Research On Fast Encoding Algorithm In HEVC Video Coding

People are enjoying video service in computers,TVs,mobile phones,and other different kinds of equipments.A video sequence is able to be decomposed into a series of pictures containing huge amount of information.No system is able to handle raw video materials in such a large scale.It is inevitable to compress video into an easily stored format using video coding technology.High effcient video coding(HEVC)standard was proposed by Joint Collaborative Team on Video Coding(JCT-VC),which is composed by group of video coding experts from ITU-T Study Group 16(VCEG)and ISO/IEC,in 2013.HEVC is also called h.265 in ITU-T standard list.The goal of HEVC is acquiring half of the bitrate while keeping the same video quality compared with H.264/AVC.HEVC meets the requirement in the aspects of coding performance.However,HEVC faces the challenge of heavily time consumption.Experiment has shown that compressing one frame of a 1080p sequence takes more than one minute with HEVC test model(HM)under Random Access configuration,which is unacceptable in actual application.The heavy time-consumption shows the remained room for optimizating the compressing process of HEVC is still huge.The paper proposes four methods to reduce time-consumption of HEVC.The experiments have shown that the proposed methods in this paper reducing coding time in different degrees.The research work is composed of following four parts:Firstly,the quad-tree structure of image partition is studied in-depth.To meet the demand of high resolution video,quad-tree image structure is introduced in HEVC standard.Experiment had shown that the partition method adapt both flat place and detailed place.However the quad-tree partition also brings huge coding complexity.To solve this problem,the paper raises two solutions to optimize time-comsumption brought by image structure searching:the first solution is based on histogram estimation.The second method is based on Adaboost.In both method,the encoder will predict the structure of each coding unit(CU)and skip redundant modes.Experiments had shown both proposed method shorten encoding time of HE VC codec.Secondly,motion estimation technic is studied in this paper.Motion estimation,as one of the most time consuming modules,is the core process of inter coding.An adaptive motion vector searching range decision algorithm is proposed in this paper.With proposed method,the codec is able to predict searching range for every motion estimation blocks and encoding time could be saved.Thirdly,the "Merge" mode is studied in the paper.Merge mode,introduced by HEVC,is able to remove redundancy existing between adjacent blocks by merging the motion information of neighborhood blocks.Experience had shown the possibility of adopting Merge mode as the best mode for a block is high.Our proposed method takes the fact into consideration,and automatically decides whether the current block adopts Merge mode or motion estimation process.If the method decides the block adopts Merge mode as the best mode,motion estimation will not be performed for current block,and encoding time will be saved.Experiment shows with proposed method,encoding time will save 10.9%at the cost of 0.27%BD-rate loss.