HEVC帧内预测参考相邻帧代码解析
2024-06-11 08:10:57
作者：66
（转载请注明出处）
参考链接：http://blog.csdn.net/hevc_cjl/article/details/8200793
亮度分量的帧内预测涉及到的模块比较多，CU->PU的划分，参考像素的填充，哈达玛变换计算STAD（代替率失真），参考预测模式的选择，最佳模式的选择等。
先了解相关的知识：1.CU->PU的划分，在帧内预测中，2Nx2N的CU有两种PU划分方式：2Nx2N，NxN。帧间预测模式有八种：分别为：2Nx2N，NxN，2NxN，Nx2N，nRx2N（左右3:1），nLx2N（左右1:3），2NxnU（上下1:3），2NxnD（上下3:1）。
2. 参考像素的填充：前面说过，HEVC较后h.264增加了左下边界，配合增加的角度模式。
3. 哈达玛变换，WHT（Walsh-Hadamard Transform），广义傅里叶变换的一种，变换矩阵Hm是个大小为2的次幂的矩阵，如下：
图一、hadamard矩阵
这种矩阵的特性主要有，元素+1、-1，计算方便；正交对称；奇行（列）偶对称，偶行（列）奇对称；变换前后能量守恒；因残差信号经WHT后的求绝对值和与DCT之后系数绝对值之和接近，可以作为视频编码中的快速模式选择。
4. 参考预测模式的选择：
图二、参考PU与当前PU的位置关系
HEVC中共35种亮度预测模式，包括planar（0）、DC（1）和33种角度模式（2~34）。
如上图，帧内预测参考其他PU，并建立候选列表candModeList[3]，三个备选项。
选择的规则如下：
①modeA预测模式与modeB相同
1. 同为planar或DC
candModeList[3] = {planar, DC, 角度26（垂直）}
2. 同为角度模式
candModeList[3] = {modeA, modeA-1, modeA+1}
注：mode2相邻为3和32，mode34相邻为33和3。
②modeA与modeB不同
candModeList = {modeA, modeB, choice}
关于choice的选择：
1. modeA与modeB都不为planar
choice = planar
2. modeA与modeB都不为DC
choice = DC
3. 当前两个都不符合时,
choice = 26
在candModelist建立后，对当前PU信息进行编码，经过率失真遍历计算后，
①最优预测模式modeC在列表中，仅编码模式在列表中的位置
②若不在，依次进行如下步骤：
1. candModeList中的候选模式从小到大重排
2. 比较最优预测模式modeC与列表中每个的标号大小，若modeC >= candModeList[i]，则令modeC = modeC - 1，遍历结束后，对modeC最终值进行编码。
整体上看，思路为：
当前CU划分为PU ->遍历每个PU ->对每个PU计算35个模式中的最优模式，与预测对比，最后确定最佳预测模式。
详细情况见代码，目前对代码还存有疑问，之后搞清楚了来修改：
//亮度分量帧内预测VoidTEncSearch::estIntraPredQT( TComDataCU* pcCU,TComYuv* pcOrgYuv,TComYuv* pcPredYuv,TComYuv* pcResiYuv,TComYuv* pcRecoYuv,UInt& ruiDistC,Bool bLumaOnly ){UInt uiDepth = pcCU->getDepth(0);//当前CU的深度//uiNumPU,划分后pu的个数,PU的划分模式，帧内有2种：2NX2N,NxN;帧间有8种：4种对称模式：2Nx2N,2NxN,Nx2N,NxN,四种非对称模式,2NxnU（上下1：3）,2NxnD（上下3：1）,nLx2N（左右1：3）,nRx2N（左右3：1）。帧间还有一种skip模式，即不需要编码残差信息时。UInt uiNumPU = pcCU->getNumPartitions();//当前cu划分为pu的数目UInt uiInitTrDepth = pcCU->getPartitionSize(0) == SIZE_2Nx2N ? 0 : 1;//计算变换深度，实际为uiDepthUInt uiWidth = pcCU->getWidth (0) >> uiInitTrDepth;//当前cu的宽度UInt uiHeight = pcCU->getHeight(0) >> uiInitTrDepth;//当前cu的长度UInt uiQNumParts = pcCU->getTotalNumPart() >> 2;//当前cu包含的最小分区4x4的数目。UInt uiWidthBit = pcCU->getIntraSizeIdx(0);UInt uiOverallDistY = 0;UInt uiOverallDistC = 0;UInt CandNum;Double CandCostList[ FAST_UDI_MAX_RDMODE_NUM ];//===== set QP and clear Cbf =====if ( pcCU->getSlice()->getPPS()->getUseDQP() == true){pcCU->setQPSubParts( pcCU->getQP(0), 0, uiDepth );}else{pcCU->setQPSubParts( pcCU->getSlice()->getSliceQp(), 0, uiDepth );}//===== loop over partitions =====遍历UInt uiPartOffset = 0;//记录当前pu的Zorder坐标for( UInt uiPU = 0; uiPU < uiNumPU; uiPU++, uiPartOffset += uiQNumParts ){//===== init pattern for luma prediction =====Bool bAboveAvail = false;//Bool bLeftAvail = false;pcCU->getPattern()->initPattern ( pcCU, uiInitTrDepth, uiPartOffset );//获取当前PU邻域的可用性，对参考像素进行滤波,代码里的宽长都为当前cu的宽长，但是pu与tu的划分是以depth为基础的隐式划分，名字上仍以Cu表示，实际此Cu已经代表了PU或TU VpcCU->getPattern()->initAdiPattern( pcCU, uiPartOffset, uiInitTrDepth, m_piYuvExt, m_iYuvExtStride, m_iYuvExtHeight, bAboveAvail, bLeftAvail );//===== determine set of modes to be tested (using prediction signal only) =====Int numModesAvailable = 35; //total number of Intra modesPel* piOrg = pcOrgYuv ->getLumaAddr( uiPU, uiWidth );Pel* piPred = pcPredYuv->getLumaAddr( uiPU, uiWidth );UInt uiStride = pcPredYuv->getStride();UInt uiRdModeList[FAST_UDI_MAX_RDMODE_NUM];Int numModesForFullRD = g_aucIntraModeNumFast[ uiWidthBit ];//MPM的数目,像是候选预测模式数目// g_aucIntraModeNumFast[]={3,8,8,3,3,3,3};2x2,4x4,8x8,16x16,32x32,64x64,128x128Bool doFastSearch = (numModesForFullRD != numModesAvailable);//这里doFastSearch恒为真if (doFastSearch)//此时是肯定会进入{assert(numModesForFullRD < numModesAvailable);//确定numModesForFullRD < numModesAvailablefor( Int i=0; i < numModesForFullRD; i++ ){CandCostList[ i ] = MAX_DOUBLE;//初始化率失真表，全部为最大值，方便后面比较。}CandNum = 0;for( Int modeIdx = 0; modeIdx < numModesAvailable; modeIdx++ )//遍历35中预测模式{UInt uiMode = modeIdx;//调用亮度帧内预测函数 VpredIntraLumaAng( pcCU->getPattern(), uiMode, piPred, uiStride, uiWidth, uiHeight, bAboveAvail, bLeftAvail );// use hadamard transform here哈达玛矩阵计算率失真UInt uiSad = m_pcRdCost->calcHAD(g_bitDepthY, piOrg, uiStride, piPred, uiStride, uiWidth, uiHeight );//计算SATD（残差经HAD后的绝对值总和）UInt iModeBits = xModeBitsIntra( pcCU, uiMode, uiPU, uiPartOffset, uiDepth, uiInitTrDepth );//计算编码当面所需的bitsDouble cost = (Double)uiSad + (Double)iModeBits * m_pcRdCost->getSqrtLambda();//率失真代价//iModeBits编码当前模式需要的bits，CandNum += xUpdateCandList( uiMode, cost, numModesForFullRD, uiRdModeList, CandCostList );//帧内预测模式候选列表}#if FAST_UDI_USE_MPMInt uiPreds[3] = {-1, -1, -1};Int iMode = -1;//分两种情况，如果前两个相同iMode=1，否则iMode=2Int numCand = pcCU->getIntraDirLumaPredictor( uiPartOffset, uiPreds, &iMode );//获取亮度预测的前三个MPMsif( iMode >= 0 ){numCand = iMode;}for( Int j=0; j < numCand; j++){Bool mostProbableModeIncluded = false;Int mostProbableMode = uiPreds[j];//取出预测的MPMfor( Int i=0; i < numModesForFullRD; i++){mostProbableModeIncluded |= (mostProbableMode == uiRdModeList[i]);//检查MPMs，是否被uiRdModeList包含}if (!mostProbableModeIncluded)//若没被包含，则将该MPM包含进uiRdModeList里{uiRdModeList[numModesForFullRD++] = mostProbableMode;//计算率失真的备选模式表}}#endif // FAST_UDI_USE_MPM}else{for( Int i=0; i < numModesForFullRD; i++){uiRdModeList[i] = i;}}//check modes 确定帧内预测模式的最佳值主要有以下几个步骤：//1. 对numModesForFullRD中预测模式进行遍历，算出RDcosts，但至多对depth=1的CU进行遍历，提高了速度。//2.得到最优，有可能包括次优的两个。//3.最佳模式下的分割模式遍历，以得最优结果。//===== check modes (using r-d costs) =====#if HHI_RQT_INTRA_SPEEDUP_MODUInt uiSecondBestMode = MAX_UINT;Double dSecondBestPUCost = MAX_DOUBLE;#endifUInt uiBestPUMode = 0;//最佳预测模式UInt uiBestPUDistY = 0;//最佳预测模式对应的亮度失真UInt uiBestPUDistC = 0;//最佳预测模式色度失真Double dBestPUCost = MAX_DOUBLE;//RDcostsfor( UInt uiMode = 0; uiMode < numModesForFullRD; uiMode++ ){// set luma prediction modeUInt uiOrgMode = uiRdModeList[uiMode];pcCU->setLumaIntraDirSubParts ( uiOrgMode, uiPartOffset, uiDepth + uiInitTrDepth );// set context modelsm_pcRDGoOnSbacCoder->load( m_pppcRDSbacCoder[uiDepth][CI_CURR_BEST] );// determine residual for partitionUInt uiPUDistY = 0;//当前预测模式的亮度失真UInt uiPUDistC = 0;//当前色度失真Double dPUCost = 0.0;//当前预测RDcost#if HHI_RQT_INTRA_SPEEDUPxRecurIntraCodingQT( pcCU, uiInitTrDepth, uiPartOffset, bLumaOnly, pcOrgYuv, pcPredYuv, pcResiYuv, uiPUDistY, uiPUDistC, true, dPUCost );#elsexRecurIntraCodingQT( pcCU, uiInitTrDepth, uiPartOffset, bLumaOnly, pcOrgYuv, pcPredYuv, pcResiYuv, uiPUDistY, uiPUDistC, dPUCost );#endif// check r-d costif( dPUCost < dBestPUCost )//更新最佳预测模式相关参数{#if HHI_RQT_INTRA_SPEEDUP_MOD//次优模式uiSecondBestMode = uiBestPUMode;dSecondBestPUCost = dBestPUCost;#endifuiBestPUMode = uiOrgMode;uiBestPUDistY = uiPUDistY;uiBestPUDistC = uiPUDistC;dBestPUCost = dPUCost;xSetIntraResultQT( pcCU, uiInitTrDepth, uiPartOffset, bLumaOnly, pcRecoYuv );UInt uiQPartNum = pcCU->getPic()->getNumPartInCU() >> ( ( pcCU->getDepth(0) + uiInitTrDepth ) << 1 );::memcpy( m_puhQTTempTrIdx, pcCU->getTransformIdx() + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[0], pcCU->getCbf( TEXT_LUMA ) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[1], pcCU->getCbf( TEXT_CHROMA_U ) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[2], pcCU->getCbf( TEXT_CHROMA_V ) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[0], pcCU->getTransformSkip(TEXT_LUMA) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[1], pcCU->getTransformSkip(TEXT_CHROMA_U) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[2], pcCU->getTransformSkip(TEXT_CHROMA_V) + uiPartOffset, uiQPartNum * sizeof( UChar ) );}#if HHI_RQT_INTRA_SPEEDUP_MODelse if( dPUCost < dSecondBestPUCost ){uiSecondBestMode = uiOrgMode;dSecondBestPUCost = dPUCost;}#endif} // Mode loop#if HHI_RQT_INTRA_SPEEDUP#if HHI_RQT_INTRA_SPEEDUP_MODfor( UInt ui =0; ui < 2; ++ui )#endif{#if HHI_RQT_INTRA_SPEEDUP_MODUInt uiOrgMode = ui ? uiSecondBestMode : uiBestPUMode;if( uiOrgMode == MAX_UINT ){break;}#elseUInt uiOrgMode = uiBestPUMode;//设置为最佳模式#endifpcCU->setLumaIntraDirSubParts ( uiOrgMode, uiPartOffset, uiDepth + uiInitTrDepth );// set context modelsm_pcRDGoOnSbacCoder->load( m_pppcRDSbacCoder[uiDepth][CI_CURR_BEST] );// determine residual for partitionUInt uiPUDistY = 0;UInt uiPUDistC = 0;Double dPUCost = 0.0;xRecurIntraCodingQT( pcCU, uiInitTrDepth, uiPartOffset, bLumaOnly, pcOrgYuv, pcPredYuv, pcResiYuv, uiPUDistY, uiPUDistC, false, dPUCost );//此时倒数第二个参数为false// check r-d costif( dPUCost < dBestPUCost ){uiBestPUMode = uiOrgMode;uiBestPUDistY = uiPUDistY;uiBestPUDistC = uiPUDistC;dBestPUCost = dPUCost;xSetIntraResultQT( pcCU, uiInitTrDepth, uiPartOffset, bLumaOnly, pcRecoYuv );UInt uiQPartNum = pcCU->getPic()->getNumPartInCU() >> ( ( pcCU->getDepth(0) + uiInitTrDepth ) << 1 );::memcpy( m_puhQTTempTrIdx, pcCU->getTransformIdx() + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[0], pcCU->getCbf( TEXT_LUMA ) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[1], pcCU->getCbf( TEXT_CHROMA_U ) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[2], pcCU->getCbf( TEXT_CHROMA_V ) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[0], pcCU->getTransformSkip(TEXT_LUMA) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[1], pcCU->getTransformSkip(TEXT_CHROMA_U) + uiPartOffset, uiQPartNum * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[2], pcCU->getTransformSkip(TEXT_CHROMA_V) + uiPartOffset, uiQPartNum * sizeof( UChar ) );}} // Mode loop#endif//--- update overall distortion ---uiOverallDistY += uiBestPUDistY;uiOverallDistC += uiBestPUDistC;//--- update transform index and cbf ---UInt uiQPartNum = pcCU->getPic()->getNumPartInCU() >> ( ( pcCU->getDepth(0) + uiInitTrDepth ) << 1 );::memcpy( pcCU->getTransformIdx() + uiPartOffset, m_puhQTTempTrIdx, uiQPartNum * sizeof( UChar ) );::memcpy( pcCU->getCbf( TEXT_LUMA ) + uiPartOffset, m_puhQTTempCbf[0], uiQPartNum * sizeof( UChar ) );::memcpy( pcCU->getCbf( TEXT_CHROMA_U ) + uiPartOffset, m_puhQTTempCbf[1], uiQPartNum * sizeof( UChar ) );::memcpy( pcCU->getCbf( TEXT_CHROMA_V ) + uiPartOffset, m_puhQTTempCbf[2], uiQPartNum * sizeof( UChar ) );::memcpy( pcCU->getTransformSkip(TEXT_LUMA) + uiPartOffset, m_puhQTTempTransformSkipFlag[0], uiQPartNum * sizeof( UChar ) );::memcpy( pcCU->getTransformSkip(TEXT_CHROMA_U) + uiPartOffset, m_puhQTTempTransformSkipFlag[1], uiQPartNum * sizeof( UChar ) );::memcpy( pcCU->getTransformSkip(TEXT_CHROMA_V) + uiPartOffset, m_puhQTTempTransformSkipFlag[2], uiQPartNum * sizeof( UChar ) );//--- set reconstruction for next intra prediction blocks ---if( uiPU != uiNumPU - 1 ){Bool bSkipChroma = false;Bool bChromaSame = false;UInt uiLog2TrSize = g_aucConvertToBit[ pcCU->getSlice()->getSPS()->getMaxCUWidth() >> ( pcCU->getDepth(0) + uiInitTrDepth ) ] + 2;if( !bLumaOnly && uiLog2TrSize == 2 ){assert( uiInitTrDepth > 0 );bSkipChroma = ( uiPU != 0 );bChromaSame = true;}UInt uiCompWidth = pcCU->getWidth ( 0 ) >> uiInitTrDepth;UInt uiCompHeight = pcCU->getHeight( 0 ) >> uiInitTrDepth;UInt uiZOrder = pcCU->getZorderIdxInCU() + uiPartOffset;Pel* piDes = pcCU->getPic()->getPicYuvRec()->getLumaAddr( pcCU->getAddr(), uiZOrder );UInt uiDesStride = pcCU->getPic()->getPicYuvRec()->getStride();Pel* piSrc = pcRecoYuv->getLumaAddr( uiPartOffset );UInt uiSrcStride = pcRecoYuv->getStride();for( UInt uiY = 0; uiY < uiCompHeight; uiY++, piSrc += uiSrcStride, piDes += uiDesStride ){for( UInt uiX = 0; uiX < uiCompWidth; uiX++ ){piDes[ uiX ] = piSrc[ uiX ];}}if( !bLumaOnly && !bSkipChroma ){if( !bChromaSame ){uiCompWidth >>= 1;uiCompHeight >>= 1;}piDes = pcCU->getPic()->getPicYuvRec()->getCbAddr( pcCU->getAddr(), uiZOrder );uiDesStride = pcCU->getPic()->getPicYuvRec()->getCStride();piSrc = pcRecoYuv->getCbAddr( uiPartOffset );uiSrcStride = pcRecoYuv->getCStride();for( UInt uiY = 0; uiY < uiCompHeight; uiY++, piSrc += uiSrcStride, piDes += uiDesStride ){for( UInt uiX = 0; uiX < uiCompWidth; uiX++ ){piDes[ uiX ] = piSrc[ uiX ];}}piDes = pcCU->getPic()->getPicYuvRec()->getCrAddr( pcCU->getAddr(), uiZOrder );piSrc = pcRecoYuv->getCrAddr( uiPartOffset );for( UInt uiY = 0; uiY < uiCompHeight; uiY++, piSrc += uiSrcStride, piDes += uiDesStride ){for( UInt uiX = 0; uiX < uiCompWidth; uiX++ ){piDes[ uiX ] = piSrc[ uiX ];}}}}//=== update PU data ====pcCU->setLumaIntraDirSubParts ( uiBestPUMode, uiPartOffset, uiDepth + uiInitTrDepth );pcCU->copyToPic ( uiDepth, uiPU, uiInitTrDepth );} // PU loopif( uiNumPU > 1 ){ // set Cbf for all blocksUInt uiCombCbfY = 0;UInt uiCombCbfU = 0;UInt uiCombCbfV = 0;UInt uiPartIdx = 0;for( UInt uiPart = 0; uiPart < 4; uiPart++, uiPartIdx += uiQNumParts ){uiCombCbfY |= pcCU->getCbf( uiPartIdx, TEXT_LUMA, 1 );uiCombCbfU |= pcCU->getCbf( uiPartIdx, TEXT_CHROMA_U, 1 );uiCombCbfV |= pcCU->getCbf( uiPartIdx, TEXT_CHROMA_V, 1 );}for( UInt uiOffs = 0; uiOffs < 4 * uiQNumParts; uiOffs++ ){pcCU->getCbf( TEXT_LUMA )[ uiOffs ] |= uiCombCbfY;pcCU->getCbf( TEXT_CHROMA_U )[ uiOffs ] |= uiCombCbfU;pcCU->getCbf( TEXT_CHROMA_V )[ uiOffs ] |= uiCombCbfV;}}//===== reset context models =====m_pcRDGoOnSbacCoder->load(m_pppcRDSbacCoder[uiDepth][CI_CURR_BEST]);//===== set distortion (rate and r-d costs are determined later) =====ruiDistC = uiOverallDistC;pcCU->getTotalDistortion() = uiOverallDistY + uiOverallDistC;}VoidTEncSearch::estIntraPredChromaQT( TComDataCU* pcCU,TComYuv* pcOrgYuv,TComYuv* pcPredYuv,TComYuv* pcResiYuv,TComYuv* pcRecoYuv,UInt uiPreCalcDistC ){UInt uiDepth = pcCU->getDepth(0);UInt uiBestMode = 0;UInt uiBestDist = 0;Double dBestCost = MAX_DOUBLE;//----- init mode list -----UInt uiMinMode = 0;UInt uiModeList[ NUM_CHROMA_MODE ];pcCU->getAllowedChromaDir( 0, uiModeList );UInt uiMaxMode = NUM_CHROMA_MODE;//----- check chroma modes -----for( UInt uiMode = uiMinMode; uiMode < uiMaxMode; uiMode++ ){//----- restore context models -----m_pcRDGoOnSbacCoder->load( m_pppcRDSbacCoder[uiDepth][CI_CURR_BEST] );//----- chroma coding -----UInt uiDist = 0;pcCU->setChromIntraDirSubParts ( uiModeList[uiMode], 0, uiDepth );xRecurIntraChromaCodingQT ( pcCU, 0, 0, pcOrgYuv, pcPredYuv, pcResiYuv, uiDist );if( pcCU->getSlice()->getPPS()->getUseTransformSkip() ){m_pcRDGoOnSbacCoder->load( m_pppcRDSbacCoder[uiDepth][CI_CURR_BEST] );}UInt uiBits = xGetIntraBitsQT( pcCU, 0, 0, false, true, false );Double dCost = m_pcRdCost->calcRdCost( uiBits, uiDist );//----- compare -----if( dCost < dBestCost ){dBestCost = dCost;uiBestDist = uiDist;uiBestMode = uiModeList[uiMode];UInt uiQPN = pcCU->getPic()->getNumPartInCU() >> ( uiDepth << 1 );xSetIntraResultChromaQT( pcCU, 0, 0, pcRecoYuv );::memcpy( m_puhQTTempCbf[1], pcCU->getCbf( TEXT_CHROMA_U ), uiQPN * sizeof( UChar ) );::memcpy( m_puhQTTempCbf[2], pcCU->getCbf( TEXT_CHROMA_V ), uiQPN * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[1], pcCU->getTransformSkip( TEXT_CHROMA_U ), uiQPN * sizeof( UChar ) );::memcpy( m_puhQTTempTransformSkipFlag[2], pcCU->getTransformSkip( TEXT_CHROMA_V ), uiQPN * sizeof( UChar ) );}}//----- set data -----UInt uiQPN = pcCU->getPic()->getNumPartInCU() >> ( uiDepth << 1 );::memcpy( pcCU->getCbf( TEXT_CHROMA_U ), m_puhQTTempCbf[1], uiQPN * sizeof( UChar ) );::memcpy( pcCU->getCbf( TEXT_CHROMA_V ), m_puhQTTempCbf[2], uiQPN * sizeof( UChar ) );::memcpy( pcCU->getTransformSkip( TEXT_CHROMA_U ), m_puhQTTempTransformSkipFlag[1], uiQPN * sizeof( UChar ) );::memcpy( pcCU->getTransformSkip( TEXT_CHROMA_V ), m_puhQTTempTransformSkipFlag[2], uiQPN * sizeof( UChar ) );pcCU->setChromIntraDirSubParts( uiBestMode, 0, uiDepth );pcCU->getTotalDistortion () += uiBestDist - uiPreCalcDistC;//----- restore context models -----m_pcRDGoOnSbacCoder->load( m_pppcRDSbacCoder[uiDepth][CI_CURR_BEST] );}HEVC帧内预测参考相邻帧代码解析相关推荐
- HEVC帧内预测参考像素检测获取和滤波
作者:66 (转载请注明出处) 还是参考HEVC_CJL的博客,理论都清楚,跟着他的进度看代码,感谢前辈的分享,他的代码里没有强滤波过程,在此我稍加补充. 原文链接:http://blog.csdn. ...
- 关于帧内预测模式的视频隐写代码介绍
关于帧内预测模式的视频隐写代码介绍 前言 一.H.265/HEVC的帧内预测过程 二.论文[1]的介绍以及如何复现 前言 在早期的基于帧内预测模式(IPM)的H.265/HEVC视频中,大多是基于自定 ...
- HEVC算法和体系结构：预测编码之帧内预测
预测编码之帧内预测(Intra-Picture Prediction) 预测编码(Prediction Coding)是视频编码的核心技术之一,指利用已编码的一个或几个样本值,根据某种模型或方法,对当 ...
- H.266/VVC帧间预测技术学习:帧间和帧内联合预测(Combined inter and intra prediction, CIIP)
在HEVC中一个CU在预测时要么使用帧内预测要么使用帧间预测,二者只能取其一.而VVC中提出的CIIP技术,是将帧间预测信号与帧内预测信号相结合. 在VVC中,当CU以Merge模式编码时,且CU包含 ...
- SVAC1.0帧内预测技术分析
##Date:2017/10/21 ##Content:SVAC1.0帧内预测技术分析 帧内预测主要是利用视频图像的空域相关性来进行预测编码.通常I帧只采用帧内预测进行编码.利用宏块之间的相关性,对当 ...
- DCC2022：高阶帧内预测
本文来自DCC2022论文<High-order Intra Prediction for Future Video Coding> 帧内预测 帧内预测是去除空域冗余的重要工具,它从上方或 ...
- 【H2645】帧内预测
1.帧内预测的原理 帧内预测的原理:预测值是该像素周围像素值加权求和(比如平均值)P,它和实际值相减后得到的差值q,如果差值q很小,说明该像素的值可以通过预测得出,可以丢弃了,这就达到压缩编码的目的. ...
- HEVC亮度分量帧内预测模式代码详解
作者:66 (转载请注明出处) 从我自身的学习历程来看,建议大家先把理论扎实了再去理解代码,代码里也有注释,与理论的地方一对应加上一点C编程基础很容易就能理解. 我们先了解相关理论,此处可参考这一篇博 ...
- 【HEVC代码阅读】帧内预测
HEVC的帧内预测的架构分为三个步骤: ①构建参考像素数组:②生成预测像素:③后处理操作. HEVC标准将这三个步骤进行了精密设计,以求达到较高的编码效率,同时降低编码和解码端的运算要求.HEVC标准 ...
最新文章
- python库缺少pkg_resource_ImportError: No module named pkg_resources解决方案
- php自动轮播图代码,JavaScript如何实现动态轮播图效果？（代码示例）
- SAP RETAIL 如何确定自动补货触发的单据类型 II
- 车载全景可视系统surroundview
- 初步了解BIM模型和超图相关操作
- 查看、关闭当前服务器上启动服务 / 进程
- AI深入应用，生态越加开放，开发者的机会在哪里？
- 气愤ing，身份证丢了到底要不要挂失？
- 高性能Mysql数据库表设计原则
- Android启动的init进程
- 戴尔计算机没有硬盘驱动,如果Dell计算机无法从硬盘驱动器启动怎么办
- 蔡勒公式与Python
- 硅谷真假u盘测试软件,硅谷硅谷真假u盘测试
- 计算机在哪里修改储存默认磁盘,Win7资源管理器修改默认显示磁盘小教程
- 接口测试二（App抓包）
- 荣耀magic book忘记pin，只能用pin登录，还着急用电脑里的文件
- 路由与交换技术-18-热备份路由选择协议HSRP
- python二进制格式追加_Python追加/填充二进制文件添加garbag
- Regsvr32注册DLL和Regasm
- 计算机系统英语作文结尾万能句子,英语作文结尾万能句型整理