Distributed Source Coding Based on Denoising Theory

Abstract

Aiming at the error propagation problem and other shortcomings of existing distributed source coding (DSC) schemes, we propose a DSC scheme based on the denoising theory (DSCBDT). First, the amount of data that need to be transmitted is reduced by sampling the source output sequence. The un-sampled sequence is compressed by conventional arithmetic coder. Finally, the receiver recovers the sampled symbols by making full use of the correlation between the side-information (SI) sequence and the un-sampled sequence. The experiment results show that compared with the low-density parity-check (LDPC) code-based DSC scheme, the proposed scheme shows higher compression ratios and lower error rate when the correlation between the sources is weak. Specifically, the error propagation problem is effectively avoided. The DSCBDT scheme is highly efficient, practical and easy to implement.