sys_encoder_vopseenc.cpp

完整的RTP RTSP代码库

	// m_pfdct = new CFwdBlockDCT(volmd.nBits);

	// B-VOP MB buffer
	m_puciDirectPredMB = new CU8Image (CRct (0, 0, MB_SIZE, MB_SIZE));
	m_puciInterpPredMB = new CU8Image (CRct (0, 0, MB_SIZE, MB_SIZE));
	m_ppxlcDirectPredMBY = (PixelC*) m_puciDirectPredMB->pixels (); 
	m_ppxlcInterpPredMBY = (PixelC*) m_puciInterpPredMB->pixels (); 

	m_piiDirectErrorMB = new CIntImage (CRct (0, 0, MB_SIZE, MB_SIZE)); 
	m_piiInterpErrorMB = new CIntImage (CRct (0, 0, MB_SIZE, MB_SIZE)); 
	m_ppxliDirectErrorMBY = (PixelI*) m_piiDirectErrorMB->pixels (); 
	m_ppxliInterpErrorMBY = (PixelI*) m_piiInterpErrorMB->pixels (); 

	// with shape
	if (m_volmd.fAUsage != RECTANGLE) {
		m_rgiSubBlkIndx16x16 = computeShapeSubBlkIndex (4, 16);
		m_rgiSSubBlkIndx16x16 = computeShapeSubBlkIndex (2, 16);
		m_rgiSubBlkIndx18x18 = computeShapeSubBlkIndex (4, 18);
		m_rgiSubBlkIndx20x20 = computeShapeSubBlkIndex (4, 20);
		m_rgiPxlIndx12x12 = new Int [8 * 8];		
		Int* piPxl = m_rgiPxlIndx12x12;
		UInt i, j;
		for (i = 2; i < 10; i++)
			for (j = 2; j < 10; j++)
				*piPxl++ = i * 12 + j;
		m_rgiPxlIndx8x8 = new Int [4 * 4];		
		piPxl = m_rgiPxlIndx8x8;
		for (i = 2; i < 6; i++)
			for (j = 2; j < 6; j++)
				*piPxl++ = i * 8 + j;
	}
	
// NEWPRED
	if(m_volmd.bNewpredEnable) {
		g_pNewPredEnc->SetObject(
					m_iNumBitsTimeIncr,
					iSessionWidth,	
					iSessionHeight,
					uiVOId,
					m_volmd.cNewpredRefName,
					m_volmd.cNewpredSlicePoint,
					m_volmd.bNewpredSegmentType,
					m_volmd.fAUsage,
					m_volmd.bShapeOnly,
					m_pvopcRefQ0,
					m_pvopcRefQ1,
					m_rctRefFrameY,	
					m_rctRefFrameUV
		);
	}
// ~NEWPRED

	m_statsVOL.reset ();
	if(m_volmd.bCodeSequenceHead)
		codeSequenceHead ();

	codeVOHead ();
	codeVOLHead (iSessionWidth, iSessionHeight);//, rctSprite);

	// Added for error resilient mode by Toshiba(1997-11-14): Moved (1998-1-16)
	g_iMaxHeading = MAXHEADING_ERR;
	g_iMaxMiddle = MAXMIDDLE_ERR;
	g_iMaxTrailing = MAXTRAILING_ERR;

	// End Toshiba(1997-11-14)

	m_statsVOL.nBitsStuffing += m_pbitstrmOut->flush ();
	m_statRC.resetSkipMode ();
	m_rgdSNR = (m_volmd.fAUsage == EIGHT_BIT) ? new Double [3+m_volmd.iAuxCompCount] : new Double [3];

	// some fixed variables	
	m_iFrameWidthZoomY = m_iFrameWidthY * 2;
	m_iFrameWidthZoomUV = m_iFrameWidthUV * 2;
	m_iFrameWidthZoomYx2Minus2MB = m_iFrameWidthY * 2 * 2 - 2 * MB_SIZE;
	m_iFrameWidthZoomYx2Minus2Blk = m_iFrameWidthY * 2 * 2 - 2 * BLOCK_SIZE;

// RRV insertion
	Int iScale	= (m_vopmd.RRVmode.iOnOff == 1) ? (2) : (1);
// ~RRV 
	if (m_volmd.fAUsage == RECTANGLE) {
		m_rctCurrVOPY = CRct (0, 0, iSessionWidthRound, iSessionHeightRound);
		m_pvopcOrig->setBoundRct (m_rctCurrVOPY);
		m_rctCurrVOPUV = m_rctCurrVOPY.downSampleBy2 ();
		
		m_rctRefVOPY0 = m_rctCurrVOPY;
// RRV modification
		m_rctRefVOPY0.expand (EXPANDY_REFVOP *iScale);
//		m_rctRefVOPY0.expand (EXPANDY_REFVOP);
// ~RRV
		m_rctRefVOPUV0 = m_rctRefVOPY0.downSampleBy2 ();
		
		m_rctRefVOPY1 = m_rctRefVOPY0;
		m_rctRefVOPUV1 = m_rctRefVOPUV0;
		
		m_rctRefVOPZoom0 = m_rctRefVOPY0.upSampleBy2 ();
		m_rctRefVOPZoom1 = m_rctRefVOPY1.upSampleBy2 ();
		m_pvopcRefOrig0->setBoundRct (m_rctRefVOPY0);
		m_pvopcRefOrig1->setBoundRct (m_rctRefVOPY1);

	if ((m_uiSprite == 0) || (m_uiSprite == 2) || ((m_uiSprite == 1) && (m_sptMode == BASIC_SPRITE)) ) // GMC

		computeVOLConstMembers (); // these VOP members are the same for all frames
	}
//OBSS_SAIT_991015
	else if(m_volmd.fAUsage == ONE_BIT && m_volmd.volType == ENHN_LAYER) {
		m_rctCurrVOPY = CRct (0, 0, iSessionWidthRound, iSessionHeightRound);
		m_pvopcOrig->setBoundRct (m_rctCurrVOPY);
		m_rctCurrVOPUV = m_rctCurrVOPY.downSampleBy2 ();
		
		m_rctRefVOPY0 = m_rctCurrVOPY;
		m_rctRefVOPY0.expand (EXPANDY_REFVOP);
		m_rctRefVOPUV0 = m_rctRefVOPY0.downSampleBy2 ();
		
		m_rctRefVOPY1 = m_rctRefVOPY0;
		m_rctRefVOPUV1 = m_rctRefVOPUV0;
		
		m_rctRefVOPZoom0 = m_rctRefVOPY0.upSampleBy2 ();
		m_rctRefVOPZoom1 = m_rctRefVOPY1.upSampleBy2 ();
		m_pvopcRefOrig0->setBoundRct (m_rctRefVOPY0);
		m_pvopcRefOrig1->setBoundRct (m_rctRefVOPY1);
	}
//~OBSS_SAIT_991015
	m_vopmd.SpriteXmitMode = STOP ;	
    // tentative solution for indicating the first Sprite VOP
    // tentativeFirstSpriteVop = 0;

	// Open motion vector file if used
	m_pchMVFileName = pchMVFileName;
	if (m_iMVFileUsage == iMVFileUsage) {
		m_fMVFile = fopen(m_pchMVFileName, (m_iMVFileUsage == 1) ? "r" : "w");
		assert(m_fMVFile != NULL);
		m_iMVLineNo = 0;
	}
}

// for back/forward shape
CVideoObjectEncoder::CVideoObjectEncoder (
	UInt uiVOId, 
	VOLMode& volmd, 
	VOPMode& vopmd,
	Int iSessionWidth,
	Int iSessionHeight
) :
  CVideoObject (),  
  m_statsVOL(volmd.iAuxCompCount),
  m_statsVOP(volmd.iAuxCompCount),
  m_statsMB (volmd.iAuxCompCount)

{
  m_pvopcOrig = NULL;
  m_rgiSubBlkIndx16x16 =NULL;
  m_rgiSSubBlkIndx16x16 = NULL;
  m_rgiSubBlkIndx18x18 = NULL;
  m_rgiSubBlkIndx20x20 = NULL;
  m_rgiPxlIndx12x12 = NULL;
  m_rgiPxlIndx8x8  = NULL;

	// sprite stuff
    	m_uiSprite = 0; // sprite off

	m_pchBitsBuffer = NULL;
	m_pbitstrmOut = NULL;
	m_pentrencSet = NULL;

        // shape cache
	m_pchShapeBitsBuffer = NULL;
	m_pbitstrmShape = NULL;
	m_pbitstrmShapeMBOut = NULL;
	
	m_uiVOId = uiVOId;
	m_volmd = volmd;

	m_vopmd = vopmd;
//	decideMVInfo ();
	m_vopmd.iRoundingControl = 0;
	m_vopmd.iRoundingControlEncSwitch = 0;
	Int iMod = iSessionWidth % MB_SIZE;
	Int iSessionWidthRound = (iMod > 0) ? iSessionWidth + MB_SIZE - iMod : iSessionWidth;
	iMod = iSessionHeight % MB_SIZE;
	Int iSessionHeightRound = (iMod > 0) ? iSessionHeight + MB_SIZE - iMod : iSessionHeight;
	m_rctRefFrameY = CRct (
		-EXPANDY_REF_FRAME, -EXPANDY_REF_FRAME, 
		EXPANDY_REF_FRAME + iSessionWidthRound, EXPANDY_REF_FRAME + iSessionHeightRound
	);
	m_rctRefFrameUV = m_rctRefFrameY.downSampleBy2 ();
	m_pvopcOrig = new CVOPU8YUVBA (m_volmd.fAUsage, m_rctRefFrameY, m_volmd.iAuxCompCount);
	m_pvopcRefOrig0 = new CVOPU8YUVBA (m_volmd.fAUsage, m_rctRefFrameY, m_volmd.iAuxCompCount);
	m_pvopcRefOrig1 = new CVOPU8YUVBA (m_volmd.fAUsage, m_rctRefFrameY, m_volmd.iAuxCompCount);
	allocateVOLMembers (iSessionWidth, iSessionHeight);

	CRct rctFrameZoom = m_rctRefFrameY.upSampleBy2 ();
	m_puciRefQZoom0 = new CU8Image (rctFrameZoom);

	m_puciRefQZoom1 = new CU8Image (rctFrameZoom);

	// m_pfdct = new CFwdBlockDCT;
	// HHI Schueuer: sadct
	if (m_volmd.fAUsage != RECTANGLE && (!m_volmd.bSadctDisable)) {
		m_pfdct = new CFwdSADCT();
		m_pscanSelector = new CScanSelectorForSADCT(m_rgiCurrMBCoeffWidth);
	}
	else {
		m_pfdct = new CFwdBlockDCT();
		m_pscanSelector = new CScanSelector;
	}
	// end HHI

	// B-VOP MB buffer
	m_puciDirectPredMB = new CU8Image (CRct (0, 0, MB_SIZE, MB_SIZE));
	m_puciInterpPredMB = new CU8Image (CRct (0, 0, MB_SIZE, MB_SIZE));
	m_ppxlcDirectPredMBY = (PixelC*) m_puciDirectPredMB->pixels (); 
	m_ppxlcInterpPredMBY = (PixelC*) m_puciInterpPredMB->pixels (); 

	m_piiDirectErrorMB = new CIntImage (CRct (0, 0, MB_SIZE, MB_SIZE)); 
	m_piiInterpErrorMB = new CIntImage (CRct (0, 0, MB_SIZE, MB_SIZE)); 
	m_ppxliDirectErrorMBY = (PixelI*) m_piiDirectErrorMB->pixels (); 
	m_ppxliInterpErrorMBY = (PixelI*) m_piiInterpErrorMB->pixels (); 

	// with shape
	if (m_volmd.fAUsage != RECTANGLE) {
		m_rgiSubBlkIndx16x16 = computeShapeSubBlkIndex (4, 16);
		m_rgiSSubBlkIndx16x16 = computeShapeSubBlkIndex (2, 16);
		m_rgiSubBlkIndx18x18 = computeShapeSubBlkIndex (4, 18);
		m_rgiSubBlkIndx20x20 = computeShapeSubBlkIndex (4, 20);
		m_rgiPxlIndx12x12 = new Int [8 * 8];		
		Int* piPxl = m_rgiPxlIndx12x12;
		UInt i, j;
		for (i = 2; i < 10; i++)
			for (j = 2; j < 10; j++)
				*piPxl++ = i * 12 + j;
		m_rgiPxlIndx8x8 = new Int [4 * 4];		
		piPxl = m_rgiPxlIndx8x8;
		for (i = 2; i < 6; i++)
			for (j = 2; j < 6; j++)
				*piPxl++ = i * 8 + j;
	}

	m_statRC.resetSkipMode ();
	m_rgdSNR = (m_volmd.fAUsage == EIGHT_BIT) ? new Double [4] : new Double [3];

	// some fixed variables	
	m_iFrameWidthZoomY = m_iFrameWidthY * 2;
	m_iFrameWidthZoomUV = m_iFrameWidthUV * 2;
	m_iFrameWidthZoomYx2Minus2MB = m_iFrameWidthY * 2 * 2 - 2 * MB_SIZE;
	m_iFrameWidthZoomYx2Minus2Blk = m_iFrameWidthY * 2 * 2 - 2 * BLOCK_SIZE;
	
	if (m_volmd.fAUsage == RECTANGLE) {
		m_rctCurrVOPY = CRct (0, 0, iSessionWidthRound, iSessionHeightRound);
		m_pvopcOrig->setBoundRct (m_rctCurrVOPY);
		m_rctCurrVOPUV = m_rctCurrVOPY.downSampleBy2 ();
		
		m_rctRefVOPY0 = m_rctCurrVOPY;
		m_rctRefVOPY0.expand (EXPANDY_REFVOP);
		m_rctRefVOPUV0 = m_rctRefVOPY0.downSampleBy2 ();
		
		m_rctRefVOPY1 = m_rctRefVOPY0;
		m_rctRefVOPUV1 = m_rctRefVOPUV0;
		
		m_rctRefVOPZoom0 = m_rctRefVOPY0.upSampleBy2 ();
		m_pvopcRefOrig0->setBoundRct (m_rctRefVOPY0);
		m_pvopcRefOrig1->setBoundRct (m_rctRefVOPY1);
		computeVOLConstMembers (); // these VOP members are the same for all frames
	}
}

Bool CVideoObjectEncoder::skipTest(Time t,VOPpredType vopPredType)
{
	if (m_uiRateControl==RC_MPEG4) {
		if(vopPredType!=IVOP)
			if (m_statRC.skipThisFrame ())
				return TRUE;
	}
	return FALSE;
}

Void CVideoObjectEncoder::encode (
	Bool bVOPVisible, 
	Time t,
	VOPpredType vopPredType,
	const CVOPU8YUVBA* pvopfRefBaseLayer)
{
	m_t = t;
	m_vopmd.vopPredType = vopPredType;
// RRV insertion
    if(m_vopmd.RRVmode.iOnOff == 1)
	{
		m_vopmd.RRVmode.iQave		= 0;
		if(m_statsVOP.nQMB != 0)
		{
			m_vopmd.RRVmode.iQave	= m_statsVOP.nQp /m_statsVOP.nQMB;
		}
		m_vopmd.RRVmode.iNumBits	= m_statsVOP.nBitsTotal;
	}
	m_iRRVScale	= 1;	// default value
// ~RRV	
    m_statsVOP.reset ();

	Bool bTemporalScalability = m_volmd.bTemporalScalability; // added by Sharp (98/2/10)
	Bool bPrevRefVopWasCoded = m_bCodedFutureRef;

//OBSS_SAIT_991015
	if (m_volmd.volType == ENHN_LAYER && m_volmd.fAUsage == ONE_BIT && (m_volmd.bSpatialScalability && m_volmd.iHierarchyType==0) && m_volmd.iuseRefShape ==0 ) { 
		if (!m_bCodedBaseRef  || !bPrevRefVopWasCoded ) 
			m_vopmd.vopPredType = PVOP;
	}
//~OBSS_SAIT_991015

	//__LOW_LATENCY_SPRITE_
	if (m_uiSprite == 1 && m_sptMode != BASIC_SPRITE && m_vopmd.SpriteXmitMode != STOP)	{
		setRefStartingPointers (); // need to compute the starting pointers of the refVOP's (corresponding to the current Rct)
		computeVOPMembers ();
		m_iNumSptMB = 0;
		encodeVOP ();
	}
	else {
	    if (m_uiSprite == 1 && m_t>=0)
		    m_vopmd.vopPredType = SPRITE;
	
	    m_pbitstrmOut->reset ();
#if 0
	    cout << "\nVOP " << m_uiVOId << "\n";
	    cout << "\tVOL " << m_volmd.volType << "\n"; // added by Sharp (98/2/10)
#endif

		if(m_vopmd.vopPredType==IVOP || m_vopmd.vopPredType==PVOP || ((m_uiSprite == 2) && (m_vopmd.vopPredType == SPRITE))) // GMC
			m_bCodedFutureRef = bVOPVisible; // flag used by bvop prediction
//OBSS_SAIT_991015
		if (m_volmd.volType == ENHN_LAYER && m_volmd.fAUsage == ONE_BIT && (m_volmd.bSpatialScalability && m_volmd.iHierarchyType==0) && m_volmd.iuseRefShape ==0 )
			m_bCodedFutureRef = bVOPVisible; // flag used by OBSS bvop prediction
//~OBSS_SAIT_991015

	    if(!bVOPVisible)
	    {
#ifdef __TRACE_AND_STATS_
		    m_pbitstrmOut -> trace (m_t, "VOP_Time");
#endif // __TRACE_AND_STATS_
		    codeNonCodedVOPHead ();
#if 0
		    cout << "\tTime..." << m_t << " (" << m_t/m_volmd.dFrameHz << " sec)" 
				<< "\n\tNot coded.\n\n";
		    cout.flush ();
#endif
		    m_statsVOP.nBitsStuffing += m_pbitstrmOut->flush ();
#ifdef __TRACE_AND_STATS_
		    m_statsVOP.dSNRY = m_statsVOP.dSNRU = m_statsVOP.dSNRV = 0;
		    m_statsVOP.print (TRUE);
		    m_statsVOL += m_statsVOP;
#endif // __TRACE_AND_STATS_
	    }