低复杂度的图像信源-信道联合译码算法

该文提出将图像编码后残留冗余的马尔可夫场模型分解为4个方向的马尔可夫链，并结合简化的模型及低密度奇偶校验码(LDPC)译码的软输出进行信源-信道联合译码。将分解后信源中多个方向上同时存在的相关性看作一种特殊的天然信道编码方式，利用前向-后向算法、和积算法以及信道译码软输出分别对信源符号进行串行和并行的译码。仿真实验表明，与传统利用马尔可夫场模型的联合译码算法相比，该联合译码算法降低了复杂度，同时提高了重建图像的峰值信噪比。

Low-Complexity Joint Source Channel Decoding of Image

Markov Random Field Model (MRFM) is separated into four Markov chains which are used to represent the residuals of encoded image source. Combined with the soft output of Low Density Parity Check (LDPC) code, this simplified model is used in joint source channel decoding. Different correlation in different direction in source is regarded as a kind of “natural” channel code. In order to utilize the correlation, a serial decoding using forward-backward algorithm and a parallel decoding using sum-product algorithm are proposed respectively. Simulations show that compared with the traditional joint source channel decoding algorithm based on the MRFM, the proposed algorithm has lower complexity and better PSNR of the rebuilt images.