Distributed Compressed Sensing Based On The Assumption Of Ensemble Prediction

The intra frame coding technology is used in traditional video coding and decoding technology,which includes a lot of motion estimation and motion compensation operation that greatly increases the complexity of coding,so it is not suitable for the encoding of limited resources in wireless multimedia sensor networks.To solve the above problems,we apply the theory of compressed sensing to the video coding and decoding technology.In the video compression sensing technology: video signals can be directly sampled by a measurement matrix,the sampling process is the process of encoding,and the amount of sampling data is far less than the required amount of data sampling theorem.At last we can rely on a special algorithm to reconstruct the original signal accurately by using the sampling data.Based on the low sampling rate and low coding complexity,video compression technology is especially suitable for wireless multimedia sensor networks.In the existing video compression sensing technology,the multi-hypothesis prediction technology has been paid more and more attention because of its high quality of reconstruction.However there are some defects in the multi-hypothesis prediction technology,such as high time complexity and slow reconstruction time.In order to ensure the reconstruction quality and reduce the time complexity,we propose an scheme that dynamically change the size of the searching window in the reference frame(DCSWRF)in the process of multi-hypothesis prediction.The proposed scheme utilizes the correlation between videos and takes sample in block units,in the process of multi-hypothesis prediction,the reconstructed window dynamically changes the searching window size according to the change degree of the corresponding location sub blocks in the relative reference frames of each sub block.Further more,in order to improve the reconstruction quality further,we also propose an scheme that the selection of the optimal reference frame under multiple reference frames(SOFUMF).In this scheme,the key frames are regarded as reference frames from start to end in the process of multi-hypotesis prediction but we also can use the reconstructed non key frames as reference frames.The premise is to meet the conditions of correlation,at the same time in order to balance the reconstruction quality and speed of reconstruction,we only choose one best reconstructed non key frame as the reference frame.The simulation results show the DCSWRF scheme not only effectively reduce the time complexity of the algorithm but also improve the performance in low sampling rate.Compared with the DCSWRF scheme,the SOFUMF scheme improves the performance of reconstruction.And also the SOFUMF method improve the speed of reconstruction compared to the ever schemes that not using the DCSWRF method.