The embedded codestream structures, associated with a variety of progressive image compression methods, suffer from severe error sensitivity when transmitted over wireless channels. In addition to employing forward error correction (FEC) codes, joint source-channel coding (JSCC) is a typical approach to combat the significant quality degradation. In this paper, a joint source-channel decoding (JSCD) method is designed when transmitting the JPEG2000 codestream over uncorrelated flat fading channels using low-density parity-check (LDPC) codes. Error resilience modes implemented in the JPEG2000 codec makes it possible to provide new messages as feedback to the channel decoder. Thus, the log-domain sum-product decoding procedure for LDPC codes is updated during each iteration according to the feedback information from the JPEG2000 decoder. Two adaptive modification algorithms are proposed to update the log-likelihood ratio (LLR) of the bit nodes at the receiver with and without channel side information (CSI), respectively. Simulation results show that the joint source-channel decoding methods not only improve the quality of the reconstructed image greatly but also reduce the number of iterations significantly.
展开▼
