Font Size: a A A

The Research On Fast Intra Prediction Algorithm For Virtual Reality 360 Degree Video Based On Improved RMD

Posted on:2020-05-22Degree:MasterType:Thesis
Country:ChinaCandidate:C XuFull Text:PDF
GTID:2428330575474267Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Virtual Reality(VR)360-degree video is a spherical video.Compared with traditional video,the VR 360-degree video has larger amount of data and higher redundancy.The existing coding framework lacks optimization of such video,and has problems such as low coding efficiency and long coding time.In this paper,the characteristics of VR 360-degree video are deeply studied.The coding speed of the encoder is improved by reducing the computational complexity of intra mode decision and coding unit(CU)partition.The main work of the paper includes the following two aspects:(1)Aiming at the problem of too many candidate modes in the intra mode decision process in the existing coding framework,an improved two-stage rough decision process(RMD)is designed by analyzing the statistical relationship between the texture characteristics of VR 360-degree video and the optimal prediction mode.The algorithm reduces the number of candidate modes of the RMD process and RDO(Rate-Distortion Optimization)process,and speeds up the intra mode decision process of the VR 360-degree video.Firstly,using the statistical relationship between VR 360-degree video texture characteristics and CU size,representative prediction modes are selected to participate in the first-stage RMD,and a possible prediction direction range is roughly selected.Then,according to the characteristics of the candidate mode selected by the first stage RMD,the number of modes participating in the second stage RMD is increased or decreased,the prediction accuracy is improved,and the mode of the candidate list for the RDO process is reduced,thereby further reducing the computational complexity.Compared with test model HM16.16+360Lib 4.0,proposed algorithm can save average of 27.7%coding time and BD-rate only increases by 0.8%.(2)Since the improved algorithm above relies on the size of CU,in order to further accelerate the speed of intra prediction process,this paper studies the CU partitioning process based on the characteristics of VR 360-degree video and a CU partition early termination algorithm based on coding bits is improved.The CU partition early termination algorithm uses the coding bits as a partitioned reference factor,and for different CUs with 64×64,32×32 and 16×16,the CU coding bits thresholds suitable for the VR 360-degree video is obtained statistically.Through the CU coding bits thresholds,the optimal CU partitioning can be achieved faster.The experimental results show that compared with test model HM16.16+360Lib 4.0,the algorithm researched in this paper saves average 49%of the coding time in All-Intra mode,and only 1.7%of the BD-rate increases.
Keywords/Search Tags:VR 360-degree video, fast intra prediction, intra mode decision, CU partition
PDF Full Text Request
Related items