Research On Intra/Inter Prediction Accelerating Algorithm Of HEVC

With the increase on consuming demand for high-definition and ultra-high definition video,people have imposed higher requirements on the compression encoding efficiency of digital video.On account of limitations in terms of coding efficiency and parallel processing in H.264/AVC,the compression performance of H.264/AVC can hardly meet the requirements for high-definition video.The new generation of HEVC(High Efficient Video Coding)solves the problems well at the expense of high computational complexity.How to reduce the amount of coding computation and achieve encoder acceleration on the premise of guaranteeing HEVC encoding performance,which is the key to expanding the application of HEVC in the video field.In the paper,intra prediction coding and inter prediction coding are studied for HEVC deeply,proposed two improved algorithms to reduce encoding complexity of HEVC sharply.In order to reduce the complexity of mode selection in intra prediction,this paper proposed an accelerating mode algorithm based on the boundary direction in intra prediction.The algorithm is improved from both rough mode decision(RMD)and rate-distortion optimization(RDO)early termination.In the rough mode decision process,in order to reduce the complexity at the RMD process,firstly we find a direction of the current prediction unit(PU),then reduce the number of intra prediction modes by determining the boundary direction.RDO early termination is making use of the correlation between the adjacent mode number of candidate modes to reduce the number of traversal from other irrelevant candidate modes,to reduce the RDO calculation process,so that the algorithm complexity is reduced in intra prediction mode and intra prediction coding process is speeded up.Compared with HEVC Test Model-16.9(HM-16.9),this algorithm reduces coding time by about 42% on average and the peak signal to noise ratio of video decreases only 0.11 dB on average in the case of the output bitrate is hardly increased.Aiming at the increase of the coding unit(CU)and PU structural division types bring about a mount of encoding computation in inter prediction process,an accelerating algorithm for inter mode decision based on motion features is proposed.First,for CU block at different motion areas,utilizing depth correlation between the current CU and the adjacent CU in the spatio-temporal to reduce the current CU depth traversal range.Then,accord to the similarity at spatial partition of the PU mode,among the current CU,the adjacent CU in the spatio-temporal and the previous depth CU,to reduce PU mode traversal range for accelerating inter prediction process.The experiment results demonstrate that compared with HM-16.9,at different access ways,the proposed algorithm can reduce about 54% coding time on average,the video quality is basically unchanged and ensure the output bitrate is less increased.