Research On Method Of Image Transmission Based On Quantized Compressive Sensing In Wireless Sensor Network

This thesis rclys on (Compressing Sensing, CS) theory by Candes, Tao, Romberg: Sampling and compression happened at the same time, which makes sampling the signal compression ratio reduce greatly. As the most effective and intuitive parameter for users to observe the site, the image has been paid close attention by researchers in wireless sensor network. Because of the large amount of image data, the traditional sampling theorem has been higher required on node processing speed, storage capacity and energy consumption, which puts resistance on the multi-media sensor network promotion. Compressive Sensing theory (CS) uses the image signal’s sparsity and compressibility to remove data redundancy, save storage space, simplify data processing and prolong network life, etc. But after being quantitative and channel transmission, the sample values will be influenced by quantization noise and channel noise. If adopting reconstruction algorithm to reconstruct directly, then the restored signal will produce large distortion. So WSN image requires not only optimal quantization coding method, but also robust reconstruction algorithm.At present the main research on the measured value of Huffman coding and uniform coding, the order of magnitude and reconstruction algorithm of signal recovery, but rarely consider the impact of transmission noise on the signal reconstruction. The above research only considers a single, it doesn’t give a solution to image from the point of view of system in WSN. This paper briefly describes related theoretical knowledge of compressive sensing, background and significance of the research, reviews current status at home and abroad. Considering the characteristics of sensor network, with the help of the existing communication model and image:image has general characteristics of the ordinary signals and high complexity. This thesis main research measured value encoding and transmission problems, and the encode and decode adaptive quantization coding and Huffman coding. And the upper bound of the error system is deduced in the paper. And it gives the design scheme for optimal nonlinear quantizer according to the prior information of the transmitted data, and the performance of nonlinear quantization is superior to the average quantization is proved by simulation. In the condition of large channel effects, this paper puts forward a robust reconstructed method based on subspace by using the incoherentness between channel noise and signals. Finally, the simulation use optimized quantizer and de-noising method could effectively improve the performance of system under the condition of high channel noise.