Research On Compressed Sensing Based Robust Image Transmission

With the rapid development of wireless communication technologies and mobile Internet,wireless communication traffic has grown explosively.In particular,high bandwidth demand services represented by multimedia transmission have brought enor-mous pressure to today's wireless networks.Besides,because of the channel time varia-tion and the channel heterogeneity in the broadcast scenario,the traditional digital trans-mission suffers from the cliff effect.In the digital system,the cliff effect not only affects broadcast efficiency and fairness,but also seriously impairs users' experience.In re-cent years,to avoid such cliff effect,a pseudo-analog transmission scheme has attracted much attention of the researchers.Different from traditional digital transmission,in the pseudo-analog transmission system,quantization,entropy coding,and channel coding are removed and a series of linear processing is used to make the distortion of the whole system is linearly changed with the channel quality.So the pseudo-analog transmission scheme has a good adaptability and provides users with reconstruction performance that scales with the channel qualities.Although pseudo-analog transmission technology has good characteristics in wire-less multimedia transmission,there are still existing many problems in actual wireless transmission applications.One of the common problems is the bandwidth matching in a bandwidth-constrained environment.Due to the different importance of the transmit-ted coefficient blocks,it's not an optimal approach for the bandwidth matching through discarding some coefficient blocks in the pseudo-analog transmission scheme.This is-sue is studied firstly in this dissertation.Considering the characteristics of images in transform domain,compressed sensing is used for source compression,so that more coefficient blocks could be transmitted.Furthermore,to improve the transmission per-formance,we perform bandwidth allocation according to the coefficient block weights,and develop a power allocation algorithm to minimize the system distortion.Finally,the simulation results show that the proposed scheme has better transmission perfor-mance than other pseudo-analog transmission schemes under various channel condi-tions and different bandwidth budgets,which proves the effectiveness and superiority of our scheme.In addition,although the pseudo-analog transmission scheme can eliminate the cliff effect and provide robust performance that match the channel conditions,it's a challenge to solve the problem of antenna heterogeneity existing in current wireless broadcasting systems.With the popularity of the multi-antenna technology,different equipments may have different numbers of antennas.Therefore,the pseudo-analog transmission scheme needs to adapt to the diverse channel quality and different numbers of antennas simultaneously.To solve this problem,we propose a compressed sensing based pseudo-analog transmission scheme,which uses a spatial multiplexing method to transmit linear measurements of the coefficient block at the sender,and implement reconstruction at the decoder according to compressed sensing reconstruction algorithm.In this way,the single-antenna user can still performs signal recovery through the reconstruction algo-rithm.Furthermore,the more antennas at the receiver,the more measurements can be received,and the quality of reconstruction will be increased accordingly.In the image transform domain,considering the poor sparseness of some coefficient blocks which makes a large system distortion,we directly transmit such blocks without compressed sensing processing,and these blocks can be reconstructed at the decoder with a mini-mum mean-squared error method.Based on this,we propose a coefficient block clas-sification strategy.Furthermore,to minimize the total system distortion,we formulate and solve the power allocation problem in our scheme.Compared with other schemes,the simulation results demonstrate that the proposed scheme can achieve better trans-mission performance.Furthermore the reconstruction quality in our scheme scales with the channel quality and number of antennas.