The Design Of Multiple Description Coding For Image Transmission

With the rapid development of wireless communication network technology andmultimedia applications, especially the personal smart mobile terminal has become morepopuler since2009, people obtain the required images and other multimedia informa-tion more frequently from the network. Wireless communication demand has ledto serious network congestion. In addition, the wireless network environment is usual-ly complex and time-varying, these make user have a terrible experience, especiallyin the use of real-time multimedia applications, such as the FIFA World Cup live webcasting and so on.In order to get a satisfactory experience in the complex environment, telecommuni-cation enginneer must not only consider the real-time feature of information transmission,but also take the robustness into consideration when designing the image or video codecs.Multiple description coding and joint source-channel coding technology enables usersto get a good experience in the complex and time-varying network environment.Followings are the major work in this article:(1) A kind of multiple description coding based on quincunx sampling has beendesigned. Firstly, we sample the matrix of the original image by the quincunx samplingmethod. This internal redundant information in single description will be conducive toestimate the lost one. For the case of the two descriptions, it also can make thereconstruction of the image easy and fast. Besides these, the designed structure caneffectively improve the quality of reconstructed image in the case of two availabledescriptions.(2) We reconstruct the image by three decoding methods in the situation of twoavailable descriptions when we use the multiple description coding based on quincunxsampling. Method3, in which we use the filter matrix, can effectively remove the errorinformation in the reconstruction description, so that we can improve the reconstructedimage quality. It can gain about3dB as compared to method2.Filter matrix has acontrollable size and is easy to be designed and applied.(3)We make a performance comparison between odd-even column sampling MDC andquincunx sampling MDC separately with the same and the different code rate underdifferent SNR, we get the performance for the different combinations of source andchannel coding，and give their advantages and disadvantages. The combination, whichconsists of quincunx sampling multiple description coding and1/4code rate Turbo code,can obtain the maximum peak signal to noise ratio, but the highest complexity and the largest amount of computation.This combination is more suitable for low signal noise ratioconditions; However, the combination which consists of parity column sampling multipledescription coding and1/3code rate Turbo code can get the minimum peak signal to noiseratio, but the least amount of calculation and the best real-time. This combination is morefit for high signal noise ratio conditions.(4) A joint source-channel coding scheme based on multiple description coding andTurbo code has been proposed. For different channel state information, we give thespecific adjustments standards for the source coding and channel coding to making imagereliable transmission at low SNR and real-time transmission at high SNR.