From 2ee8b0289b941f964012e32b874a7a1492131959 Mon Sep 17 00:00:00 2001
From: Xiang Li <xlxiangli@google.com>
Date: Tue, 21 Jun 2022 10:24:25 -0700
Subject: [PATCH] Y4M support at both encoder and decoder
---
.clang-format | 8 +-
source/App/DecoderApp/DecApp.cpp | 8 ++
source/App/DecoderApp/DecAppCfg.cpp | 8 ++
source/App/DecoderApp/DecAppCfg.h | 1 +
source/App/EncoderApp/EncApp.cpp | 24 ++++
source/Lib/Utilities/VideoIOYuv.cpp | 167 ++++++++++++++++++++++++++++
source/Lib/Utilities/VideoIOYuv.h | 19 +++-
7 files changed, 229 insertions(+), 6 deletions(-)
diff --git a/.clang-format b/.clang-format
index 46c86d74bc..b5ac8af9d2 100644
--- a/.clang-format
+++ b/.clang-format
@@ -79,7 +79,7 @@ MacroBlockBegin: ''
MacroBlockEnd: ''
MaxEmptyLinesToKeep: 1
NamespaceIndentation: All
-ObjCBinPackProtocolList: Auto
+# ObjCBinPackProtocolList: Auto
ObjCBlockIndentWidth: 2
ObjCSpaceAfterProperty: false
ObjCSpaceBeforeProtocolList: true
@@ -97,10 +97,10 @@ SortUsingDeclarations: true
SpaceAfterCStyleCast: true
SpaceAfterTemplateKeyword: false
SpaceBeforeAssignmentOperators: true
-SpaceBeforeCtorInitializerColon: true
-SpaceBeforeInheritanceColon: true
+# SpaceBeforeCtorInitializerColon: true
+# SpaceBeforeInheritanceColon: true
SpaceBeforeParens: ControlStatements
-SpaceBeforeRangeBasedForLoopColon: false
+# SpaceBeforeRangeBasedForLoopColon: false
SpaceInEmptyParentheses: false
SpacesBeforeTrailingComments: 3
SpacesInAngles: false
diff --git a/source/App/DecoderApp/DecApp.cpp b/source/App/DecoderApp/DecApp.cpp
index 30c5a825d1..62364e84aa 100644
--- a/source/App/DecoderApp/DecApp.cpp
+++ b/source/App/DecoderApp/DecApp.cpp
@@ -418,6 +418,14 @@ uint32_t DecApp::decode()
}
if( ( m_cDecLib.getVPS() != nullptr && ( m_cDecLib.getVPS()->getMaxLayers() == 1 || xIsNaluWithinTargetOutputLayerIdSet( &nalu ) ) ) || m_cDecLib.getVPS() == nullptr )
{
+ if (isY4mFileExt(reconFileName))
+ {
+ const auto sps = pcListPic->front()->cs->sps;
+ const auto pps = pcListPic->front()->cs->pps;
+ m_cVideoIOYuvReconFile[nalu.m_nuhLayerId].setOutputY4mInfo(
+ pps->getPicWidthInLumaSamples(), pps->getPicHeightInLumaSamples(), m_outputFrameRate, m_outputBitDepth[0],
+ sps->getChromaFormatIdc());
+ }
m_cVideoIOYuvReconFile[nalu.m_nuhLayerId].open( reconFileName, true, m_outputBitDepth, m_outputBitDepth, bitDepths.recon ); // write mode
}
}
diff --git a/source/App/DecoderApp/DecAppCfg.cpp b/source/App/DecoderApp/DecAppCfg.cpp
index 8988da1f34..cf310da048 100644
--- a/source/App/DecoderApp/DecAppCfg.cpp
+++ b/source/App/DecoderApp/DecAppCfg.cpp
@@ -40,6 +40,7 @@
#include <string>
#include "DecAppCfg.h"
#include "Utilities/program_options_lite.h"
+#include "Utilities/VideoIOYuv.h"
#include "CommonLib/ChromaFormat.h"
#include "CommonLib/dtrace_next.h"
@@ -87,6 +88,7 @@ bool DecAppCfg::parseCfg( int argc, char* argv[] )
("SkipFrames,s", m_iSkipFrame, 0, "number of frames to skip before random access")
("OutputBitDepth,d", m_outputBitDepth[CHANNEL_TYPE_LUMA], 0, "bit depth of YUV output luma component (default: use 0 for native depth)")
("OutputBitDepthC,d", m_outputBitDepth[CHANNEL_TYPE_CHROMA], 0, "bit depth of YUV output chroma component (default: use luma output bit-depth)")
+ ("OutputFrameRate", m_outputFrameRate, 0, "output frame rate, used to generate decoded Y4M")
("OutputColourSpaceConvert", outputColourSpaceConvert, string(""), "Colour space conversion to apply to input 444 video. Permitted values are (empty string=UNCHANGED) " + getListOfColourSpaceConverts(false))
("MaxTemporalLayer,t", m_iMaxTemporalLayer, 500, "Maximum Temporal Layer to be decoded. -1 to decode all layers")
("TargetOutputLayerSet,p", m_targetOlsIdx, 500, "Target output layer set index")
@@ -245,6 +247,12 @@ bool DecAppCfg::parseCfg( int argc, char* argv[] )
{
m_targetOlsIdx = -1;
}
+
+ if (m_outputFrameRate == 0 && isY4mFileExt(m_reconFileName))
+ {
+ msg(ERROR, "OutputFrameRate needs to be set when outputting to Y4M file\n");
+ return false;
+ }
return true;
}
diff --git a/source/App/DecoderApp/DecAppCfg.h b/source/App/DecoderApp/DecAppCfg.h
index f9af97fb68..e5ae2ddc17 100644
--- a/source/App/DecoderApp/DecAppCfg.h
+++ b/source/App/DecoderApp/DecAppCfg.h
@@ -63,6 +63,7 @@ protected:
int m_iSkipFrame; ///< counter for frames prior to the random access point to skip
int m_outputBitDepth[MAX_NUM_CHANNEL_TYPE]; ///< bit depth used for writing output
+ int m_outputFrameRate = 0; ///< used to generate decoded Y4M
InputColourSpaceConversion m_outputColourSpaceConvert;
int m_targetOlsIdx; ///< target output layer set
std::vector<int> m_targetOutputLayerIdSet; ///< set of LayerIds to be outputted
diff --git a/source/App/EncoderApp/EncApp.cpp b/source/App/EncoderApp/EncApp.cpp
index a0e4ab3380..c3a2ad1c5b 100644
--- a/source/App/EncoderApp/EncApp.cpp
+++ b/source/App/EncoderApp/EncApp.cpp
@@ -1334,6 +1334,11 @@ void EncApp::xCreateLib( std::list<PelUnitBuf*>& recBufList, const int layerId )
reconFileName.append( std::to_string( layerId ) );
}
}
+ if (isY4mFileExt(reconFileName))
+ {
+ m_cVideoIOYuvReconFile.setOutputY4mInfo(m_sourceWidth, m_sourceHeight, m_iFrameRate, m_internalBitDepth[0],
+ m_chromaFormatIDC);
+ }
m_cVideoIOYuvReconFile.open( reconFileName, true, m_outputBitDepth, m_outputBitDepth, m_internalBitDepth ); // write mode
}
@@ -1380,6 +1385,25 @@ void EncApp::xInitLib()
void EncApp::createLib( const int layerIdx )
{
+ if (isY4mFileExt(m_inputFileName))
+ {
+ int width = 0, height = 0, frameRate = 0, inputBitDepth = 0;
+ ChromaFormat chromaFormat = CHROMA_420;
+ m_cVideoIOYuvInputFile.parseY4mFileHeader(m_inputFileName, width, height, frameRate, inputBitDepth, chromaFormat);
+ if (width != m_sourceWidth || height != m_sourceHeight || frameRate != m_iFrameRate
+ || inputBitDepth != m_inputBitDepth[0] || chromaFormat != m_chromaFormatIDC)
+ {
+ printf("Warning: Y4M file info is different from input.\n");
+ m_sourceWidth = width;
+ m_sourceHeight = height;
+ m_iFrameRate = frameRate;
+ m_inputBitDepth[0] = inputBitDepth;
+ m_inputBitDepth[1] = inputBitDepth;
+ m_chromaFormatIDC = chromaFormat;
+ m_MSBExtendedBitDepth[0] = m_inputBitDepth[0];
+ m_MSBExtendedBitDepth[1] = m_inputBitDepth[1];
+ }
+ }
const int sourceHeight = m_isField ? m_iSourceHeightOrg : m_sourceHeight;
UnitArea unitArea( m_chromaFormatIDC, Area( 0, 0, m_sourceWidth, sourceHeight ) );
diff --git a/source/Lib/Utilities/VideoIOYuv.cpp b/source/Lib/Utilities/VideoIOYuv.cpp
index 9b98e092d1..836c87c308 100644
--- a/source/Lib/Utilities/VideoIOYuv.cpp
+++ b/source/Lib/Utilities/VideoIOYuv.cpp
@@ -50,6 +50,12 @@ using namespace std;
#define FLIP_PIC 0
+constexpr int Y4M_SIGNATURE_LENGTH = 10;
+const char y4mSignature[] = "YUV4MPEG2 ";
+constexpr int Y4M_MAX_HEADER_LENGTH = 128;
+constexpr int Y4M_FRAME_HEADER_LENGTH = 6; // basic Y4m frame header, "FRAME" + '\n'
+const char y4mFrameHeader[] = { 'F', 'R', 'A', 'M', 'E', '\n' };
+
// ====================================================================================================================
// Local Functions
// ====================================================================================================================
@@ -154,20 +160,158 @@ void VideoIOYuv::open( const std::string &fileName, bool bWriteMode, const int f
{
EXIT( "Failed to write reconstructed YUV file: " << fileName.c_str() );
}
+ if (isY4mFileExt(fileName))
+ {
+ writeY4mFileHeader();
+ m_outY4m = true;
+ }
}
else
{
+ if (isY4mFileExt(fileName))
+ {
+ if (m_inY4mFileHeaderLength == 0)
+ {
+ int dummyWidth = 0, dummyHeight = 0, dummyFrameRate = 0, dummyBitDepth = 0;
+ ChromaFormat dummyChromaFormat = CHROMA_420;
+ parseY4mFileHeader(fileName, dummyWidth, dummyHeight, dummyFrameRate, dummyBitDepth, dummyChromaFormat);
+ }
+ }
m_cHandle.open( fileName.c_str(), ios::binary | ios::in );
if( m_cHandle.fail() )
{
EXIT( "Failed to open input YUV file: " << fileName.c_str() );
}
+
+ if (m_inY4mFileHeaderLength)
+ {
+ m_cHandle.seekg(m_inY4mFileHeaderLength, ios::cur);
+ }
}
return;
}
+void VideoIOYuv::parseY4mFileHeader(const std::string &fileName, int &width, int &height, int &frameRate, int &bitDepth,
+ ChromaFormat &chromaFormat)
+{
+ m_cHandle.open(fileName.c_str(), ios::binary | ios::in);
+ CHECK(m_cHandle.fail(), "File open failed.")
+
+ char header[Y4M_MAX_HEADER_LENGTH];
+ m_cHandle.read(header, sizeof(header));
+ CHECK(strncmp(header, y4mSignature, Y4M_SIGNATURE_LENGTH), "The input is not a Y4M file!");
+
+ // locate the end of the header
+ for (int i = Y4M_SIGNATURE_LENGTH + 1; i < Y4M_MAX_HEADER_LENGTH; i++)
+ {
+ if (header[i] == '\n')
+ {
+ header[i] = ' '; // space is used as token end later
+ m_inY4mFileHeaderLength = i + 1;
+ break;
+ }
+ }
+ // parse Y4M header info
+ for (int i = Y4M_SIGNATURE_LENGTH; i < m_inY4mFileHeaderLength; i++)
+ {
+ int numerator = 0, denominator = 0, pos = 0;
+ switch (header[i])
+ {
+ case 'W': sscanf(header + i + 1, "%d", &width); break;
+ case 'H': sscanf(header + i + 1, "%d", &height); break;
+ case 'C':
+ if (strncmp(&header[i + 1], "mono", 4) == 0)
+ {
+ chromaFormat = CHROMA_400;
+ pos = i + 5;
+ }
+ else if (strncmp(&header[i + 1], "420", 3) == 0)
+ {
+ chromaFormat = CHROMA_420;
+ pos = i + 4;
+ if (strncmp(&header[pos], "jpeg", 4) == 0)
+ {
+ pos += 4;
+ }
+ else if (strncmp(&header[pos], "paldv", 5) == 0)
+ {
+ pos += 5;
+ }
+ }
+ else if (strncmp(&header[i + 1], "422", 3) == 0)
+ {
+ chromaFormat = CHROMA_422;
+ pos = i + 4;
+ }
+ else if (strncmp(&header[i + 1], "444", 3) == 0)
+ {
+ chromaFormat = CHROMA_444;
+ pos = i + 4;
+ }
+ bitDepth = 8;
+ if (header[pos] == 'p')
+ {
+ sscanf(&header[pos + 1], "%d", &bitDepth);
+ }
+ break;
+ case 'F':
+ if (sscanf(header + i + 1, "%d:%d", &numerator, &denominator) == 2)
+ {
+ if (denominator != 0)
+ {
+ frameRate = (int) (1.0 * numerator / denominator + 0.5);
+ }
+ }
+ break;
+ case 'I': CHECK(header[i + 1] != 'p', "Interlaced Y4M is not supported yet");
+ case 'A': // not support, ignore
+ case 'X': // not support, ignore
+ break;
+ default: CHECK(true, "Wrong Y4M file header!")
+ }
+ i = (int) (strchr(header + i + 1, ' ') - header);
+ }
+
+ m_cHandle.close();
+}
+
+void VideoIOYuv::setOutputY4mInfo(int width, int height, int frameRate, int bitDepth, ChromaFormat chromaFormat)
+{
+ m_outPicWidth = width;
+ m_outPicHeight = height;
+ m_outBitDepth = bitDepth;
+ m_outFrameRate = frameRate;
+ m_outChromaFormat = chromaFormat;
+}
+
+void VideoIOYuv::writeY4mFileHeader()
+{
+ CHECK(m_outPicWidth == 0 || m_outPicHeight == 0 || m_outBitDepth == 0 || m_outFrameRate == 0,
+ "Output Y4M file into has not been set");
+ std::string header = y4mSignature;
+ header += "W" + std::to_string(m_outPicWidth) + " ";
+ header += "H" + std::to_string(m_outPicHeight) + " ";
+ header += "F" + std::to_string(m_outFrameRate) + ":1 ";
+ header += "Ip A0:0 ";
+ switch (m_outChromaFormat)
+ {
+ case CHROMA_400: header += "Cmono"; break;
+ case CHROMA_420: header += "C420"; break;
+ case CHROMA_422: header += "C422"; break;
+ case CHROMA_444: header += "C444"; break;
+ }
+ if (m_outBitDepth > 8)
+ {
+ header += "p" + std::to_string(m_outBitDepth);
+ }
+ header += "\n";
+ // not write extension/comment
+
+ m_cHandle.write(header.c_str(), header.length());
+}
+
void VideoIOYuv::close()
{
m_cHandle.close();
@@ -215,6 +359,10 @@ void VideoIOYuv::skipFrames(uint32_t numFrames, uint32_t width, uint32_t height,
}
frameSize *= wordsize;
//------------------
+ if (m_inY4mFileHeaderLength)
+ {
+ frameSize += Y4M_FRAME_HEADER_LENGTH;
+ }
const streamoff offset = frameSize * numFrames;
@@ -912,6 +1060,13 @@ bool VideoIOYuv::read ( PelUnitBuf& pic, PelUnitBuf& picOrg, const InputColourSp
}
}
+ if (m_inY4mFileHeaderLength)
+ {
+ char y4mFrameHeader[Y4M_FRAME_HEADER_LENGTH];
+ m_cHandle.read(y4mFrameHeader, Y4M_FRAME_HEADER_LENGTH);
+ CHECK(strncmp(y4mFrameHeader, "FRAME", Y4M_FRAME_HEADER_LENGTH - 1), "Wrong Y4M frame header!");
+ }
+
const PelBuf areaBufY = picOrg.get(COMPONENT_Y);
#if !EXTENSION_360_VIDEO
const uint32_t stride444 = areaBufY.stride;
@@ -1060,6 +1215,11 @@ bool VideoIOYuv::write( uint32_t orgWidth, uint32_t orgHeight, const CPelUnitBuf
msg( WARNING, "\nWarning: writing %d x %d luma sample output picture!", width444, height444);
}
+ if (m_outY4m)
+ {
+ m_cHandle.write(y4mFrameHeader, Y4M_FRAME_HEADER_LENGTH);
+ }
+
for(uint32_t comp=0; retval && comp < ::getNumberValidComponents(format); comp++)
{
const ComponentID compID = ComponentID(comp);
@@ -1325,3 +1485,10 @@ bool VideoIOYuv::writeUpscaledPicture( const SPS& sps, const PPS& pps, const CPe
return ret;
}
+
+bool isY4mFileExt(const std::string &fileName)
+{
+ auto pos = fileName.find(".y4m");
+ // ".y4m" must be at the end of the file name
+ return (pos != std::string::npos && pos + 4 == fileName.length());
+}
diff --git a/source/Lib/Utilities/VideoIOYuv.h b/source/Lib/Utilities/VideoIOYuv.h
index b18d1f8eff..48697fda1f 100644
--- a/source/Lib/Utilities/VideoIOYuv.h
+++ b/source/Lib/Utilities/VideoIOYuv.h
@@ -61,13 +61,26 @@ private:
int m_fileBitdepth[MAX_NUM_CHANNEL_TYPE]; ///< bitdepth of input/output video file
int m_MSBExtendedBitDepth[MAX_NUM_CHANNEL_TYPE]; ///< bitdepth after addition of MSBs (with value 0)
int m_bitdepthShift[MAX_NUM_CHANNEL_TYPE]; ///< number of bits to increase or decrease image by before/after write/read
+ int m_inY4mFileHeaderLength = 0;
+ int m_outPicWidth = 0;
+ int m_outPicHeight = 0;
+ int m_outBitDepth = 0;
+ int m_outFrameRate = 0;
+ ChromaFormat m_outChromaFormat = CHROMA_420;
+ bool m_outY4m = false;
public:
VideoIOYuv() {}
virtual ~VideoIOYuv() {}
- void open ( const std::string &fileName, bool bWriteMode, const int fileBitDepth[MAX_NUM_CHANNEL_TYPE], const int MSBExtendedBitDepth[MAX_NUM_CHANNEL_TYPE], const int internalBitDepth[MAX_NUM_CHANNEL_TYPE] ); ///< open or create file
- void close (); ///< close file
+ void parseY4mFileHeader(const std::string &fileName, int &width, int &height, int &frameRate, int &bitDepth,
+ ChromaFormat &chromaFormat);
+ void setOutputY4mInfo(int width, int height, int frameRate, int bitDepth, ChromaFormat chromaFormat);
+ void writeY4mFileHeader();
+ void open(const std::string &fileName, bool bWriteMode, const int fileBitDepth[MAX_NUM_CHANNEL_TYPE],
+ const int MSBExtendedBitDepth[MAX_NUM_CHANNEL_TYPE],
+ const int internalBitDepth[MAX_NUM_CHANNEL_TYPE]); ///< open or create file
+ void close(); ///< close file
#if EXTENSION_360_VIDEO
void skipFrames(int numFrames, uint32_t width, uint32_t height, ChromaFormat format);
#else
@@ -105,5 +118,7 @@ public:
};
+bool isY4mFileExt(const std::string &fileName);
+
#endif // __VIDEOIOYUV__
--
GitLab