Parse Slice headers for multiple layers in HEVC and VVC

IENT · Feb 20, 2024 · 0bad52e · 0bad52e
1 parent 35ff669
commit 0bad52e
Show file tree

Hide file tree

Showing 10 changed files with 67 additions and 24 deletions.
diff --git a/YUViewLib/src/common/Typedef.h b/YUViewLib/src/common/Typedef.h
@@ -218,6 +218,8 @@ typedef QPair<unsigned int, unsigned int> QUIntPair;
 template <typename T> using umap_1d = std::map<unsigned, T>;
 template <typename T> using umap_2d = std::map<unsigned, umap_1d<T>>;
 template <typename T> using umap_3d = std::map<unsigned, umap_2d<T>>;
+template <typename T> using umap_4d = std::map<unsigned, umap_3d<T>>;
+template <typename T> using umap_5d = std::map<unsigned, umap_4d<T>>;
 
 template <typename T> using vector   = std::vector<T>;
 template <typename T> using vector2d = std::vector<vector<T>>;
@@ -226,6 +228,8 @@ template <typename T> using vector4d = std::vector<vector3d<T>>;
 
 template <typename T, size_t N> using array               = std::array<T, N>;
 template <typename T, size_t N1, size_t N2> using array2d = std::array<std::array<T, N2>, N1>;
+template <typename T, size_t N1, size_t N2, size_t N3>
+using array3d = std::array<std::array<std::array<T, N3>, N2>, N1>;
 
 template <typename T> int indexInVec(const std::vector<T> &vec, const T &item)
 {

diff --git a/YUViewLib/src/parser/AVC/ParserAnnexBAVC.cpp b/YUViewLib/src/parser/AVC/ParserAnnexBAVC.cpp
@@ -193,7 +193,8 @@ ParserAnnexBAVC::parseAndAddNALUnit(int
       // Save the info of the last frame
       if (!this->addFrameToList(*this->curFrameData->poc,
                                 this->curFrameData->fileStartEndPos,
-                                this->curFrameData->isRandomAccess))
+                                this->curFrameData->isRandomAccess,
+                                0))
       {
         if (parent)
           parent->createChildItem("Error - POC " + std::to_string(*this->curFrameData->poc) +
@@ -481,7 +482,8 @@ ParserAnnexBAVC::parseAndAddNALUnit(int
       // Save the info of the last frame
       if (!this->addFrameToList(*this->curFrameData->poc,
                                 this->curFrameData->fileStartEndPos,
-                                this->curFrameData->isRandomAccess))
+                                this->curFrameData->isRandomAccess,
+                                0))
       {
         throw std::logic_error("Error - POC " + std::to_string(*this->curFrameData->poc) +
                                " already in the POC list");

diff --git a/YUViewLib/src/parser/HEVC/ParserAnnexBHEVC.cpp b/YUViewLib/src/parser/HEVC/ParserAnnexBHEVC.cpp
@@ -283,18 +283,21 @@ ParserAnnexBHEVC::parseAndAddNALUnit(int
     if (curFramePOC != -1)
     {
       // Save the info of the last frame
-      if (!this->addFrameToList(curFramePOC, curFrameFileStartEndPos, curFrameIsRandomAccess))
+      if (!this->addFrameToList(
+              curFramePOC, curFrameFileStartEndPos, curFrameIsRandomAccess, curFrameLayerID))
       {
         throw std::logic_error("Error - POC " + std::to_string(curFramePOC) +
                                " alread in the POC list.");
       }
       if (curFrameFileStartEndPos)
         DEBUG_HEVC("ParserAnnexBHEVC::parseAndAddNALUnit Adding start/end "
                    << curFrameFileStartEndPos->first << "/" << curFrameFileStartEndPos->second
-                   << " - POC " << curFramePOC << (curFrameIsRandomAccess ? " - ra" : ""));
+                   << " - POC " << curFramePOC << " layer " << curFrameLayerID
+                   << (curFrameIsRandomAccess ? " - ra" : ""));
       else
         DEBUG_HEVC("ParserAnnexBHEVC::parseAndAddNALUnit Adding start/end NA/NA - POC "
-                   << curFramePOC << (curFrameIsRandomAccess ? " - ra" : ""));
+                   << curFramePOC << " layer " << curFrameLayerID
+                   << (curFrameIsRandomAccess ? " - ra" : ""));
     }
     // The file ended
     return parseResult;
@@ -445,22 +448,26 @@ ParserAnnexBHEVC::parseAndAddNALUnit(int
         if (curFramePOC != -1)
         {
           // Save the info of the last frame
-          if (!this->addFrameToList(curFramePOC, curFrameFileStartEndPos, curFrameIsRandomAccess))
+          if (!this->addFrameToList(
+                  curFramePOC, curFrameFileStartEndPos, curFrameIsRandomAccess, curFrameLayerID))
           {
             throw std::logic_error("Error - POC " + std::to_string(curFramePOC) +
                                    " alread in the POC list.");
           }
           if (curFrameFileStartEndPos)
             DEBUG_HEVC("ParserAnnexBHEVC::parseAndAddNALUnit Adding start/end "
                        << curFrameFileStartEndPos->first << "/" << curFrameFileStartEndPos->second
-                       << " - POC " << curFramePOC << (curFrameIsRandomAccess ? " - ra" : ""));
+                       << " - POC " << curFramePOC << curFramePOC << " layer "
+                       << (curFrameIsRandomAccess ? " - ra" : ""));
           else
             DEBUG_HEVC("ParserAnnexBHEVC::parseAndAddNALUnit Adding start/end NA/NA - POC "
-                       << curFramePOC << (curFrameIsRandomAccess ? " - ra" : ""));
+                       << curFramePOC << curFramePOC << " layer "
+                       << (curFrameIsRandomAccess ? " - ra" : ""));
         }
         curFrameFileStartEndPos = nalStartEndPosFile;
         curFramePOC             = newSlice->sliceSegmentHeader.globalPOC;
         curFrameIsRandomAccess  = nalHEVC->header.isIRAP();
+        curFrameLayerID         = nalHEVC->header.nuh_layer_id;
 
         {
           std::shared_ptr<seq_parameter_set_rbsp> refSPS;

diff --git a/YUViewLib/src/parser/HEVC/ParserAnnexBHEVC.h b/YUViewLib/src/parser/HEVC/ParserAnnexBHEVC.h
@@ -138,8 +138,9 @@ class ParserAnnexBHEVC : public ParserAnnexB
                                // in case the frame has multiple NAL units
   // The POC of the current frame. We save this we encounter a NAL from the next POC; then we add
   // it.
-  int  curFramePOC{-1};
-  bool curFrameIsRandomAccess{false};
+  int      curFramePOC{-1};
+  bool     curFrameIsRandomAccess{false};
+  unsigned curFrameLayerID{0};
 
   hevc::AUDelimiterDetector auDelimiterDetector;
 

diff --git a/YUViewLib/src/parser/HEVC/slice_segment_header.cpp b/YUViewLib/src/parser/HEVC/slice_segment_header.cpp
@@ -97,7 +97,18 @@ void slice_segment_header::parse(SubByteReaderLogging & reader,
 
   if (!dependent_slice_segment_flag)
   {
-    for (unsigned int i = 0; i < pps->num_extra_slice_header_bits; i++)
+    unsigned i = 0;
+    if (pps->num_extra_slice_header_bits > i)
+    {
+      i++;
+      this->discardable_flag = reader.readFlag("discardable_flag");
+    }
+    if (pps->num_extra_slice_header_bits > i)
+    {
+      i++;
+      this->cross_layer_bla_flag = reader.readFlag("cross_layer_bla_flag");
+    }
+    for (; i < pps->num_extra_slice_header_bits; i++)
       this->slice_reserved_flag.push_back(reader.readFlag(formatArray("slice_reserved_flag", i)));
 
     auto sliceTypeIdx = reader.readUEV(
@@ -174,6 +185,12 @@ void slice_segment_header::parse(SubByteReaderLogging & reader,
         this->slice_temporal_mvp_enabled_flag = reader.readFlag("slice_temporal_mvp_enabled_flag");
     }
 
+    // This is not 100% done yet.
+    // if (nalUnitHeader.nuh_layer_id > 0 && !default_ref_layers_active_flag &&
+    //     NumDirectRefLayers[nuh_layer_id] > 0)
+    // {
+    // }
+
     if (sps->sample_adaptive_offset_enabled_flag)
     {
       this->slice_sao_luma_flag = reader.readFlag("slice_sao_luma_flag");

diff --git a/YUViewLib/src/parser/HEVC/slice_segment_header.h b/YUViewLib/src/parser/HEVC/slice_segment_header.h
@@ -75,6 +75,8 @@ class slice_segment_header
   bool              dependent_slice_segment_flag{};
   unsigned          slice_pic_parameter_set_id{};
   unsigned          slice_segment_address{};
+  bool              discardable_flag{};
+  bool              cross_layer_bla_flag{};
   vector<bool>      slice_reserved_flag;
   SliceType         slice_type{};
   bool              pic_output_flag{true};

diff --git a/YUViewLib/src/parser/ParserAnnexB.cpp b/YUViewLib/src/parser/ParserAnnexB.cpp
@@ -62,10 +62,11 @@ std::string ParserAnnexB::getShortStreamDescription(const int) const
 
 bool ParserAnnexB::addFrameToList(int                       poc,
                                   std::optional<pairUint64> fileStartEndPos,
-                                  bool                      randomAccessPoint)
+                                  bool                      randomAccessPoint,
+                                  unsigned                  layerID)
 {
   for (const auto &f : this->frameListCodingOrder)
-    if (f.poc == poc)
+    if (f.poc == poc && f.layerID == layerID)
       return false;
 
   if (!this->pocOfFirstRandomAccessFrame && randomAccessPoint)
@@ -77,6 +78,7 @@ bool ParserAnnexB::addFrameToList(int                       poc,
     newFrame.poc               = poc;
     newFrame.fileStartEndPos   = fileStartEndPos;
     newFrame.randomAccessPoint = randomAccessPoint;
+    newFrame.layerID           = layerID;
     this->frameListCodingOrder.push_back(newFrame);
     this->frameListDisplayOder.clear();
   }

diff --git a/YUViewLib/src/parser/ParserAnnexB.h b/YUViewLib/src/parser/ParserAnnexB.h
@@ -137,17 +137,20 @@ class ParserAnnexB : public Parser
   struct AnnexBFrame
   {
     AnnexBFrame() = default;
-    int poc{-1}; //< The poc of this frame
-    std::optional<pairUint64>
-         fileStartEndPos;          //< The start and end position of all slice NAL units (if known)
-    bool randomAccessPoint{false}; //< Can we start decoding here?
+    int                       poc{-1};
+    std::optional<pairUint64> fileStartEndPos;
+    bool                      randomAccessPoint{false};
+    unsigned                  layerID{};
 
     bool operator<(AnnexBFrame const &b) const { return (this->poc < b.poc); }
     bool operator==(AnnexBFrame const &b) const { return (this->poc == b.poc); }
   };
 
   // Returns false if the POC was already present int the list
-  bool addFrameToList(int poc, std::optional<pairUint64> fileStartEndPos, bool randomAccessPoint);
+  bool addFrameToList(int                       poc,
+                      std::optional<pairUint64> fileStartEndPos,
+                      bool                      randomAccessPoint,
+                      unsigned                  layerID);
 
   static void logNALSize(const ByteVector &        data,
                          std::shared_ptr<TreeItem> root,

diff --git a/YUViewLib/src/parser/VVC/ParserAnnexBVVC.cpp b/YUViewLib/src/parser/VVC/ParserAnnexBVVC.cpp
@@ -473,6 +473,7 @@ ParserAnnexBVVC::parseAndAddNALUnit(int
 
       nalVVC->rbsp = newSliceLayer;
 
+      updatedParsingState.currentAU.layerID = nalVVC->header.nuh_layer_id;
       updatedParsingState.currentAU.isKeyframe =
           (nalType == NalType::IDR_W_RADL || nalType == NalType::IDR_N_LP ||
            nalType == NalType::CRA_NUT);
@@ -596,22 +597,25 @@ int ParserAnnexBVVC::calculateAndUpdateGlobalPOC(bool isIRAP, unsigned PicOrderC
 
 bool ParserAnnexBVVC::handleNewAU(ParsingState &parsingState)
 {
-  DEBUG_VVC("Start of new AU. Adding bitrate " << parsingState.currentAU.sizeBytes << " POC "
-                                               << parsingState.currentAU.poc << " AU "
-                                               << parsingState.currentAU.counter);
+  DEBUG_VVC("Start of new AU. Adding bitrate "
+            << parsingState.currentAU.sizeBytes << " POC " << parsingState.currentAU.poc << " AU "
+            << parsingState.currentAU.counter << " Layer " << parsingState.currentAU.layerID);
 
   if (!this->addFrameToList(parsingState.currentAU.poc,
                             parsingState.currentAU.fileStartEndPos,
-                            parsingState.currentAU.isKeyframe))
+                            parsingState.currentAU.isKeyframe,
+                            parsingState.currentAU.layerID))
     return false;
 
   if (this->parsingState.currentAU.fileStartEndPos)
     DEBUG_VVC("Adding start/end " << parsingState.currentAU.fileStartEndPos->first << "/"
                                   << parsingState.currentAU.fileStartEndPos->second << " - AU "
-                                  << parsingState.currentAU.counter
+                                  << parsingState.currentAU.counter << " Layer "
+                                  << parsingState.currentAU.layerID
                                   << (parsingState.currentAU.isKeyframe ? " - ra" : ""));
   else
-    DEBUG_VVC("Adding start/end NA/NA - AU " << parsingState.currentAU.counter
+    DEBUG_VVC("Adding start/end NA/NA - AU " << parsingState.currentAU.counter << " Layer "
+                                             << parsingState.currentAU.layerID
                                              << (parsingState.currentAU.isKeyframe ? " - ra" : ""));
 
   return true;

diff --git a/YUViewLib/src/parser/VVC/ParserAnnexBVVC.h b/YUViewLib/src/parser/VVC/ParserAnnexBVVC.h
@@ -65,6 +65,7 @@ struct ParsingState
     size_t                    counter{};
     size_t                    sizeBytes{};
     int                       poc{-1};
+    unsigned                  layerID{0};
     bool                      isKeyframe{};
     std::optional<pairUint64> fileStartEndPos;
   };