Vector Gaussian Successive Refinement With Degraded Side Information

In this paper, we consider Successive refinement for the Wyner-Ziv source coding problem, in which sources are vector Gaussian distributed and side information follows a degraded order assumption. We fully characterize its ratedistortion region with covariance mean square error distortions. We use jointly Gaussian auxiliary random variables to evaluate the existing single-letter description of the rate region by Tian and Diggavi, which gives the achievability. The converse relies on the information-estimation relationship, with which the tight outer bound is obtained.