From 2861e1313114ac2d4bdea3f0aa193c40767dadec Mon Sep 17 00:00:00 2001
From: charles Salmon-Legagneur <charles.salmon-legagneur@interdigital.com>
Date: Wed, 1 Feb 2023 22:53:35 +0000
Subject: [PATCH] add options --SourceScalingRatio to rescale source video
---
doc/software-manual.tex | 10 ++++++
source/App/EncoderApp/EncApp.cpp | 56 ++++++++++++++++++++++++++---
source/App/EncoderApp/EncApp.h | 2 ++
source/App/EncoderApp/EncAppCfg.cpp | 15 ++++++++
source/App/EncoderApp/EncAppCfg.h | 4 +++
5 files changed, 83 insertions(+), 4 deletions(-)
diff --git a/doc/software-manual.tex b/doc/software-manual.tex
index 58ace85d79..6be5b18bd6 100644
--- a/doc/software-manual.tex
+++ b/doc/software-manual.tex
@@ -639,6 +639,16 @@ Specifies the output locally reconstructed video file. If more than one layer is
Specifies the width and height of the input video in luma samples.
\\
+
+\Option{SourceScalingRatioHor}%
+\Option{SourceScalingRatioVer} &
+\Default{1.0}%
+\Default{1.0} &
+Specifies a scaling ratio to apply in hor and vert direction to the pictures read from input video file.
+Note: The SourceWidth and SourceHeight are multiplied by these scaling factors. This option is useful for spatial scalability in a multi layer scenario to use enhancement layer source when base layer source is not available.
+\\
+
+
\Option{InputBitDepth}
&
%\ShortOption{\None} &
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index a0e9349b77..c72accaaad 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -1403,8 +1403,16 @@ void EncApp::xCreateLib( std::list<PelUnitBuf*>& recBufList, const int layerId )
m_cVideoIOYuvInputFile.skipFrames(m_frameSkip, m_inputFileWidth, m_inputFileHeight, m_inputChromaFormatIDC);
#else
const int sourceHeight = m_isField ? m_iSourceHeightOrg : m_sourceHeight;
- m_cVideoIOYuvInputFile.skipFrames(m_frameSkip, m_sourceWidth - m_sourcePadding[0], sourceHeight - m_sourcePadding[1],
- m_inputChromaFormatIDC);
+ if (m_sourceScalingRatioHor != 1.0 || m_sourceScalingRatioVer != 1.0)
+ {
+ m_cVideoIOYuvInputFile.skipFrames(m_frameSkip, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale,
+ m_inputChromaFormatIDC);
+ }
+ else
+ {
+ m_cVideoIOYuvInputFile.skipFrames(m_frameSkip, m_sourceWidth - m_sourcePadding[0],
+ sourceHeight - m_sourcePadding[1], m_inputChromaFormatIDC);
+ }
#endif
if (!m_reconFileName.empty())
{
@@ -1497,6 +1505,14 @@ void EncApp::createLib( const int layerIdx )
m_trueOrgPic = new PelStorage;
m_orgPic->create( unitArea );
m_trueOrgPic->create( unitArea );
+ if (m_sourceScalingRatioHor != 1.0 || m_sourceScalingRatioVer != 1.0)
+ {
+ UnitArea unitAreaPrescale( m_chromaFormatIDC, Area( 0, 0, m_sourceWidthBeforeScale, m_sourceHeightBeforeScale) );
+ m_orgPicBeforeScale = new PelStorage;
+ m_trueOrgPicBeforeScale = new PelStorage;
+ m_orgPicBeforeScale->create( unitAreaPrescale );
+ m_trueOrgPicBeforeScale->create( unitAreaPrescale );
+ }
if ( m_gopBasedTemporalFilterEnabled || m_bimEnabled )
{
m_filteredOrgPic = new PelStorage;
@@ -1588,6 +1604,15 @@ void EncApp::destroyLib()
m_trueOrgPic->destroy();
delete m_trueOrgPic;
delete m_orgPic;
+
+ if (m_sourceScalingRatioHor != 1.0 || m_sourceScalingRatioVer != 1.0)
+ {
+ m_orgPicBeforeScale->destroy();
+ m_trueOrgPicBeforeScale->destroy();
+ delete m_trueOrgPicBeforeScale;
+ delete m_orgPicBeforeScale;
+ }
+
if (m_resChangeInClvsEnabled && m_gopBasedRPREnabledFlag)
{
for (int i = 0; i < 2; i++)
@@ -1641,8 +1666,31 @@ bool EncApp::encodePrep( bool& eos )
m_clipInputVideoToRec709Range);
}
#else
- m_cVideoIOYuvInputFile.read(*m_orgPic, *m_trueOrgPic, ipCSC, m_sourcePadding, m_inputChromaFormatIDC,
- m_clipInputVideoToRec709Range);
+ if (m_sourceScalingRatioHor != 1.0 || m_sourceScalingRatioVer != 1.0)
+ {
+ int noPadding[2] = { 0 };
+ m_cVideoIOYuvInputFile.read(*m_orgPicBeforeScale, *m_trueOrgPicBeforeScale, ipCSC, noPadding, m_inputChromaFormatIDC,
+ 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 xScale = ((w0 << ScalingRatio::BITS) + (w1 >> 1)) / w1;
+ int yScale = ((h0 << ScalingRatio::BITS) + (h1 >> 1)) / h1;
+ ScalingRatio scalingRatio = { xScale, yScale };
+ Window conformanceWindow1;
+ conformanceWindow1.setWindow(m_confWinLeft, m_confWinRight, m_confWinTop, m_confWinBottom);
+
+ bool downsampling = (m_sourceWidthBeforeScale > m_sourceWidth) || (m_sourceHeightBeforeScale > m_sourceHeight);
+ bool useLumaFilter = downsampling;
+ Picture::rescalePicture(scalingRatio, *m_orgPicBeforeScale, Window(), *m_orgPic, conformanceWindow1, m_inputChromaFormatIDC , m_internalBitDepth,useLumaFilter,downsampling,m_horCollocatedChromaFlag,m_verCollocatedChromaFlag );
+ m_trueOrgPic->copyFrom(*m_orgPic);
+ }
+ else
+ {
+ m_cVideoIOYuvInputFile.read(*m_orgPic, *m_trueOrgPic, ipCSC, m_sourcePadding, m_inputChromaFormatIDC,
+ m_clipInputVideoToRec709Range);
+ }
#endif
if (m_fgcSEIAnalysisEnabled && m_fgcSEIExternalDenoised.empty())
diff --git a/source/App/EncoderApp/EncApp.h b/source/App/EncoderApp/EncApp.h
index e48ce07b6c..3b63cf63c4 100644
--- a/source/App/EncoderApp/EncApp.h
+++ b/source/App/EncoderApp/EncApp.h
@@ -99,6 +99,8 @@ private:
int m_numEncoded;
PelStorage* m_trueOrgPic;
PelStorage* m_orgPic;
+ PelStorage* m_trueOrgPicBeforeScale;
+ PelStorage* m_orgPicBeforeScale;
PelStorage* m_filteredOrgPic;
PelStorage* m_rprPic[2];
#if EXTENSION_360_VIDEO
diff --git a/source/App/EncoderApp/EncAppCfg.cpp b/source/App/EncoderApp/EncAppCfg.cpp
index 331399be2e..bc6e11f1d0 100644
--- a/source/App/EncoderApp/EncAppCfg.cpp
+++ b/source/App/EncoderApp/EncAppCfg.cpp
@@ -772,6 +772,8 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
#endif
("SourceWidth,-wdt", m_sourceWidth, 0, "Source picture width")
("SourceHeight,-hgt", m_sourceHeight, 0, "Source picture height")
+ ("SourceScalingRatioHor", m_sourceScalingRatioHor, 1.0, "Source picture horizontal scaling ratio")
+ ("SourceScalingRatioVer", m_sourceScalingRatioVer, 1.0, "Source picture vertical scaling ratio")
("InputBitDepth", m_inputBitDepth[ChannelType::LUMA], 8, "Bit-depth of input file")
("OutputBitDepth", m_outputBitDepth[ChannelType::LUMA], 0, "Bit-depth of output file (default:InternalBitDepth)")
("MSBExtendedBitDepth", m_msbExtendedBitDepth[ChannelType::LUMA], 0, "bit depth of luma component after addition of MSBs of value 0 (used for synthesising High Dynamic Range source material). (default:InputBitDepth)")
@@ -2112,6 +2114,19 @@ bool EncAppCfg::parseCfg( int argc, char* argv[] )
/*
* Set any derived parameters
*/
+
+ if ( m_sourceScalingRatioHor != 1.0 || m_sourceScalingRatioVer != 1.0 )
+ {
+ m_sourceWidthBeforeScale = m_sourceWidth;
+ m_sourceHeightBeforeScale = m_sourceHeight;
+ m_sourceWidth = int(round(m_sourceWidth*m_sourceScalingRatioHor));
+ m_sourceHeight = int(round(m_sourceHeight*m_sourceScalingRatioVer));
+ }
+ else
+ {
+ m_sourceWidthBeforeScale = 0;
+ m_sourceHeightBeforeScale = 0;
+ }
#if EXTENSION_360_VIDEO
m_inputFileWidth = m_sourceWidth;
m_inputFileHeight = m_sourceHeight;
diff --git a/source/App/EncoderApp/EncAppCfg.h b/source/App/EncoderApp/EncAppCfg.h
index e6b82a3aee..275c85c0f3 100644
--- a/source/App/EncoderApp/EncAppCfg.h
+++ b/source/App/EncoderApp/EncAppCfg.h
@@ -95,6 +95,10 @@ protected:
uint32_t m_temporalSubsampleRatio; ///< temporal subsample ratio, 2 means code every two frames
int m_sourceWidth; ///< source width in pixel
int m_sourceHeight; ///< source height in pixel (when interlaced = field height)
+ double m_sourceScalingRatioHor; ////< source scaling ratio Horizontal
+ double m_sourceScalingRatioVer; ////< source scaling ratio Vertical
+ int m_sourceWidthBeforeScale; ///< source width in pixel before applying source scaling ratio Horizontal
+ int m_sourceHeightBeforeScale; ///< source height in pixel before applying source scaling ratio Vertical (when interlaced = field height)
#if EXTENSION_360_VIDEO
int m_inputFileWidth; ///< width of image in input file (this is equivalent to sourceWidth, if sourceWidth is not subsequently altered due to padding)
int m_inputFileHeight; ///< height of image in input file (this is equivalent to sourceHeight, if sourceHeight is not subsequently altered due to padding)
--
GitLab