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

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 0.061 6 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.