EncGOP.cpp

    {
      msg( NOTICE, "\nPSNR2: [Y %6.4lf dB    U %6.4lf dB    V %6.4lf dB]", upscaledPSNR[COMPONENT_Y], upscaledPSNR[COMPONENT_Cb], upscaledPSNR[COMPONENT_Cr] );
    }
  }
  else if( g_verbosity >= INFO )
  {
    std::cout << "\r\t" << pcSlice->getPOC();
    std::cout.flush();
  }
}

#if JVET_O0756_CALCULATE_HDRMETRICS
void EncGOP::xCalculateHDRMetrics( Picture* pcPic, double deltaE[hdrtoolslib::NB_REF_WHITE], double psnrL[hdrtoolslib::NB_REF_WHITE])
{
  copyBuftoFrame(pcPic);

  ChromaFormat chFmt =  pcPic->chromaFormat;

  if (chFmt != CHROMA_444)
  {
    m_pcConvertFormat->process(m_ppcFrameOrg[1], m_ppcFrameOrg[0]);
    m_pcConvertFormat->process(m_ppcFrameRec[1], m_ppcFrameRec[0]);
  }

  m_pcConvertIQuantize->process(m_ppcFrameOrg[2], m_ppcFrameOrg[1]);
  m_pcConvertIQuantize->process(m_ppcFrameRec[2], m_ppcFrameRec[1]);

  m_pcColorTransform->process(m_ppcFrameOrg[3], m_ppcFrameOrg[2]);
  m_pcColorTransform->process(m_ppcFrameRec[3], m_ppcFrameRec[2]);

  m_pcTransferFct->forward(m_ppcFrameOrg[4], m_ppcFrameOrg[3]);
  m_pcTransferFct->forward(m_ppcFrameRec[4], m_ppcFrameRec[3]);

  // Calculate the Metrics
  m_pcDistortionDeltaE->computeMetric(m_ppcFrameOrg[4], m_ppcFrameRec[4]);

  *deltaE = m_pcDistortionDeltaE->getDeltaE();
  *psnrL  = m_pcDistortionDeltaE->getPsnrL();

}

void EncGOP::copyBuftoFrame( Picture* pcPic )
{
  int cropOffsetLeft   = m_pcCfg->getCropOffsetLeft();
  int cropOffsetTop    = m_pcCfg->getCropOffsetTop();
  int cropOffsetRight  = m_pcCfg->getCropOffsetRight();
  int cropOffsetBottom = m_pcCfg->getCropOffsetBottom();

  int height = pcPic->getOrigBuf(COMPONENT_Y).height - cropOffsetLeft + cropOffsetRight;
  int width = pcPic->getOrigBuf(COMPONENT_Y).width - cropOffsetTop + cropOffsetBottom;

  ChromaFormat chFmt =  pcPic->chromaFormat;

  Pel* pOrg = pcPic->getOrigBuf(COMPONENT_Y).buf;
  Pel* pRec = pcPic->getRecoBuf(COMPONENT_Y).buf;

  uint16_t* yOrg = m_ppcFrameOrg[0]->m_ui16Comp[hdrtoolslib::Y_COMP];
  uint16_t* yRec = m_ppcFrameRec[0]->m_ui16Comp[hdrtoolslib::Y_COMP];
  uint16_t* uOrg = m_ppcFrameOrg[0]->m_ui16Comp[hdrtoolslib::Cb_COMP];
  uint16_t* uRec = m_ppcFrameRec[0]->m_ui16Comp[hdrtoolslib::Cb_COMP];
  uint16_t* vOrg = m_ppcFrameOrg[0]->m_ui16Comp[hdrtoolslib::Cr_COMP];
  uint16_t* vRec = m_ppcFrameRec[0]->m_ui16Comp[hdrtoolslib::Cr_COMP];

  if(chFmt == CHROMA_444){
    yOrg = m_ppcFrameOrg[1]->m_ui16Comp[hdrtoolslib::Y_COMP];
    yRec = m_ppcFrameRec[1]->m_ui16Comp[hdrtoolslib::Y_COMP];
    uOrg = m_ppcFrameOrg[1]->m_ui16Comp[hdrtoolslib::Cb_COMP];
    uRec = m_ppcFrameRec[1]->m_ui16Comp[hdrtoolslib::Cb_COMP];
    vOrg = m_ppcFrameOrg[1]->m_ui16Comp[hdrtoolslib::Cr_COMP];
    vRec = m_ppcFrameRec[1]->m_ui16Comp[hdrtoolslib::Cr_COMP];
  }

  for (int i = 0; i < height; i++) {
    for (int j = 0; j < width; j++) {
      yOrg[i*width + j] = static_cast<uint16_t>(pOrg[(i + cropOffsetTop) * pcPic->getOrigBuf(COMPONENT_Y).stride + j + cropOffsetLeft]);
      yRec[i*width + j] = static_cast<uint16_t>(pRec[(i + cropOffsetTop) * pcPic->getRecoBuf(COMPONENT_Y).stride + j + cropOffsetLeft]);
    }
  }

  if (chFmt != CHROMA_444) {
    height >>= 1;
    width  >>= 1;
    cropOffsetLeft >>= 1;
    cropOffsetTop >>= 1;
  }

  pOrg = pcPic->getOrigBuf(COMPONENT_Cb).buf;
  pRec = pcPic->getRecoBuf(COMPONENT_Cb).buf;

  for (int i = 0; i < height; i++) {
    for (int j = 0; j < width; j++) {
      uOrg[i*width + j] = static_cast<uint16_t>(pOrg[(i + cropOffsetTop) * pcPic->getOrigBuf(COMPONENT_Cb).stride + j + cropOffsetLeft]);
      uRec[i*width + j] = static_cast<uint16_t>(pRec[(i + cropOffsetTop) * pcPic->getRecoBuf(COMPONENT_Cb).stride + j + cropOffsetLeft]);
    }
  }

  pOrg = pcPic->getOrigBuf(COMPONENT_Cr).buf;
  pRec = pcPic->getRecoBuf(COMPONENT_Cr).buf;

  for (int i = 0; i < height; i++) {
    for (int j = 0; j < width; j++) {
      vOrg[i*width + j] = static_cast<uint16_t>(pOrg[(i + cropOffsetTop) * pcPic->getOrigBuf(COMPONENT_Cr).stride + j + cropOffsetLeft]);
      vRec[i*width + j] = static_cast<uint16_t>(pRec[(i + cropOffsetTop) * pcPic->getRecoBuf(COMPONENT_Cr).stride + j + cropOffsetLeft]);
    }
  }
}
#endif

void EncGOP::xCalculateInterlacedAddPSNR( Picture* pcPicOrgFirstField, Picture* pcPicOrgSecondField,
                                          PelUnitBuf cPicRecFirstField, PelUnitBuf cPicRecSecondField,
                                          const InputColourSpaceConversion conversion, const bool printFrameMSE, double* PSNR_Y
                                        , bool isEncodeLtRef
)
{
  const SPS &sps = *pcPicOrgFirstField->cs->sps;
  const ChromaFormat format = sps.getChromaFormatIdc();
  double  dPSNR[MAX_NUM_COMPONENT];
  Picture    *apcPicOrgFields[2] = {pcPicOrgFirstField, pcPicOrgSecondField};
  PelUnitBuf acPicRecFields[2]   = {cPicRecFirstField, cPicRecSecondField};
#if ENABLE_QPA
  const bool    useWPSNR = m_pcEncLib->getUseWPSNR();
#endif
  for(int i=0; i<MAX_NUM_COMPONENT; i++)
  {
    dPSNR[i]=0.0;
  }

  PelStorage cscd[2 /* first/second field */];
  if (conversion!=IPCOLOURSPACE_UNCHANGED)
  {
    for(uint32_t fieldNum=0; fieldNum<2; fieldNum++)
    {
      PelUnitBuf& reconField= (acPicRecFields[fieldNum]);
      cscd[fieldNum].create( reconField.chromaFormat, Area( Position(), reconField.Y()) );
      VideoIOYuv::ColourSpaceConvert(reconField, cscd[fieldNum], conversion, false);
      acPicRecFields[fieldNum]=cscd[fieldNum];
    }
  }

  //===== calculate PSNR =====
  double MSEyuvframe[MAX_NUM_COMPONENT] = {0, 0, 0};

  CHECK(!(acPicRecFields[0].chromaFormat==acPicRecFields[1].chromaFormat), "Unspecified error");
  const uint32_t numValidComponents = ::getNumberValidComponents( acPicRecFields[0].chromaFormat );

  for (int chan = 0; chan < numValidComponents; chan++)
  {
    const ComponentID ch=ComponentID(chan);
    CHECK(!(acPicRecFields[0].get(ch).width==acPicRecFields[1].get(ch).width), "Unspecified error");
    CHECK(!(acPicRecFields[0].get(ch).height==acPicRecFields[0].get(ch).height), "Unspecified error");

    uint64_t uiSSDtemp=0;
    const uint32_t width    = acPicRecFields[0].get(ch).width - (m_pcEncLib->getPad(0) >> ::getComponentScaleX(ch, format));
    const uint32_t height   = acPicRecFields[0].get(ch).height - ((m_pcEncLib->getPad(1) >> 1) >> ::getComponentScaleY(ch, format));
    const uint32_t bitDepth = sps.getBitDepth(toChannelType(ch));

    for(uint32_t fieldNum=0; fieldNum<2; fieldNum++)
    {
      CHECK(!(conversion == IPCOLOURSPACE_UNCHANGED), "Unspecified error");
#if ENABLE_QPA
      uiSSDtemp += xFindDistortionPlane( acPicRecFields[fieldNum].get(ch), apcPicOrgFields[fieldNum]->getOrigBuf().get(ch), useWPSNR ? bitDepth : 0, ::getComponentScaleX(ch, format) );
#else
      uiSSDtemp += xFindDistortionPlane( acPicRecFields[fieldNum].get(ch), apcPicOrgFields[fieldNum]->getOrigBuf().get(ch), 0 );
#endif
    }
    const uint32_t maxval = 255 << (bitDepth - 8);
    const uint32_t size   = width * height * 2;
    const double fRefValue = (double)maxval * maxval * size;
    dPSNR[ch]         = uiSSDtemp ? 10.0 * log10(fRefValue / (double)uiSSDtemp) : 999.99;
    MSEyuvframe[ch]   = (double)uiSSDtemp / size;
  }

  uint32_t uibits = 0; // the number of bits for the pair is not calculated here - instead the overall total is used elsewhere.

  //===== add PSNR =====
  m_gcAnalyzeAll_in.addResult (dPSNR, (double)uibits, MSEyuvframe
    , MSEyuvframe
    , isEncodeLtRef
  );

  *PSNR_Y = dPSNR[COMPONENT_Y];

  msg( DETAILS, "\n                                      Interlaced frame %d: [Y %6.4lf dB    U %6.4lf dB    V %6.4lf dB]", pcPicOrgSecondField->getPOC()/2 , dPSNR[COMPONENT_Y], dPSNR[COMPONENT_Cb], dPSNR[COMPONENT_Cr] );
  if (printFrameMSE)
  {
    msg( DETAILS, " [Y MSE %6.4lf  U MSE %6.4lf  V MSE %6.4lf]", MSEyuvframe[COMPONENT_Y], MSEyuvframe[COMPONENT_Cb], MSEyuvframe[COMPONENT_Cr] );
  }

  for(uint32_t fieldNum=0; fieldNum<2; fieldNum++)
  {
    cscd[fieldNum].destroy();
  }
}

/** Function for deciding the nal_unit_type.
 * \param pocCurr POC of the current picture
 * \param lastIDR  POC of the last IDR picture
 * \param isField  true to indicate field coding
 * \returns the NAL unit type of the picture
 * This function checks the configuration and returns the appropriate nal_unit_type for the picture.
 */
NalUnitType EncGOP::getNalUnitType(int pocCurr, int lastIDR, bool isField)
{
  if (pocCurr == 0)
  {
    return NAL_UNIT_CODED_SLICE_IDR_N_LP;
  }

  if (m_pcCfg->getEfficientFieldIRAPEnabled() && isField && pocCurr == (m_pcCfg->getUseCompositeRef() ? 2: 1))
  {
    // to avoid the picture becoming an IRAP
    return NAL_UNIT_CODED_SLICE_TRAIL;
  }

  if (m_pcCfg->getDecodingRefreshType() != 3 && (pocCurr - isField) % (m_pcCfg->getIntraPeriod() * (m_pcCfg->getUseCompositeRef() ? 2 : 1)) == 0)
  {
    if (m_pcCfg->getDecodingRefreshType() == 1)
    {
      return NAL_UNIT_CODED_SLICE_CRA;
    }
    else if (m_pcCfg->getDecodingRefreshType() == 2)
    {
      return NAL_UNIT_CODED_SLICE_IDR_W_RADL;
    }
  }
  if(m_pocCRA>0)
  {
    if(pocCurr<m_pocCRA)
    {
      // All leading pictures are being marked as TFD pictures here since current encoder uses all
      // reference pictures while encoding leading pictures. An encoder can ensure that a leading
      // picture can be still decodable when random accessing to a CRA/CRANT/BLA/BLANT picture by
      // controlling the reference pictures used for encoding that leading picture. Such a leading
      // picture need not be marked as a TFD picture.
      return NAL_UNIT_CODED_SLICE_RASL;
    }
  }
  if (lastIDR>0)
  {
    if (pocCurr < lastIDR)
    {
      return NAL_UNIT_CODED_SLICE_RADL;
    }
  }
  return NAL_UNIT_CODED_SLICE_TRAIL;
}

void EncGOP::xUpdateRasInit(Slice* slice)
{
  slice->setPendingRasInit( false );
  if ( slice->getPOC() > m_lastRasPoc )
  {
    m_lastRasPoc = MAX_INT;
    slice->setPendingRasInit( true );
  }
  if ( slice->isIRAP() )
  {
    m_lastRasPoc = slice->getPOC();
  }
}

double EncGOP::xCalculateRVM()
{
  double dRVM = 0;

  if( m_pcCfg->getGOPSize() == 1 && m_pcCfg->getIntraPeriod() != 1 && m_pcCfg->getFramesToBeEncoded() > RVM_VCEGAM10_M * 2 )
  {
    // calculate RVM only for lowdelay configurations
    std::vector<double> vRL , vB;
    size_t N = m_vRVM_RP.size();
    vRL.resize( N );
    vB.resize( N );

    int i;
    double dRavg = 0 , dBavg = 0;
    vB[RVM_VCEGAM10_M] = 0;
    for( i = RVM_VCEGAM10_M + 1 ; i < N - RVM_VCEGAM10_M + 1 ; i++ )
    {
      vRL[i] = 0;
      for( int j = i - RVM_VCEGAM10_M ; j <= i + RVM_VCEGAM10_M - 1 ; j++ )
      {
        vRL[i] += m_vRVM_RP[j];
      }
      vRL[i] /= ( 2 * RVM_VCEGAM10_M );
      vB[i] = vB[i-1] + m_vRVM_RP[i] - vRL[i];
      dRavg += m_vRVM_RP[i];
      dBavg += vB[i];
    }

    dRavg /= ( N - 2 * RVM_VCEGAM10_M );
    dBavg /= ( N - 2 * RVM_VCEGAM10_M );

    double dSigamB = 0;
    for( i = RVM_VCEGAM10_M + 1 ; i < N - RVM_VCEGAM10_M + 1 ; i++ )
    {
      double tmp = vB[i] - dBavg;
      dSigamB += tmp * tmp;
    }
    dSigamB = sqrt( dSigamB / ( N - 2 * RVM_VCEGAM10_M ) );

    double f = sqrt( 12.0 * ( RVM_VCEGAM10_M - 1 ) / ( RVM_VCEGAM10_M + 1 ) );

    dRVM = dSigamB / dRavg * f;
  }

  return( dRVM );
}

/** Attaches the input bitstream to the stream in the output NAL unit
    Updates rNalu to contain concatenated bitstream. rpcBitstreamRedirect is cleared at the end of this function call.
 *  \param codedSliceData contains the coded slice data (bitstream) to be concatenated to rNalu
 *  \param rNalu          target NAL unit
 */
void EncGOP::xAttachSliceDataToNalUnit (OutputNALUnit& rNalu, OutputBitstream* codedSliceData)
{
  // Byte-align
  rNalu.m_Bitstream.writeByteAlignment();   // Slice header byte-alignment

  // Perform bitstream concatenation
  if (codedSliceData->getNumberOfWrittenBits() > 0)
  {
    rNalu.m_Bitstream.addSubstream(codedSliceData);
  }
  codedSliceData->clear();
}


void EncGOP::arrangeCompositeReference(Slice* pcSlice, PicList& rcListPic, int pocCurr)
{
  Picture* curPic = NULL;
  PicList::iterator  iterPic = rcListPic.begin();
  const PreCalcValues *pcv = pcSlice->getPPS()->pcv;
  m_bgPOC = pocCurr + 1;
  if (m_picBg->getSpliceFull())
  {
    return;
  }
  while (iterPic != rcListPic.end())
  {
    curPic = *(iterPic++);
    if (curPic->getPOC() == pocCurr)
    {
      break;
    }
  }
  if (pcSlice->isIRAP())
  {
    return;
  }

  int width = pcv->lumaWidth;
  int height = pcv->lumaHeight;
  int stride = curPic->getOrigBuf().get(COMPONENT_Y).stride;
  int cStride = curPic->getOrigBuf().get(COMPONENT_Cb).stride;
  Pel* curLumaAddr = curPic->getOrigBuf().get(COMPONENT_Y).buf;
  Pel* curCbAddr = curPic->getOrigBuf().get(COMPONENT_Cb).buf;
  Pel* curCrAddr = curPic->getOrigBuf().get(COMPONENT_Cr).buf;
  Pel* bgOrgLumaAddr = m_picOrig->getOrigBuf().get(COMPONENT_Y).buf;
  Pel* bgOrgCbAddr = m_picOrig->getOrigBuf().get(COMPONENT_Cb).buf;
  Pel* bgOrgCrAddr = m_picOrig->getOrigBuf().get(COMPONENT_Cr).buf;
  int cuMaxWidth = pcv->maxCUWidth;
  int cuMaxHeight = pcv->maxCUHeight;
  int maxReplace = (pcv->sizeInCtus) / 2;
  maxReplace = maxReplace < 1 ? 1 : maxReplace;
  typedef struct tagCostStr
  {
    double cost;
    int ctuIdx;
  }CostStr;
  CostStr* minCtuCost = new CostStr[maxReplace];
  for (int i = 0; i < maxReplace; i++)
  {
    minCtuCost[i].cost = 1e10;
    minCtuCost[i].ctuIdx = -1;
  }
  int bitIncrementY = pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA) - 8;
  int bitIncrementUV = pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_CHROMA) - 8;
  for (int y = 0; y < height; y += cuMaxHeight)
  {
    for (int x = 0; x < width; x += cuMaxWidth)
    {
      double lcuDist = 0.0;
      double lcuDistCb = 0.0;
      double lcuDistCr = 0.0;
      int    realPixelCnt = 0;
      double lcuCost = 1e10;
      int largeDist = 0;

      for (int tmpy = 0; tmpy < cuMaxHeight; tmpy++)
      {
        if (y + tmpy >= height)
        {
          break;
        }
        for (int tmpx = 0; tmpx < cuMaxWidth; tmpx++)
        {
          if (x + tmpx >= width)
          {
            break;
          }

          realPixelCnt++;
          lcuDist += abs(curLumaAddr[(y + tmpy)*stride + x + tmpx] - bgOrgLumaAddr[(y + tmpy)*stride + x + tmpx]);
          if (abs(curLumaAddr[(y + tmpy)*stride + x + tmpx] - bgOrgLumaAddr[(y + tmpy)*stride + x + tmpx]) >(20 << bitIncrementY))
          {
            largeDist++;
          }

          if (tmpy % 2 == 0 && tmpx % 2 == 0)
          {
            lcuDistCb += abs(curCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] - bgOrgCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2]);
            lcuDistCr += abs(curCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] - bgOrgCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2]);
          }
        }
      }

      //Test the vertical or horizontal edge for background patches candidates
      int yInLCU = y / cuMaxHeight;
      int xInLCU = x / cuMaxWidth;
      int iLCUIdx = yInLCU * pcv->widthInCtus + xInLCU;
      if ((largeDist / (double)realPixelCnt < 0.01 &&lcuDist / realPixelCnt < (3.5 * (1 << bitIncrementY)) && lcuDistCb / realPixelCnt < (0.5 * (1 << bitIncrementUV)) && lcuDistCr / realPixelCnt < (0.5 * (1 << bitIncrementUV)) && m_picBg->getSpliceIdx(iLCUIdx) == 0))
      {
        lcuCost = lcuDist / realPixelCnt + lcuDistCb / realPixelCnt + lcuDistCr / realPixelCnt;
        //obtain the maxReplace smallest cost
        //1) find the largest cost in the maxReplace candidates
        for (int i = 0; i < maxReplace - 1; i++)
        {
          if (minCtuCost[i].cost > minCtuCost[i + 1].cost)
          {
            swap(minCtuCost[i].cost, minCtuCost[i + 1].cost);
            swap(minCtuCost[i].ctuIdx, minCtuCost[i + 1].ctuIdx);
          }
        }
        // 2) compare the current cost with the largest cost
        if (lcuCost < minCtuCost[maxReplace - 1].cost)
        {
          minCtuCost[maxReplace - 1].cost = lcuCost;
          minCtuCost[maxReplace - 1].ctuIdx = iLCUIdx;
        }
      }
    }
  }

  // modify QP for background CTU
  {
    for (int i = 0; i < maxReplace; i++)
    {
      if (minCtuCost[i].ctuIdx != -1)
      {
        m_picBg->setSpliceIdx(minCtuCost[i].ctuIdx, pocCurr);
      }
    }
  }
  delete[]minCtuCost;
}

void EncGOP::updateCompositeReference(Slice* pcSlice, PicList& rcListPic, int pocCurr)
{
  Picture* curPic = NULL;
  const PreCalcValues *pcv = pcSlice->getPPS()->pcv;
  PicList::iterator  iterPic = rcListPic.begin();
  iterPic = rcListPic.begin();
  while (iterPic != rcListPic.end())
  {
    curPic = *(iterPic++);
    if (curPic->getPOC() == pocCurr)
    {
      break;
    }
  }
  assert(curPic->getPOC() == pocCurr);

  int width = pcv->lumaWidth;
  int height = pcv->lumaHeight;
  int stride = curPic->getRecoBuf().get(COMPONENT_Y).stride;
  int cStride = curPic->getRecoBuf().get(COMPONENT_Cb).stride;

  Pel* bgLumaAddr = m_picBg->getRecoBuf().get(COMPONENT_Y).buf;
  Pel* bgCbAddr = m_picBg->getRecoBuf().get(COMPONENT_Cb).buf;
  Pel* bgCrAddr = m_picBg->getRecoBuf().get(COMPONENT_Cr).buf;
  Pel* curLumaAddr = curPic->getRecoBuf().get(COMPONENT_Y).buf;
  Pel* curCbAddr = curPic->getRecoBuf().get(COMPONENT_Cb).buf;
  Pel* curCrAddr = curPic->getRecoBuf().get(COMPONENT_Cr).buf;

  int maxCuWidth = pcv->maxCUWidth;
  int maxCuHeight = pcv->maxCUHeight;

  // Update background reference
  if (pcSlice->isIRAP())//(pocCurr == 0)
  {
    curPic->extendPicBorder();
    curPic->setBorderExtension(true);

    m_picBg->getRecoBuf().copyFrom(curPic->getRecoBuf());
    m_picOrig->getOrigBuf().copyFrom(curPic->getOrigBuf());
  }
  else
  {
    //cout << "update B" << pocCurr << endl;
    for (int y = 0; y < height; y += maxCuHeight)
    {
      for (int x = 0; x < width; x += maxCuWidth)
      {
        if (m_picBg->getSpliceIdx((y / maxCuHeight)*pcv->widthInCtus + x / maxCuWidth) == pocCurr)
        {
          for (int tmpy = 0; tmpy < maxCuHeight; tmpy++)
          {
            if (y + tmpy >= height)
            {
              break;
            }
            for (int tmpx = 0; tmpx < maxCuWidth; tmpx++)
            {
              if (x + tmpx >= width)
              {
                break;
              }
              bgLumaAddr[(y + tmpy)*stride + x + tmpx] = curLumaAddr[(y + tmpy)*stride + x + tmpx];
              if (tmpy % 2 == 0 && tmpx % 2 == 0)
              {
                bgCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] = curCbAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2];
                bgCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2] = curCrAddr[(y + tmpy) / 2 * cStride + (x + tmpx) / 2];
              }
            }
          }
        }
      }
    }
    m_picBg->setBorderExtension(false);
    m_picBg->extendPicBorder();
    m_picBg->setBorderExtension(true);

    curPic->extendPicBorder();
    curPic->setBorderExtension(true);
    m_picOrig->getOrigBuf().copyFrom(curPic->getOrigBuf());

    m_picBg->setBorderExtension(false);
    m_picBg->extendPicBorder();
    m_picBg->setBorderExtension(true);
  }
}

void EncGOP::applyDeblockingFilterMetric( Picture* pcPic, uint32_t uiNumSlices )
{
  PelBuf cPelBuf = pcPic->getRecoBuf().get( COMPONENT_Y );
  Pel* Rec    = cPelBuf.buf;
  const int  stride = cPelBuf.stride;
  const uint32_t picWidth = cPelBuf.width;
  const uint32_t picHeight = cPelBuf.height;

  Pel* tempRec = Rec;
  const Slice* pcSlice = pcPic->slices[0];
  const uint32_t log2maxTB = pcSlice->getSPS()->getLog2MaxTbSize();
  const uint32_t maxTBsize = (1<<log2maxTB);
  const uint32_t minBlockArtSize = 8;
  const uint32_t noCol = (picWidth>>log2maxTB);
  const uint32_t noRows = (picHeight>>log2maxTB);
  CHECK(!(noCol > 1), "Unspecified error");
  CHECK(!(noRows > 1), "Unspecified error");
  std::vector<uint64_t> colSAD(noCol,  uint64_t(0));
  std::vector<uint64_t> rowSAD(noRows, uint64_t(0));
  uint32_t colIdx = 0;
  uint32_t rowIdx = 0;
  Pel p0, p1, p2, q0, q1, q2;

  int qp = pcSlice->getSliceQp();
  const int bitDepthLuma=pcSlice->getSPS()->getBitDepth(CHANNEL_TYPE_LUMA);
  int bitdepthScale = 1 << (bitDepthLuma-8);
  int beta = LoopFilter::getBeta( qp ) * bitdepthScale;
  const int thr2 = (beta>>2);
  const int thr1 = 2*bitdepthScale;
  uint32_t a = 0;

  if (maxTBsize > minBlockArtSize)
  {
    // Analyze vertical artifact edges
    for(int c = maxTBsize; c < picWidth; c += maxTBsize)
    {
      for(int r = 0; r < picHeight; r++)
      {
        p2 = Rec[c-3];
        p1 = Rec[c-2];
        p0 = Rec[c-1];
        q0 = Rec[c];
        q1 = Rec[c+1];
        q2 = Rec[c+2];
        a = ((abs(p2-(p1<<1)+p0)+abs(q0-(q1<<1)+q2))<<1);
        if ( thr1 < a && a < thr2)
        {
          colSAD[colIdx] += abs(p0 - q0);
        }
        Rec += stride;
      }
      colIdx++;
      Rec = tempRec;
    }

    // Analyze horizontal artifact edges
    for(int r = maxTBsize; r < picHeight; r += maxTBsize)
    {
      for(int c = 0; c < picWidth; c++)
      {
        p2 = Rec[c + (r-3)*stride];
        p1 = Rec[c + (r-2)*stride];
        p0 = Rec[c + (r-1)*stride];
        q0 = Rec[c + r*stride];
        q1 = Rec[c + (r+1)*stride];
        q2 = Rec[c + (r+2)*stride];
        a = ((abs(p2-(p1<<1)+p0)+abs(q0-(q1<<1)+q2))<<1);
        if (thr1 < a && a < thr2)
        {
          rowSAD[rowIdx] += abs(p0 - q0);
        }
      }
      rowIdx++;
    }
  }

  uint64_t colSADsum = 0;
  uint64_t rowSADsum = 0;
  for(int c = 0; c < noCol-1; c++)
  {
    colSADsum += colSAD[c];
  }
  for(int r = 0; r < noRows-1; r++)
  {
    rowSADsum += rowSAD[r];
  }

  colSADsum <<= 10;
  rowSADsum <<= 10;
  colSADsum /= (noCol-1);
  colSADsum /= picHeight;
  rowSADsum /= (noRows-1);
  rowSADsum /= picWidth;

  uint64_t avgSAD = ((colSADsum + rowSADsum)>>1);
  avgSAD >>= (bitDepthLuma-8);

  if ( avgSAD > 2048 )
  {
    avgSAD >>= 9;
    int offset = Clip3(2,6,(int)avgSAD);
    for (int i=0; i<uiNumSlices; i++)
    {
      Slice* pcLocalSlice = pcPic->slices[i];
      pcLocalSlice->setDeblockingFilterOverrideFlag   ( true);
      pcLocalSlice->setDeblockingFilterDisable        ( false);
      pcLocalSlice->setDeblockingFilterBetaOffsetDiv2 ( offset );
      pcLocalSlice->setDeblockingFilterTcOffsetDiv2   ( offset );
    }
  }
  else
  {
    for (int i=0; i<uiNumSlices; i++)
    {
      Slice* pcLocalSlice = pcPic->slices[i];
      const PPS* pcPPS = pcSlice->getPPS();
      pcLocalSlice->setDeblockingFilterOverrideFlag  ( false);
      pcLocalSlice->setDeblockingFilterDisable       ( pcPPS->getPPSDeblockingFilterDisabledFlag() );
      pcLocalSlice->setDeblockingFilterBetaOffsetDiv2( pcPPS->getDeblockingFilterBetaOffsetDiv2() );
      pcLocalSlice->setDeblockingFilterTcOffsetDiv2  ( pcPPS->getDeblockingFilterTcOffsetDiv2()   );
    }
  }
}

#if W0038_DB_OPT
void EncGOP::applyDeblockingFilterParameterSelection( Picture* pcPic, const uint32_t numSlices, const int gopID )
{
  enum DBFltParam
  {
    DBFLT_PARAM_AVAILABLE = 0,
    DBFLT_DISABLE_FLAG,
    DBFLT_BETA_OFFSETD2,
    DBFLT_TC_OFFSETD2,
    //NUM_DBFLT_PARAMS
  };
  const int MAX_BETA_OFFSET = 3;
  const int MIN_BETA_OFFSET = -3;
  const int MAX_TC_OFFSET = 3;
  const int MIN_TC_OFFSET = -3;

  PelUnitBuf reco = pcPic->getRecoBuf();

  const int currQualityLayer = (!pcPic->slices[0]->isIRAP()) ? m_pcCfg->getGOPEntry(gopID).m_temporalId+1 : 0;
  CHECK(!(currQualityLayer <MAX_ENCODER_DEBLOCKING_QUALITY_LAYERS), "Unspecified error");

  CodingStructure& cs = *pcPic->cs;

  if(!m_pcDeblockingTempPicYuv)
  {
    m_pcDeblockingTempPicYuv = new PelStorage;
    m_pcDeblockingTempPicYuv->create( cs.area );
    memset(m_DBParam, 0, sizeof(m_DBParam));
  }

  //preserve current reconstruction
  m_pcDeblockingTempPicYuv->copyFrom ( reco );

  const bool bNoFiltering      = m_DBParam[currQualityLayer][DBFLT_PARAM_AVAILABLE] && m_DBParam[currQualityLayer][DBFLT_DISABLE_FLAG]==false /*&& pcPic->getTLayer()==0*/;
  const int  maxBetaOffsetDiv2 = bNoFiltering? Clip3(MIN_BETA_OFFSET, MAX_BETA_OFFSET, m_DBParam[currQualityLayer][DBFLT_BETA_OFFSETD2]+1) : MAX_BETA_OFFSET;
  const int  minBetaOffsetDiv2 = bNoFiltering? Clip3(MIN_BETA_OFFSET, MAX_BETA_OFFSET, m_DBParam[currQualityLayer][DBFLT_BETA_OFFSETD2]-1) : MIN_BETA_OFFSET;
  const int  maxTcOffsetDiv2   = bNoFiltering? Clip3(MIN_TC_OFFSET, MAX_TC_OFFSET, m_DBParam[currQualityLayer][DBFLT_TC_OFFSETD2]+2)       : MAX_TC_OFFSET;
  const int  minTcOffsetDiv2   = bNoFiltering? Clip3(MIN_TC_OFFSET, MAX_TC_OFFSET, m_DBParam[currQualityLayer][DBFLT_TC_OFFSETD2]-2)       : MIN_TC_OFFSET;

  uint64_t distBetaPrevious      = std::numeric_limits<uint64_t>::max();
  uint64_t distMin               = std::numeric_limits<uint64_t>::max();
  bool   bDBFilterDisabledBest = true;
  int    betaOffsetDiv2Best    = 0;
  int    tcOffsetDiv2Best      = 0;

  for(int betaOffsetDiv2=maxBetaOffsetDiv2; betaOffsetDiv2>=minBetaOffsetDiv2; betaOffsetDiv2--)
  {
    uint64_t distTcMin = std::numeric_limits<uint64_t>::max();
    for(int tcOffsetDiv2=maxTcOffsetDiv2; tcOffsetDiv2 >= minTcOffsetDiv2; tcOffsetDiv2--)
    {
      for (int i=0; i<numSlices; i++)
      {
        Slice* pcSlice = pcPic->slices[i];
        pcSlice->setDeblockingFilterOverrideFlag  ( true);
        pcSlice->setDeblockingFilterDisable       ( false);
        pcSlice->setDeblockingFilterBetaOffsetDiv2( betaOffsetDiv2 );
        pcSlice->setDeblockingFilterTcOffsetDiv2  ( tcOffsetDiv2 );
      }

      // restore reconstruction
      reco.copyFrom( *m_pcDeblockingTempPicYuv );

      const uint64_t dist = preLoopFilterPicAndCalcDist( pcPic );

      if(dist < distMin)
      {
        distMin = dist;
        bDBFilterDisabledBest = false;
        betaOffsetDiv2Best  = betaOffsetDiv2;
        tcOffsetDiv2Best = tcOffsetDiv2;
      }
      if(dist < distTcMin)
      {
        distTcMin = dist;
      }
      else if(tcOffsetDiv2 <-2)
      {
        break;
      }
    }
    if(betaOffsetDiv2<-1 && distTcMin >= distBetaPrevious)
    {
      break;
    }
    distBetaPrevious = distTcMin;
  }

  //update:
  m_DBParam[currQualityLayer][DBFLT_PARAM_AVAILABLE] = 1;
  m_DBParam[currQualityLayer][DBFLT_DISABLE_FLAG]    = bDBFilterDisabledBest;
  m_DBParam[currQualityLayer][DBFLT_BETA_OFFSETD2]   = betaOffsetDiv2Best;
  m_DBParam[currQualityLayer][DBFLT_TC_OFFSETD2]     = tcOffsetDiv2Best;

  // restore reconstruction
  reco.copyFrom( *m_pcDeblockingTempPicYuv );

  const PPS* pcPPS = pcPic->slices[0]->getPPS();
  if(bDBFilterDisabledBest)
  {
    for (int i=0; i<numSlices; i++)
    {
      Slice* pcSlice = pcPic->slices[i];
      pcSlice->setDeblockingFilterOverrideFlag( true);
      pcSlice->setDeblockingFilterDisable     ( true);
    }
  }
  else if(betaOffsetDiv2Best == pcPPS->getDeblockingFilterBetaOffsetDiv2() &&  tcOffsetDiv2Best == pcPPS->getDeblockingFilterTcOffsetDiv2())
  {
    for (int i=0; i<numSlices; i++)
    {
      Slice*      pcSlice = pcPic->slices[i];
      pcSlice->setDeblockingFilterOverrideFlag   ( false);
      pcSlice->setDeblockingFilterDisable        ( pcPPS->getPPSDeblockingFilterDisabledFlag() );
      pcSlice->setDeblockingFilterBetaOffsetDiv2 ( pcPPS->getDeblockingFilterBetaOffsetDiv2() );
      pcSlice->setDeblockingFilterTcOffsetDiv2   ( pcPPS->getDeblockingFilterTcOffsetDiv2()   );
    }
  }
  else
  {
    for (int i=0; i<numSlices; i++)
    {
      Slice* pcSlice = pcPic->slices[i];
      pcSlice->setDeblockingFilterOverrideFlag   ( true);
      pcSlice->setDeblockingFilterDisable        ( false );
      pcSlice->setDeblockingFilterBetaOffsetDiv2 ( betaOffsetDiv2Best);
      pcSlice->setDeblockingFilterTcOffsetDiv2   ( tcOffsetDiv2Best);
    }
  }
}
#endif
//! \}