Multiple Description Image Coding Based On The Theory Of The Compression Perception Research

When the signal transmission in harsh channel, The data packet may be the wrong coded or loss packet due to noise and other problems, Hence, how to solve the above problems for Image quality declining in receiving end has become a very popular research projects in recent years. As a method of solving this problem, multiple description coding(MDC)is concerned widely, due to satisfy the requirement of real-time data transmission and reduce data distortion. Because the method of Multiple Description Coding is divide the original image into multiple independent descriptions and transmission through a separate channel to the decoder, so that each description can independently recover the image of acceptable quality, Thus it can enhance the robustness of the signal in the image communication applications.Transmission in harsh channel, especially in the wireless channel and network channel, conventional approaches to sampling sis, always follow Shannon principle. It take great costs on data storage. Compressed sensing theory brings a revolutionary breakthrough for signal collection technology, it adopts the adaptive linear projection to keep the signal of the original structure, put forward a kind of new sampling theory, can be much lower than the Nyquist sampling rate sampling signal, by means of numerical optimization problems accurately reconstruct the original signal. Compressed sensing basic argument is if the signal is sparse, can be projected onto a uncorrelated random matrix from transform based and get far less than the length of the signal measured values, and then by solving the optimization problem, accurate reconstruction signal.This paper will increase the sparsity of the image through wavelet transform firstly, then generates a random observation matrix by Matlab, By using random matrix to transform matrix of correlated observations, Then we try several image segmentation method for image segmentation of the observation matrix. Using different quantization precision combined into a plurality of description, and coder decoder through separate arithmetic. Then by using OMP algorithm and wavelet transform comparative reconstruction image compression ratio and PSNR value (peak signal-to-noise ratio), the results show that the image parity column segmentation methods can obtain the high quality of reconstructed image.