Research On Iterative Decoding Technologies For Unidirectional Distributed Video Coding

With the popularization of multimedia devices and the development of communication technologies,emerging applications such as wireless video surveillance have received more and more attention.At the same time,the relatively low computational capabilities of their encoding devices have also brought new challenges to encoding algorithms.Traditional video coding standards often have high coding complexity,and the emergence of Distributed Video Coding(DVC)makes it possible to allocate the coding complexity flexibly.Therefore,DVC has become a hot topic in video coding.The traditional DVC performs rate control at decoder based on the feedback channel,and in order to avoid the delay caused by the feedback channel,Unidirectional Distributed Video Coding(UDVC)removes it,and the code rate is estimated by the encoding rate control algorithm,which is more adaptable to applications requiring high real-time or no feedback channels.However,the independent encoding of DVC makes the encoder unable to fully understand the correlation between videos,and the code rate will always be underestimated or underestimated,which will affect the system performance.Thanks to the asymmetric structure of DVC,when the code rate estimation is inaccurate,appropriate compensation techniques can be introduced at decoder to improve the decoding effect.The accuracy of correlation noise and the quality of side information are the main factors affecting DVC.In this paper,based on the characteristics of DVC,two improvements are proposed for the decoder:(1)A correlation noise refinement algorithm based on iterative decoding is proposed,where the reconstructed value obtained by the previous decoding refines the correlation noise residual to obtain a more accurate correlation noise distribution,and then re-decodes the bit plane obtained by the last decoding to improve the reconstruction performance,and during refinement process,considering whether the reconstruction values are decoded correctly,classifies them to improve the precision of the refinement;(2)We proposed an improved side information refinement algorithm,where Error-decoded bits in the reconstructed frame are marked,and a new reconstruction frame can be achieved by replacing the original one with corresponding side information,meanwhile the refinement of the side information is refined by this reconstructed frame,so that a more accurate edge information frame is obtained.This algorithm corrects the misleading of the refinement process caused by incorrect reconstruction frames in the existing side information refinement algorithm.Finally,we propose an UDVC system based on DISCOVER video codec,combining the algorithm proposed in this paper,which is based on iterative decoding.In UDVC system,the correlation noise and side information are refined,which makes the accuracy of correlation noise model and the quality of side information improved gradually,thus the performance of system is progressively upgraded.In order to verify the availability of the proposed algorithm and UDVC system,the module tests are performed for above two algorithms on four typical video sequences.At the same time,the comparation between UDVC proposed in this paper and the state-of-the-art DVC codec is made,and the experimental results show that our proposed UDVC can achieve the performance equivalent to DISCOVER,which proves that our proposed algorithms have good research results.