Research On Video Transcoding Technology For Mobile Communication

With the rapid development of computer and digital communication technology,video communication of small mobile devices is more and more popular,but its battery capacity is limited,so the coding method used must meet the requirements of low power consumption,real-time and high efficiency.The new-generation video coding standard HEVC(High Efficiency Video Coding)adopts transformation and prediction methods in its coding part,which has high compression efficiency but higher coding complexity;Distributed Video Coding(DVC)transfers the mining of video correlation from the encoding end to the decoding end,so as to achieve the purpose of low complexity and low power consumption,but the complexity of the decoding end is very high.Therefore,neither of these two solutions can well meet the video communication requirements of low-power mobile terminal devices.Considering the respective advantages of DVC and HEVC,the DVC encoding scheme can be used at the transmitting end,and the HEVC decoding scheme can be used at the receiving end,so the DVC-HEVC code stream needs to be transcoded on the server.This transcoding process includes the decoding of the DVC code stream and the HEVC re-encoding of the decoded video stream,both of which have high complexity,resulting in the transcoding process taking too much time.Therefore,how to speed up the DVCHEVC transcoding process becomes a key issue.Based on the DVC-HEVC transcoding for mobile communication,this thesis studies the transcoding speed,mainly including the following aspects:(1)In the DVC-HEVC transcoding process,the DVC decoding needs to perform joint decoding,and the computational complexity is high.Therefore,this thesis proposes a DVC decoding acceleration processing algorithm based on the CUDA computing platform.The parallel acceleration of the DVC decoding process mainly includes the parallel acceleration design of the motion estimation module and the parallel acceleration design of the LDPCA decoding module.Experiments show that,compared with the conventional non-accelerated DVC decoding scheme,this algorithm can effectively reduce the time required for DVC decoding in the entire transcoding process,and accelerate the speed of DVC decoding without affecting the quality of video images.(2)Due to the high computational complexity of HEVC re-encoding in the DVCHEVC transcoding process,this thesis proposes a DVC-HEVC fast transcoding algorithm based on SVM classification,which uses the effective information extracted in the DVC decoding process to realize the fast splitting of CU.Experiments show that,compared with the simple cascading transcoding scheme,this thesis can greatly reduce the time spent in HEVC re-encoding process while maintaining the video reconstruction quality.(3)On the basis of summarizing the above two acceleration schemes,the overall scheme of DVC-HEVC transcoding for mobile communication is proposed,and the corresponding experimental test of the whole system scheme is carried out.Compared with other DVC-HEVC fast transcoding schemes,the algorithm in this thesis proves its superiority.