Design On Optimization For HEVC Inter Prediction

High Efficiency Video Coding(HEVC)is a new generation of video coding standard jointly developed by Video Coding Experts Group(VCEG)and Moving Picture Experts Group(MPEG).Compared with the predecessor H.264/AVC,HEVC can save about half of the bitrate under the same coding quality while the corresponding computational complexity increases greatly.How to effectively improve coding speed and ensure the coding quality has become an important issue in the research of HEVC technology.As the most complex module in HEVC,inter prediction is studied intensively in this thesis,and the following optimization is carried out:1.HEVC inter prediction module traverses the fixed Prediction Unit(PU)mode set in order at each Coding Unit(CU)partition depth to get the optimal PU mode for each depth.In this thesis,by employing the PU mode information of spatiotemporal adjacent coded CUs,the skip mode early decision based and mode complexity prediction based fast algorithms are designed,which are combined to form the proposed PU mode decision fast algorithm.The PU mode decision fast algorithm can determine the optimal PU mode in advance or reduce the traversal number of PU modes,so as to reduce the computational complexity of the inter prediction PU mode decision process effectively.2.HEVC inter prediction module needs to split CTU recursively and determine the optimal coding size combination according to RD cost,which is a very complex procedure.In this thesis,the correlation between the spatiotemporal adjacent coded CUs and the current CU is utilized to reduce the searching range of CU depths.Also combined with the RD cost of current CU,the CU depth prediction based and RD cost based fast algorithms are designed,which are combined to form the proposed CU size decision fast algorithm.And the computational complexity of the inter prediction CU size decision process can be reduced obviously.Each proposed algorithm and the complete optimized algorithm are tested using the standard test sequence set published by JCT-VC.The experimental results show that compared with the standard test model HM 16.20,the complete optimized algorithm can save 55.89%of coding time,with 2.08%increase of BDBR and 0.068dB loss of BDPSNR-Y under Random Access Main(RAM)configuration.Also the complete optimized algorithm can save 47.71%of coding time,with 2.17%increase of BDBR and 0.083dB loss of BDPSNR-Y under Low Delay Main(LDM)configuration.