图像压缩感知在信源信道联合编码系统中的应用

The Application of Image Compressed Sensing in Joint Source-Channel Coding System

提出了一种将基于离散小波变换的图像压缩感知应用在信源信道联合编码系统中的方法。该方法将双重不等差错保护与码率动态分配机制相结合,在信源编码部分,根据图像小波变换后各频带所包含重构信息量的差异,利用压缩感知算法进行不等压缩,产生渐进性的信息流,信息流通过Huffman熵编码后成为适合信道传输的二进制码流;在信道编码部分,根据信源编码后二进制码流的渐进性信息,动态分配Turbo编码码率,从而实现信道的不等差错保护。该联合编码方法在信道资源受限的情况下,可对资源进行优化分配,达到良好的端到端通信效果。以标准灰度图像Lena图为例,编码后通过高斯白噪声信道,仿真结果为:当信噪比为4 dB时,图像重构均方误差为0.061 6,重构性能良好;同时,系统获得了高达4:1的压缩比,系统耗能减少,传输效率增加。

This paper proposes a method which applies image compressed sensing based on the discrete wavelet transform(DWT)in the joint source-channel coding system.This method combines the double unequal error protection and dynamic rate allocation mechanism.According to the difference in the sub-band reconstruction information of each frequency after DWT of the image,unequal compression is performed using the compressed sensing(CS)algorithm to produce a progressive information flow in the source coding part.Consequently,the information flow is converted into binary codes which is comparably suitable for the channel transmission after the Huffman entropy coding.The rate of Turbo is distributed based on the progressive bit stream dynamically in the channel coding part,to achieve the unequal error protection of the channel.During instances when the channel resources are limited,the channel resource allocation can be optimized to achieve good end-to-end communication performance.Performing a simulation of the standard gray scale image of Lena and passing through the white Gaussian noise channel after coding,the result is as follows:When the SNR is 4 dB,the image reconstruction mean square error is 0.0616 with good reconstruction performance.Meanwhile,the system obtains the compression ratio up to 4:1,reduces the system energy consumption and increases the transmission efficiency.