Implementation And Optimization Of The Hi-definition Video Encoder

Rapid developments of information and network technologies have witness an explosion of various multimedia services,especially for high-definition video applications,which further bring great challenges to the video coding efficiencies.Therefore,the ITU-T Video Coding Experts Group(VCEG)and the ISO/IEC Moving Picture Expert Group(MPEG)standardization organizations working together as a partnership and jointly formulated the newest video coding stand High Efficiency Video Coding(H.265/HEVC).Compared with the last generation video coding standard H.264/AVC,it can achieve a bitrate reduction of nearly 50% while maintaining the same video quality by utilizing more flexible coding structures and lots of advanced coding techniques.However,it also requires much more computational complexity than the previous standards.Therefore it is necessary to develop high-performance and low-complexity coding techniques for H.265/HEVC.This thesis mainly focuses on the famous X265 encoder.Firstly,the encoding process,coding efficiency and coding speed of X265 are analyzed.Then,the rate-distortion optimization(RDO)methods and the parallel encoding schemes in X265 are introduced.Finally,an adaptive key-frame quantization parameter allocation method is proposed for X265 based on inter-frame dependency.The main works of this thesis are summarized as follows.X265 is becoming the most popular H.265/HEVC encoder since it can achieve high-efficiency and high-speed video coding performance.For the X265 encoder,the encoding process for different layers are firstly introduced,including sequence coding layer,frame coding layer,CTU group coding layer and CTU coding layer.Then,the RDO methods in mode selection and quantization process are analyzed,and the coding efficiency and coding speed of each RDO level are evaluated.Finally,the parallel encoding schemes in X265 are analyzed and evaluated,including frame-level parallel processing,CTU level wavefront parallel processing(WPP),parallel mode selection and motion estimation,which offers a guidance for properly allocating hardware resources under hardware resource limitations.Video coding framework utilizes lots of inter prediction techniques for removing the temporal redundancies,which introduces strong inter-frame dependencies among neighboring frames.Thus the coding distortion of a key frame may have a large influence on the coding performance of the whole GOP.Therefore in this thesis,an adaptive key-frame quantization parameter allocation method is proposed for X265 based on inter-frame dependency.Firstly,a measure for estimating the inter-frame dependency is established,and based on which the quantization parameter for each key frame is automatically calculated based on the estimated inter-frame dependency.Experimental results show that compared with the default quantization parameter settings in X265,the proposed method can achieve an average coding gain of 2.04%,while maintaining the same computational complexity.