Research On Acceleration Method Of Quality SHVC

With the development of Internet,video coding technology is constantly updated.In order to better adapt to the heterogeneity of the network,the diversity of terminal device and the difference of user demand,quality SHVC scalable video coding technology emerged.As a scalable extension of the new generation of High Efficiency Video Coding(HEVC)standard,Scalable High Efficiency Video Coding(SHVC)standard adopts the multi-layer structure of base layer and enhancement layer,and its coding complexity has been increased significantly,which brings great challenges to its application and development.Therefore,under the premise of negligible video quality loss,reducing coding complexity is one of the current research focuses.Based on the analysis of the HEVC framework and the key technologies of video coding,this thesis researches on the fast algorithm of quality SHVC to reduce coding complexity and accelerate the coding speed.Firstly,a fast Enhancement Layer(EL)algorithm based on depth prediction and Inter Layer Reference(ILR)mode decision is proposed.According to the spatial and inter-layer depth correlation of the Coding Tree Unit(CTU),the depth value range of the CTU is predicted,and the depth value with low depth correlation is skipped to reduce the coding complexity.At each depth value,whether the ILR mode is the optimal mode is judged according to the residual coefficient distribution in the EL,so as to skip the traversal of the intra prediction mode and reduce the coding time.Within the depth value range,the residual coefficient distribution is detected by statistical method,and the threshold value of depth termination of different sizes coding unit is studied by experiments.For the case of less than the threshold,the calculation process of the subsequent transform quantization and inverse transform inverse quantization is directly omitted to further reduce the coding complexity.The experiment results show that,compared with standard SHVC encoders,the total coding time and EL coding time of the proposed algorithm are reduced by 46.54%% and 80.49% respectively,while ensuring that the video quality loss is within an acceptable range.Secondly,this thesis proposes a fast EL algorithm based on edge detection for depth value prediction and inter Prediction Unit(PU)division in advance.The three components of the luminance component Y and the chrominance components U and V in the YUV color space are independent of each other,and the three components are separately amplified to enhance the chromatic difference,the new YUV color space using improved edge detection operator to extract the edge pixel values,predicts the depth value of the coding unit according to the proportion of the edge pixel value,and skips the calculation of the depth value with a low proportion,so as to reduce the coding time.The inter prediction mode is divided into three categories,including left and right structure,top and bottom structure and uniform structure.The residual coefficient distribution of the two parts is tested by statistical method,which reduces the range of inter prediction unit division mode to reduces the coding complexity.The experiment results show that,compared with standard SHVC encoders,the total coding time and EL coding time of the proposed algorithm are reduced by 31.53% and 58.65% respectively,while ensuring that the video quality loss is within an acceptable range.