Remote coding of correlated sources with high resolution

The authors first discuss direct coding with high resolution. They then consider remote coding with high resolution. They restrict their attention to the Gaussian case, but consider the following more general, indirect coding problem. We need to reconstruct a (memoryless) zero mean vector source θ=θ₁...θ_m, jointly Gaussian with (X, Y), from separate encodings of X and Y, with averaged squared errors D₁...D_m. We refer to θ as the remote source, and to X and Y as its noisy measurements. To gain some insight into the results, the authors examine the total rate loss caused by the separation of the encoders in the high resolution limit