Paper detail

Towards Optimality in Transform Coding

It is well-known for transform coding of multivariate Gaussian sources, that the Karhunen-Loève transform (KLT) minimizes the mean square error distortion. However, finding the optimal transform for general non-Gaussian sources has been an open problem for decades, despite several important advances that provide some partial answers regarding KLT optimality. In this paper, we present a necessary and sufficient condition for optimality of a transform when high resolution, variable rate quantizers are employed. We hence present not only a complete characterization of when KLT is optimal, but also a determining condition for optimality of a general (non-KLT) transform. This necessary and sufficient condition is shown to have direct connections to the well studied source separation problem. This observation can impact source separation itself, as illustrated with a new optimality result. We combine the transform optimality condition with algorithmic tools from source separation, to derive a practical numerical method to search for the optimal transform in source coding. Then, we focus on multiterminal settings, for which {\it conditional} KLT was shown to possess certain optimality properties for Gaussian sources. We derive the optimal orthogonal transform for the setting where side information is only available to the decoder, along with new specialized results specific to the conditions for optimality of conditional KLT. Finally, we consider distributed source coding where two correlated sources are to be transform coded separately but decoded jointly. We derive the necessary and sufficient condition of optimality of the orthogonal transforms. We specialize to find the optimal orthogonal transforms, in this setting, for specific source densities, including jointly Gaussian sources.