Multi-Mode Nested Quantization in Presence of Uncertain Side Information and Feedback

Abstract

In this paper, the problem of uniform quantization for the Wyner-Ziv (WZ) problem in presence of uncertain side information at the encoder is addressed. The uncertain side information problem arises in data compression with side information (SI) when the joint probability distribution of source and SI is not (fully) known at the encoder. The uncertainty is due to the time varying nature of the dependency of source and SI in settings such as wireless sensor networks and distributed video coding. In this paper, a solution is proposed which consists of a multi-mode WZ encoder in conjunction with a carefully designed feedback scheme from the decoder. Each encoder mode is associated with a uniform nested quantizer. In this framework, the quantization parameters, the feedback scheme, and the source coding rate are jointly optimized to minimize the average rate or distortion. Two distinct schemes with adaptive or fixed transmission rate are proposed. Simulation results are provided which demonstrate the effectiveness of the presented solutions and assess the effect of different design parameters including the number of modes (feedback rate) on the performance.