Research And Improvement Of HEVC Inter-frame Prediction Algorithm

The traditional way of relying on words and pictures to obtain information is no longer the mainstream.Because of its vivid,comprehensive and real characteristics,video enters people's field of vision in an unpredictable posture and becomes one of the main ways for people to obtain information.More efficient video coding can better cope with the diversification of application scenarios and the need for ultra-high definition resolution in the market.The High Efficient Video Coding standard is currently one of the top performance video coding standards.Compared with the previous generation video coding standard,Advanced Video Coding,it has a higher compression ratio,but at the same time,its complexity is greater.For many applications that require real-time encoding,High Efficient Video Coding is limited.In video coding,since time redundancy is much larger than spatial redundancy,most image frames use inter-frame prediction.Therefore,the complexity of inter-frame prediction will be effectively reduced,and the overall complexity of video coding is greatly reduced,which will enable High Efficient Video Coding to be better applied to real-time coding scenarios.In order to reduce the coding complexity,this thesis deeply studies the inter-frame prediction technology in High Efficient Video Coding,and improves the motion search algorithm and coding unit depth selection algorithm in inter-frame prediction.The main work of this paper is as follows:1.On the basis of reading a large number of video coding proposals,books and papers,the current research status of video coding at home and abroad is introduced,and the research status of prediction algorithms is further refined.2.Review the relevant knowledge in the efficient video coding standards.Includes coding process for High Efficient Video Coding,partitioning,intra prediction,inter prediction,and video quality assessment.3.The relationship between the selection of the motion vector of the current coding block and the motion vector of the adjacent coded block is studied in depth,and the definition of the motion complexity of the coding unit is proposed.The motion complexity is then used as a criterion to determine whether to simplify the motion search step.More over,the influence of the selection of the motion vector of the inter-frame 2N×2N mode prediction unit on the motion vector selection of the prediction unit in other inter-modes is studied.A decision condition is proposed to determine whether to simplify the search process of other inter-mode prediction unit motion vectors.The experimental results show that compared with the standard test software HM16.16,the algorithm saves an average of 23.5% of the motion estimation time.4.In this paper,the coding unit partitioning process in inter-frame prediction is deeply studied,and a fast coding unit depth decision algorithm is proposed.The core of the algorithm is the minimum risk Bayesian criterion.This paper constructs a coding unit depth selection model based on the minimum risk Bayesian criterion,and then uses the huge database information to learn Bayesian threshold and Bayesian conditional probability density offline,establish a lookup table,and select the optimal by calculating mutual information.A subset of coding features.The algorithm can reduce the time at which the current coding unit selects depth.The experimental results show that compared with the standard test software HM16.16,the algorithm saves an average of 45.9% of the total coding time.5.This paper summarizes the advantages and disadvantages of the proposed algorithm and makes further plans and prospects for future work.