Binary Wyner-Ziv and CEO Problem Coding Under the Hamming Distortion Criterion

In this paper, we present a practical encoding and decoding scheme for the binary Wyner-Ziv problem based on graph-based codes. Our proposed scheme uses low-density generator-matrix (LDGM) codes in lossy source coding part and low-density parity-check (LDPC) codes in syndrome generation and decoding part. Actually, we apply Bias-Propagation algorithm for lossy source coding or binary quantization and Sum-Product algorithm for syndrome-based channel decoding. Using appropriate degree distributions for LDGM codes and optimized degree distributions for LDPC codes, we will be able to achieve close rate-distortion performance to the theoretical Wyner-Ziv bound. Also, we extend our proposed scheme for presenting a practical coding scheme for the binary Chief Executive Officer (CEO) problem. In our scheme, encodig is based on binary-quantization and Slepian-Wolf coding using source-splitting technique. It is shown that, source-splitting technique is an efficient strategy for achieving non-corner points in Slepian-Wolf rate region. We show that, this technique along with iterative message-passing algorithms can be efficient for having close rate-distortion performance to the Berger-Tung inner bound of binary CEO problem for non-corner points too.

Keywords:

Wyner-Ziv Problem , CEO Problem , LDGM , LDPC , Hamming Distortion