From df873c462549bd90791cd12b4a22d18d58ecfc8f Mon Sep 17 00:00:00 2001
From: Fabrice URBAN <fabrice.urban@interdigital.com>
Date: Tue, 27 Aug 2024 01:31:18 +0000
Subject: [PATCH] FIX input file with rescaling (multilayer scalable) when
using "SourceScalingRatioHor/Ver" encoder option
---
source/App/EncoderApp/EncApp.cpp | 14 +++--
source/Lib/EncoderLib/EncTemporalFilter.cpp | 64 +++++++++++++++++++--
source/Lib/EncoderLib/EncTemporalFilter.h | 6 ++
3 files changed, 74 insertions(+), 10 deletions(-)
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index 4c3d0570b8..da3bc33375 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -1698,7 +1698,9 @@ void EncApp::createLib( const int layerIdx )
{
m_cEncLib.getTemporalFilter().init(m_frameSkip, m_inputBitDepth, m_msbExtendedBitDepth, m_internalBitDepth, m_sourceWidth,
sourceHeight, m_sourcePadding, m_clipInputVideoToRec709Range, m_inputFileName,
- m_chromaFormatIdc, m_inputColourSpaceConvert, m_iQP, m_gopBasedTemporalFilterStrengths,
+ m_chromaFormatIdc, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale,
+ m_horCollocatedChromaFlag, m_verCollocatedChromaFlag,
+ m_inputColourSpaceConvert, m_iQP, m_gopBasedTemporalFilterStrengths,
m_gopBasedTemporalFilterPastRefs, m_gopBasedTemporalFilterFutureRefs, m_firstValidFrame,
m_lastValidFrame, m_gopBasedTemporalFilterEnabled, m_cEncLib.getAdaptQPmap(),
m_cEncLib.getBIM(), m_ctuSize);
@@ -1707,7 +1709,9 @@ void EncApp::createLib( const int layerIdx )
{
m_cEncLib.getTemporalFilterForFG().init(m_frameSkip, m_inputBitDepth, m_msbExtendedBitDepth, m_internalBitDepth, m_sourceWidth,
sourceHeight, m_sourcePadding, m_clipInputVideoToRec709Range, m_inputFileName,
- m_chromaFormatIdc, m_inputColourSpaceConvert, m_iQP, m_fgcSEITemporalFilterStrengths,
+ m_chromaFormatIdc, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale,
+ m_horCollocatedChromaFlag, m_verCollocatedChromaFlag,
+ m_inputColourSpaceConvert, m_iQP, m_fgcSEITemporalFilterStrengths,
m_fgcSEITemporalFilterPastRefs, m_fgcSEITemporalFilterFutureRefs, m_firstValidFrame,
m_lastValidFrame, true, m_cEncLib.getAdaptQPmap(), m_cEncLib.getBIM(), m_ctuSize);
}
@@ -1797,12 +1801,12 @@ bool EncApp::encodePrep( bool& eos )
m_clipInputVideoToRec709Range);
int w0 = m_sourceWidthBeforeScale;
int h0 = m_sourceHeightBeforeScale;
- int w1 = m_orgPic->get(COMPONENT_Y).width - SPS::getWinUnitX(m_chromaFormatIdc) * (m_confWinLeft + m_confWinRight);
- int h1 = m_orgPic->get(COMPONENT_Y).height - SPS::getWinUnitY(m_chromaFormatIdc) * (m_confWinTop + m_confWinBottom);
+ int w1 = m_orgPic->get(COMPONENT_Y).width - m_sourcePadding[0];
+ int h1 = m_orgPic->get(COMPONENT_Y).height - m_sourcePadding[1];
int xScale = ((w0 << ScalingRatio::BITS) + (w1 >> 1)) / w1;
int yScale = ((h0 << ScalingRatio::BITS) + (h1 >> 1)) / h1;
ScalingRatio scalingRatio = { xScale, yScale };
- Window conformanceWindow1(m_confWinLeft, m_confWinRight, m_confWinTop, m_confWinBottom);
+ Window conformanceWindow1(0, m_sourcePadding[0] / SPS::getWinUnitX(m_inputChromaFormatIDC), 0, m_sourcePadding[1] / SPS::getWinUnitY(m_inputChromaFormatIDC));
bool downsampling = (m_sourceWidthBeforeScale > m_sourceWidth) || (m_sourceHeightBeforeScale > m_sourceHeight);
bool useLumaFilter = downsampling;
diff --git a/source/Lib/EncoderLib/EncTemporalFilter.cpp b/source/Lib/EncoderLib/EncTemporalFilter.cpp
index ce0fb82fa2..31e605e45d 100644
--- a/source/Lib/EncoderLib/EncTemporalFilter.cpp
+++ b/source/Lib/EncoderLib/EncTemporalFilter.cpp
@@ -81,6 +81,8 @@ const int EncTemporalFilter::m_cuTreeThresh[4] =
EncTemporalFilter::EncTemporalFilter()
: m_frameSkip(0)
, m_chromaFormatIdc(ChromaFormat::UNDEFINED)
+ , m_sourceWidthBeforeScale(0)
+ , m_sourceHeightBeforeScale(0)
, m_sourceWidth(0)
, m_sourceHeight(0)
, m_QP(0)
@@ -91,6 +93,8 @@ EncTemporalFilter::EncTemporalFilter()
void EncTemporalFilter::init(const int frameSkip, const BitDepths &inputBitDepth, const BitDepths &msbExtendedBitDepth,
const BitDepths &internalBitDepth, const int width, const int height, const int *pad,
const bool rec709, const std::string &filename, const ChromaFormat inputChromaFormatIDC,
+ const int sourceWidthBeforeScale, const int sourceHeightBeforeScale,
+ const int sourceHorCollocatedChromaFlag, const int sourceVerCollocatedChromaFlag,
const InputColourSpaceConversion colorSpaceConv, const int qp,
const std::map<int, double> &temporalFilterStrengths, const int pastRefs,
const int futureRefs, const int firstValidFrame, const int lastValidFrame,
@@ -111,6 +115,10 @@ void EncTemporalFilter::init(const int frameSkip, const BitDepths &inputBitDepth
m_clipInputVideoToRec709Range = rec709;
m_inputFileName = filename;
m_chromaFormatIdc = inputChromaFormatIDC;
+ m_sourceWidthBeforeScale = sourceWidthBeforeScale;
+ m_sourceHeightBeforeScale = sourceHeightBeforeScale;
+ m_sourceHorCollocatedChromaFlag = sourceHorCollocatedChromaFlag;
+ m_sourceVerCollocatedChromaFlag = sourceVerCollocatedChromaFlag;
m_inputColourSpaceConvert = colorSpaceConv;
m_area = Area(0, 0, width, height);
m_QP = qp;
@@ -155,7 +163,14 @@ bool EncTemporalFilter::filter(PelStorage *orgPic, int receivedPoc)
const int lastFrame = std::min(currentFilePoc + m_futureRefs, m_lastValidFrame);
VideoIOYuv yuvFrames;
yuvFrames.open(m_inputFileName, false, m_inputBitDepth, m_msbExtendedBitDepth, m_internalBitDepth);
- yuvFrames.skipFrames(firstFrame, m_sourceWidth - m_pad[0], m_sourceHeight - m_pad[1], m_chromaFormatIdc);
+ if (m_sourceWidthBeforeScale != 0 && m_sourceHeightBeforeScale != 0)
+ {
+ yuvFrames.skipFrames(firstFrame, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale, m_chromaFormatIdc);
+ }
+ else
+ {
+ yuvFrames.skipFrames(firstFrame, m_sourceWidth - m_pad[0], m_sourceHeight - m_pad[1], m_chromaFormatIdc);
+ }
std::deque<TemporalFilterSourcePicInfo> srcFrameInfo;
@@ -177,7 +192,14 @@ bool EncTemporalFilter::filter(PelStorage *orgPic, int receivedPoc)
{
if (poc == currentFilePoc)
{ // hop over frame that will be filtered
- yuvFrames.skipFrames(1, m_sourceWidth - m_pad[0], m_sourceHeight - m_pad[1], m_chromaFormatIdc);
+ if (m_sourceWidthBeforeScale != 0 && m_sourceHeightBeforeScale != 0)
+ {
+ yuvFrames.skipFrames(1, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale, m_chromaFormatIdc);
+ }
+ else
+ {
+ yuvFrames.skipFrames(1, m_sourceWidth - m_pad[0], m_sourceHeight - m_pad[1], m_chromaFormatIdc);
+ }
continue;
}
srcFrameInfo.push_back(TemporalFilterSourcePicInfo());
@@ -185,9 +207,41 @@ bool EncTemporalFilter::filter(PelStorage *orgPic, int receivedPoc)
PelStorage dummyPicBufferTO; // Only used temporary in yuvFrames.read
srcPic.picBuffer.create(m_chromaFormatIdc, m_area, 0, m_padding);
- dummyPicBufferTO.create(m_chromaFormatIdc, m_area, 0, m_padding);
- if (!yuvFrames.read(srcPic.picBuffer, dummyPicBufferTO, m_inputColourSpaceConvert, m_pad, m_chromaFormatIdc,
- m_clipInputVideoToRec709Range))
+ bool readOk = false;
+ if (m_sourceWidthBeforeScale != 0 && m_sourceHeightBeforeScale != 0)
+ {
+ Area areaPrescale(0, 0, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale);
+ PelStorage m_orgPicBeforeScale;
+ m_orgPicBeforeScale.create(m_chromaFormatIdc, areaPrescale, 0, m_padding);
+ dummyPicBufferTO.create(m_chromaFormatIdc, areaPrescale, 0, m_padding);
+ readOk = yuvFrames.read(m_orgPicBeforeScale, dummyPicBufferTO, m_inputColourSpaceConvert, m_pad, m_chromaFormatIdc,
+ m_clipInputVideoToRec709Range);
+ if (readOk)
+ {
+ int w0 = m_sourceWidthBeforeScale;
+ int h0 = m_sourceHeightBeforeScale;
+ int w1 = m_sourceWidth - m_pad[0];
+ int h1 = m_sourceHeight - m_pad[1];
+ int xScale = ((w0 << ScalingRatio::BITS) + (w1 >> 1)) / w1;
+ int yScale = ((h0 << ScalingRatio::BITS) + (h1 >> 1)) / h1;
+ ScalingRatio scalingRatio = { xScale, yScale };
+ Window conformanceWindow1(0, m_pad[0] / SPS::getWinUnitX(m_chromaFormatIdc), 0, m_pad[1] / SPS::getWinUnitY(m_chromaFormatIdc));
+
+ bool downsampling = (m_sourceWidthBeforeScale > m_sourceWidth) || (m_sourceHeightBeforeScale > m_sourceHeight);
+ bool useLumaFilter = downsampling;
+ Picture::rescalePicture(scalingRatio, m_orgPicBeforeScale, Window(), srcPic.picBuffer, conformanceWindow1,
+ m_chromaFormatIdc, m_internalBitDepth, useLumaFilter, downsampling,
+ m_sourceHorCollocatedChromaFlag != 0, m_sourceVerCollocatedChromaFlag != 0);
+ }
+ }
+ else
+ {
+ dummyPicBufferTO.create(m_chromaFormatIdc, m_area, 0, m_padding);
+ readOk = yuvFrames.read(srcPic.picBuffer, dummyPicBufferTO, m_inputColourSpaceConvert, m_pad, m_chromaFormatIdc,
+ m_clipInputVideoToRec709Range);
+ }
+
+ if(!readOk)
{
// eof or read fail
srcPic.picBuffer.destroy();
diff --git a/source/Lib/EncoderLib/EncTemporalFilter.h b/source/Lib/EncoderLib/EncTemporalFilter.h
index d6ac28416d..241580d7b7 100644
--- a/source/Lib/EncoderLib/EncTemporalFilter.h
+++ b/source/Lib/EncoderLib/EncTemporalFilter.h
@@ -110,6 +110,8 @@ public:
void init(const int frameSkip, const BitDepths &inputBitDepth, const BitDepths &msbExtendedBitDepth,
const BitDepths &internalBitDepth, const int width, const int height, const int *pad, const bool rec709,
const std::string &filename, const ChromaFormat inputChroma,
+ const int sourceWidthBeforeScale, const int sourceHeightBeforeScale,
+ const int sourceHorCollocatedChromaFlag, const int sourceVerCollocatedChromaFlag,
const InputColourSpaceConversion colorSpaceConv, const int qp,
const std::map<int, double> &temporalFilterStrengths, const int pastRefs, const int futureRefs,
const int firstValidFrame, const int lastValidFrame, const bool bMCTFenabled,
@@ -137,6 +139,10 @@ private:
BitDepths m_internalBitDepth;
ChromaFormat m_chromaFormatIdc;
+ int m_sourceWidthBeforeScale;
+ int m_sourceHeightBeforeScale;
+ int m_sourceHorCollocatedChromaFlag;
+ int m_sourceVerCollocatedChromaFlag;
int m_sourceWidth;
int m_sourceHeight;
int m_QP;
--
GitLab