Research On HEVC Inter Frame Predictive Coding

With the rapid development of communication technology,computer technology and multimedia technology,multimedia has been deeply integrated into people's lives,which is the largest amount of video information in multimedia,is also the most common media,however,the amount of data is very large,not compressed coding of video transmission in the Internet is difficult to directly.Video compression coding technology is the premise of video storage and transmission,and it is also the core technology of digital TV,video surveillance and network video.In recent years,more and more video resolution,high resolution video puts high demands on the storage,the ITU-T video encoding and MPEG board of the joint expert group customized video encoding standard,the new HEVC(High Efficiency Video Coding),and on January 2013 formally promulgated,while providing a decoder reference software HM,has aroused enthusiasm the research of scholars.Inter frame encoding technology is one of the key technologies in video encoding,which is the most time-consuming part,accounted for about half of the computation of the entire-video encoding,how to balance the computation and the computation accuracy is always a hot and challenging topic,scholars have put forward various fast algorithms,however,inter the prediction algorithms encoding are based on block matching,in recent years,some scholars began to introduce the affine transformation model in inter frame encoding,to further improve the encoding quality.Based on the study on the structure of the whole HEVC encoding,in-depth study of the inter frame encoding,combined with some scholars on the affine transformation of the.preliminary exploration results,to further explore the mode of division,affine transformation and motion vector computation complexity,and the validity of the proposed algorithm and the view has been proved by experiments,the the main work is as follows:(1)Propose a method to describe the affine motion model based on coordinates,and further subdivision of the affine transformation unit,the affine transform transformation block based affine transformation,a triangle based experiments show that this algorithm can get the matching accuracy is greater than the block translation;the research produced by affine transformation MV(motion vector)nature,through theory and experiment proved that the correlation between adjacent MV is poor,Merge does not have a similar pattern of predictability;proposes a method for affine motion estimation fast search method,the matching accuracy loss is small,shorten the time of motion estimation.(2)In view of the HEVC reference software is too complex,inter frame prediction encoding structure is huge,so this algorithm could not be embedded in HM,but this paper modeled HEVC structure independent design the interframe prediction encoding,including prediction unit classification,motion estimation,quantization and entropy encoding,through the experimental data to make B-D curves of different test sequences the simulation results show that the superiority of the proposed algorithm in high precision encoding mode.