一种图像多描述网格编码矢量量化算法

New Approach to Multiple Description Trellis Coded Vector Quantiza tion of Image

在互联网和无线通信环境下,图像通信常因数据包传输出错或丢生而导致质量的严重下降,多描述编码是解决这个问题的一种新途径。该文提出了一种多描述网格编码矢量量化新算法,给出了多描述分解的方法和规则,并通过构造一个代价函数,提出了最优搜索策略,用维特比算法对多描述编码最优路径进行搜索。仿真结果表明,新算法有效利用网格编码矢量量化将量化码书大小扩大一倍的特性,提高了中央通道的编码性能,而如果只收到单个描述,解码图像仍能获得较好的重建质量。

Multiple Description Coding(MDC)is a new method to solve the problem of image communication fading over erasure channels such as Internet and wireless networks,where packet loss and errors are inevitable.In this paper,a new method to image multiple description TCVQ is proposed.Firstly,a cost function is constructed to choose the code rule of MDC,and then viterbi algorithm is used to search the optimized quantization sequence.Experimental results show that by exploring TCVQ to double the size of vector quantizer codebook,this algorithm improves the performance of the centre channel effectively,and even if only one description is received correctly,the reconstructed image is still acceptable.