HEVC Video Transcoding Under The Mobile Network

Along with the rapid development of the mobile Internet revolution,watching video on the mobile terminal has become the mainstream of people's video consumption in the past 5 years.Based on the last generation standard H.264,HEVC has made many innovations in coding unit and prediction mode.In the case of the same video quality distortion,the bit rate is reduced by about one time,which make it become the mainstream of today's technology.However,the improvement of compression efficiency inevitably leads to higher coding complexity.In mobile terminals,due to the instability of wireless bandwidth,users will frequently encounter stutter when watching the video,which is a great influence on the watching experience.Considering the high cost of storage,video service providers only place a limited number of high bit rate video streams in the cloud,which is difficult to meet different needs from users.So,how to achieve fast video transcoding and make the cloud server transcodes suitable version to meet different users has gradually become one of the directions of the industry.In this paper,I introduced the development of video coding standard,and elaborated the key technology of the latest coding standard HEVC.Based on the current technology and research status,a fast decision algorithm about the coding unit and prediction unit is proposed.The research content of this article mainly includes the following two aspects.One is the fast CU decision algorithm.In the decoder,I extract high bit rate video coding unit information and combine it with machine learning theoiy.Then a multi-class adaboost classifier is used to train the video transcoding framework,so as to determine the coding unit mode of low bit rate video stream.The other is a fast PU decision algorithm.After knowing the prediction unit information,the algorithm will reduce the candidate range of the prediction unit and terminate the rate distortion cost calculation of prediction unit with low probability,so as to reduce the computational complexity.The experimental results shows the algorithm achieves 84%coding time saving,and the PSNR only increases by about 0.11 comparing with HM16.0.