedicated encoder which does not obtain any information about the other source, while the decoder receives the output of both encoders.; We treat two individual problems in multiterminal source encoding, the second being an extension of the first. The first is based on the same classical multiterminal configuration described above. Imposing less stringent rate-distortion criteria than those of the three classical results, we obtain a single-letter characterization of the admissible coding rate region. The specialization of our result to any of the classical results is relatively simple. Thus our result subsumes and unifies all the classical results in multiterminal source encoding.; The second problem treats the situation in which a particular encoder in the classical configuration breaks down. Then the decoder must estimate the sources based solely on the output of the other encoder. A single-letter characterization of the admissible coding rate region with certain distortion criteria is obtained. This result subsumes most previously reported results on multiterminal source encoding. We also specialize the general result to the case in which the decoder must reproduce both sources perfectly when both encoders are functioning. Surprisingly, we are able to show that with the same coding rates as in the Slepian-Wolf operation, it is in general possible for the decoder not only to reconstruct both sources perfectly under normal operation but also to make a non-trivial estimate on the sources in case a particular encoder breaks down. We also study the situation in which one encoder is operating at the minimum rate, with the requirement that the decoder must reconstruct both sources perfectly when both encoders are functioning. We find that under rather general conditions, if the other encoder breaks down, it is still possible for the decoder to make a non-trivial estimate of the sources based on the output of the remaining encoder. |
