Merge pull request #286 from Eyevinn/fix-avc-slice-header

Fix avc ChromeArrayType

CHANGELOG.md

@@ -7,7 +7,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-- nothing yet
+### Fixed
+
+- SPS.ChromaArrayType method
+
+### Added
+
+- AccErrEBSPReader.NrBitsRead method
 
 ## [0.39.0] - 2023-10-13
 

avc/slice.go

@@ -104,6 +104,7 @@ type SliceHeader struct {
 	ModificationOfPicNumsIDC      uint32
 	AbsDiffPicNumMinus1           uint32
 	LongTermPicNum                uint32
+	AbsDiffViewIdxMinus1          uint32
 	LumaLog2WeightDenom           uint32
 	ChromaLog2WeightDenom         uint32
 	DifferenceOfPicNumsMinus1     uint32
@@ -129,6 +130,7 @@ type SliceHeader struct {
 	AdaptiveRefPicMarkingModeFlag bool
 }
 
+// ParseSliceHeader parses AVC slice header following the syntax in ISO/IEC 14496-10 section 7.3.3
 func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PPS) (*SliceHeader, error) {
 	sh := SliceHeader{}
 	buf := bytes.NewBuffer(nalu)
@@ -204,7 +206,7 @@ func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PP
 		}
 	}
 
-	// ref_pic_list_modification (nal unit type != 20)
+	// ref_pic_list_modification (nal unit type != 20 or 21) Section G.3.3.3.1.1
 	if sliceType != SLICE_I && sliceType != SLICE_SI {
 		sh.RefPicListModificationL0Flag = r.ReadFlag()
 		if sh.RefPicListModificationL0Flag {
@@ -216,6 +218,8 @@ func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PP
 					sh.AbsDiffPicNumMinus1 = uint32(r.ReadExpGolomb())
 				case 2:
 					sh.LongTermPicNum = uint32(r.ReadExpGolomb())
+				case 4, 5:
+					sh.AbsDiffViewIdxMinus1 = uint32(r.ReadExpGolomb())
 				case 3:
 					break refPicListL0Loop
 				}
@@ -236,6 +240,8 @@ func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PP
 					sh.AbsDiffPicNumMinus1 = uint32(r.ReadExpGolomb())
 				case 2:
 					sh.LongTermPicNum = uint32(r.ReadExpGolomb())
+				case 4, 5:
+					sh.AbsDiffViewIdxMinus1 = uint32(r.ReadExpGolomb())
 				case 3:
 					break refPicListL1Loop
 				}
@@ -249,9 +255,9 @@ func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PP
 
 	if pps.WeightedPredFlag && (sliceType == SLICE_P || sliceType == SLICE_SP) ||
 		(pps.WeightedBipredIDC == 1 && sliceType == SLICE_B) {
-		// pred_weight_table
+		// pred_weight_table, section 7.3.3.2
 		sh.LumaLog2WeightDenom = uint32(r.ReadExpGolomb())
-		if sps.ChromaArrayType() != 0 {
+		if sps.ChromaArrayType() != 0 { // chroma_idc != 0 in Bento4
 			sh.ChromaLog2WeightDenom = uint32(r.ReadExpGolomb())
 		}
 
@@ -281,9 +287,9 @@ func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PP
 					_ = r.ReadExpGolomb() // luma_weight_l1[i] = SignedGolomb()
 					_ = r.ReadExpGolomb() // luma_offset_l1[i] = SignedGolomb()
 				}
-				if sps.ChromaFormatIDC != 0 {
-					chromaWeightL0Flag := r.ReadFlag()
-					if chromaWeightL0Flag {
+				if sps.ChromaArrayType() != 0 {
+					chromaWeightL1Flag := r.ReadFlag()
+					if chromaWeightL1Flag {
 						// Just parse, don't store this
 						for j := 0; j < 2; j++ {
 							_ = r.ReadSignedGolomb() // chroma_weight_l1[i][j] = SignedGolomb()
@@ -350,12 +356,12 @@ func ParseSliceHeader(nalu []byte, spsMap map[uint32]*SPS, ppsMap map[uint32]*PP
 		pps.SliceGroupMapType >= 3 &&
 		pps.SliceGroupMapType <= 5 {
 		picSizeInMapUnits := pps.PicSizeInMapUnitsMinus1 + 1
-		sliceGroupChangeRage := pps.SliceGroupChangeRateMinus1 + 1
-		nrBits := int(math.Ceil(math.Log2(float64(picSizeInMapUnits/sliceGroupChangeRage + 1))))
+		sliceGroupChangeRate := pps.SliceGroupChangeRateMinus1 + 1
+		nrBits := int(math.Ceil(math.Log2(float64(picSizeInMapUnits/sliceGroupChangeRate + 1))))
 		sh.SliceGroupChangeCycle = uint32(r.Read(nrBits))
 	}
 
-	// compute the size in bytes. The last byte may not be fully read
+	// compute the size in bytes. The last byte may not be fully parsed
 	sh.Size = uint32(r.NrBytesRead())
 	return &sh, nil
 }

avc/slice_test.go

@@ -116,3 +116,32 @@ func TestParseSliceHeader_TwoFrames(t *testing.T) {
 		t.Errorf("Got NON_IDR Slice Header: %+v\n Diff is: %v", *gotNonIdrHdr, diff)
 	}
 }
+
+func TestParseSliceHeaderLength(t *testing.T) {
+	spsHex := "6764001eacd940a02ff9610000030001000003003c8f162d96"
+	ppsHex := "68ebecb22c"
+	naluStartHex := "419a6649e10f2653022fff8700000302c8a32d32"
+	spsData, _ := hex.DecodeString(spsHex)
+	sps, err := ParseSPSNALUnit(spsData, true)
+	if err != nil {
+		t.Error(err)
+	}
+	spsMap := make(map[uint32]*SPS, 1)
+	spsMap[uint32(sps.ParameterID)] = sps
+	ppsData, _ := hex.DecodeString(ppsHex)
+	pps, err := ParsePPSNALUnit(ppsData, spsMap)
+	if err != nil {
+		t.Error(err)
+	}
+	ppsMap := make(map[uint32]*PPS, 1)
+	ppsMap[uint32(pps.PicParameterSetID)] = pps
+	naluStart, _ := hex.DecodeString(naluStartHex)
+	sh, err := ParseSliceHeader(naluStart, spsMap, ppsMap)
+	if err != nil {
+		t.Error(err)
+	}
+	wantedSliceHeaderSize := uint32(11)
+	if sh.Size != wantedSliceHeaderSize {
+		t.Errorf("got %d want %d", sh.Size, wantedSliceHeaderSize)
+	}
+}

avc/sps.go

@@ -261,9 +261,9 @@ func (s *SPS) PicStructPresent() bool {
 	return s.VUI.PicStructPresentFlag
 }
 
-// ChromaArrayType as defined in Section 7.4.2.1.1
+// ChromaArrayType as defined in Section 7.4.2.1.1 under separate_colour_plane_flag
 func (s *SPS) ChromaArrayType() byte {
-	if s.SeparateColourPlaneFlag {
+	if !s.SeparateColourPlaneFlag {
 		return byte(s.ChromaFormatIDC)
 	}
 	return 0

bits/aeebspreader.go

@@ -34,6 +34,15 @@ func (r *AccErrEBSPReader) NrBytesRead() int {
 	return r.pos + 1 // Starts at -1
 }
 
+// NrBitsRead - how many bits read into parser
+func (r *AccErrEBSPReader) NrBitsRead() int {
+	nrBits := r.NrBytesRead() * 8
+	if r.NrBitsReadInCurrentByte() != 8 {
+		nrBits += r.NrBitsReadInCurrentByte() - 8
+	}
+	return nrBits
+}
+
 // NrBitsReadInCurrentByte - how many bits have been read
 func (r *AccErrEBSPReader) NrBitsReadInCurrentByte() int {
 	return 8 - r.n

bits/aeebspreader_test.go

@@ -10,18 +10,19 @@ import (
 
 func TestAccErrEBSPReader(t *testing.T) {
 	testCases := []struct {
-		hexData             string
-		readBits            int
-		expectedValue       uint
-		expectedNrBytesRead int
-		expectedNrBitsRead  int
+		hexData                         string
+		readBits                        int
+		expectedValue                   uint
+		expectedNrBytesRead             int
+		expectedNrBitsRead              int
+		expectedNrBitsReadInCurrentByte int
 	}{
-		{"00", 0, 0, 0, 8},
-		{"0001", 1, 0, 1, 1},
-		{"0001", 8, 0x00, 1, 8},
-		{"0010", 12, 0x001, 2, 4},
-		{"0102", 16, 0x0102, 2, 8},
-		{"00000304", 24, 0x000004, 4, 8},
+		{"00", 0, 0, 0, 0, 8},
+		{"0001", 1, 0, 1, 1, 1},
+		{"0001", 8, 0x00, 1, 8, 8},
+		{"0010", 12, 0x001, 2, 12, 4},
+		{"0102", 16, 0x0102, 2, 16, 8},
+		{"00000304", 24, 0x000004, 4, 32, 8}, // Note the start-code emulation 0x03
 	}
 	for _, tc := range testCases {
 		t.Run("", func(t *testing.T) {
@@ -41,8 +42,11 @@ func TestAccErrEBSPReader(t *testing.T) {
 			if r.NrBytesRead() != tc.expectedNrBytesRead {
 				t.Errorf("expected %d bytes read, got %d", tc.expectedNrBytesRead, r.NrBytesRead())
 			}
-			if r.NrBitsReadInCurrentByte() != tc.expectedNrBitsRead {
-				t.Errorf("expected %d bits read, got %d", tc.expectedNrBitsRead, r.NrBitsReadInCurrentByte())
+			if r.NrBitsRead() != tc.expectedNrBitsRead {
+				t.Errorf("expected %d bits read, got %d", tc.expectedNrBitsRead, r.NrBitsRead())
+			}
+			if r.NrBitsReadInCurrentByte() != tc.expectedNrBitsReadInCurrentByte {
+				t.Errorf("expected %d bits read in current byte, got %d", tc.expectedNrBitsReadInCurrentByte, r.NrBitsReadInCurrentByte())
 			}
 		})
 	}