HEVC帧内模式决策及其核心技术的研究(1)

硕士学位论文HEVC帧内模式决策及其核心技术的研究RESEARCHSONFASTINTRAMODEDECISIONANDCORETECHNIQUESINHEVC赵亮哈尔滨工业大学2011年6月国内图书分类号：TP391.4学校代码：10213国际图书分类号：681.39密级：公开工学硕士学位论文HEVC帧内模式决策及其核心技术的研究硕士研究生：赵亮导师：赵德斌教授申请学位：工学硕士学科、专业：计算机科学与技术所在单位：计算机科学与技术学院答辩日期：2011年6月授予学位单位：哈尔滨工业大学ClassifiedIndex:TP391.4U.D.C.:681.39ADissertationfortheDegreeofM.Eng.RESEARCHSONFASTINTRAMODEDECISIONANDCORETECHNIQUESINHEVCCandidate：ZhaoLiangSupervisor：Prof.ZhaoDebinAcademicDegreeAppliedfor：MasterofEngineeringSpecialty：ComputerScienceandTechnologyAffiliation:SchoolofComputerScienceandTechnologyDateofDefence：June,2011DegreeConferringInstitution：HarbinInstituteofTechnology哈尔滨工业大学工学硕士学位论文摘要随着科技的不断进步，人们对视觉和听觉质量的要求越来越高，高清视频和超高清视频开始被普遍关注。以H.264/AVC为代表的第二代视频编码标准已不能满足人们对于实际应用的要求，工业界和学术界对新一代视频编码标准的渴望越来越强烈。在这一形势下，VCEG和MPEG两大标准组织开始进行合作开发，成立了称为JCT-VC(JointCollaborativeTeamonVideoCoding)的联合组织，统一制定下一代编码标准：HEVC(HighEfficiencyVideoCoding)。目前HEVC尚未制定完成，但整个框架结构已基本确定下来。HEVC仍属于预测加变换的混合编码框架，然而其采用更加灵活的编码结构来提高编码效率，包括编码单元(CodingUnit)、预测单元(PredictUnit)和变换单元(TransformUnit)。其中编码单元类似于H.264/AVC中的宏块的概念，预测单元是进行预测的基本单元，变换单元是进行变换和量化的基本单元。这三个单元的分离，有利于各个单元更优化的完成各自的功能。与H.264/AVC相比，HEVC的另外一个显著区别是它采用了更大的块结构：昀大的编码单元达到了64x64个像素。本文从降低复杂度和提高编码性能两个出发点着手，来进一步改进HEVC的编码效果。首先，本文提出了一个新的帧内快速模式决策算法，来进一步降低编码端的时间复杂度，其主要思想是通过利用相邻块预测模式之间的相关性，降低进行率失真决策的帧内模式数量。其次，本文在色度像素帧内预测前对当前预测单元的参考像素进行平滑处理，一方面是为了去除参考像素中的噪声，另一方面是为了克服对预测方向并不是很明显的预测块进行严格的方向性预测。昀后，本文根据CABAC编码帧内预测模式编码时各个Bin单元中0和1出现的概率，对帧内预测模式的上下文模型进行了进一步的优化，以提高编码性能。关键词HEVC；帧内模式决策；帧内平滑处理；熵编码；上下文优化-I-哈尔滨工业大学工学硕士学位论文AbstractWiththerapiddevelopmentofthetechnology,thepeople’sdemandonvisualandauditoryqualityincreasedramatically.Asaresult,HDTV(HighDefinitionTelevisionVideo)hasattractedalotofattention.TheH.264/AVCvideocodingstandardwhichdelegatesthesecondgenerationofvideocodingstandardintheworldcan’tsatisfypeople’sdemandonpracticalapplication.Bothindustryandacademiahasstrongeageronthenextgenerationofvideocoding.Inthissituation,bothVCEGandMPEGwhicharetwointernationalorganizationsforstandardizationbegintocooperateanddevelopthenextgenerationofvideocodingnamedasHEVC,andtheyhavefoundedajointorganizationwhichiscalledJCT-VC.Nowadays,althoughHEVChasnotbeencompletelyfinished,thewholecodingstructurehasbeenbasicallyestablished.HEVCstillbelongstoblock-basedhybridvideocodingframework,however,itadoptsmoreflexiblecodingunittoimprovethecodingefficiency.ItincludesCodingUnit(CU),PredictionUnit(PU)andTransformUnit(TU).CUisamacroblock-likeunitforregionsplitting.TheCUconceptallowsrecursivesplittingintofourequallysizedblocks,startingfromLCU.Thisprocessgivesacontent-adaptivecodingtreestructurecomprisedofCUblocks.ThePUisusedonlyfortheCUwhichistheleafnodeinquadtreestructure.InadditiontotheCUandPUdefinitions,thereisanothertransformandquantizationrelatedunit,TU,whosesizecan’texceedthatoftheCU.Theseparationofthreeunitsallowseachunittobeoptimized.What’smore,anotherdistinguisheddifferencebetweenH.264/AVCandHEVCisthatHEVCadoptslargerblockstructurewhichcanbeaslargeas64x64samples.Thispaperproceedfromtwopoints:reducingcomplexityandenhancingcodingefficiencytoimprovethecodingeffectofHEVC.First,thispaperproposesanewfastintramodedecisiontoreducetheencodingtime.ThemainideaisthatwereducethecandidateintramodesforRateDistortionOptimization(RDO)byexploitingthecorrelationbetweencurrentblockanditsneighboringblocks.Second,beforechromaintraprediction,wesmooththereferencesamplesaroundthecurrentblockinordertoremovethenoiseofthereferencesamplesandavoidusingrigiddirectionalpredictiontopredictnotrigidblock.Lastly,weoptimizethecontextmodeloftheintrapredictionmodesinCABACbycountingtheprobabilityof0ineachBin.KeywordsHEVC,intramodedecision,intrasmoothing,entropycoding,contextoptimize-II-哈尔滨工业大学工学硕士学位论文目录摘要..........................................................................................................................IAbstract........................................................................................................................II第1章绪论...............................................................................................................11.1视频编码基础.........................................................................................................11.1.1视频数据冗余..................................................................................................11.1.2视频数据压缩方法..........................................................................................21.2视频编码技术综述.................................................................................................21.2.1预测..................................................................................................................21.2.2变换与量化......................................................................................................31.2.3熵编码..............................................................................................................41.3国内外研究现状.....................................................................................................51.4本文主要研究内容及论文组织.............................................................