Merge pull request #209 from Eyevinn/hevc-sps-fix

Fix HEVC SPS scaling list parsing
Eyevinn · Jan 25, 2023 · 0d2fcb7 · 0d2fcb7
2 parents 238c080 + d46d41f
commit 0d2fcb7
Show file tree

Hide file tree

Showing 3 changed files with 148 additions and 7 deletions.
diff --git a/hevc/hevcdecoderconfigurationrecord_test.go b/hevc/hevcdecoderconfigurationrecord_test.go
@@ -77,3 +77,54 @@ func createNalusFromHex(hexStr string) [][]byte {
 	}
 	return nalus
 }
+
+func TestDecodeConfRec(t *testing.T) {
+	hexData := ("0102" +
+		"00000020b000000000009c0000000102" +
+		"0200000f03a00001001840010c01ffff" +
+		"022000000300b0000003000003009c15" +
+		"c090a100010025420101022000000300" +
+		"b0000003000003009ca001e020021c4d" +
+		"8815ee4595602d4244024020a2000100" +
+		"094401c02864b8d05324")
+	data, err := hex.DecodeString(hexData)
+	if err != nil {
+		t.Error(err)
+	}
+	hdcr, err := DecodeHEVCDecConfRec(data)
+	if err != nil {
+		t.Error(err)
+	}
+	for _, na := range hdcr.NaluArrays {
+
+		switch na.NaluType() {
+		case NALU_VPS, NALU_PPS:
+		case NALU_SPS:
+			for _, nalu := range na.Nalus {
+				sps, err := ParseSPSNALUnit(nalu)
+				if err != nil {
+					t.Error(err)
+				}
+				if !sps.ScalingListEnabledFlag || sps.ScalingListDataPresentFlag {
+					t.Error("scaling_list_enabled not properly parsed")
+				}
+				vui := sps.VUI
+				if vui == nil {
+					t.Error("no vui parsed")
+				} else {
+					if vui.TransferCharacteristics != 16 {
+						t.Errorf("vui transfer_characteristics is %d, not 16", vui.TransferCharacteristics)
+					}
+					if vui.ColourPrimaries != 9 {
+						t.Errorf("vui colour_primaries is %d, not 9", vui.ColourPrimaries)
+					}
+					if vui.MatrixCoefficients != 9 {
+						t.Errorf("vui matrix_coefficients is %d, not 9", vui.MatrixCoefficients)
+					}
+				}
+			}
+		default:
+			t.Errorf("strange nalu type %s", na.NaluType())
+		}
+	}
+}
diff --git a/hevc/sps.go b/hevc/sps.go
@@ -130,7 +130,7 @@ type BitstreamRestrictions struct {
 	Log2MaxMvLengthVertical     uint
 }
 
-// ParseSPSNALUnit - Parse HEVC SPS NAL unit starting with NAL unit header
+// ParseSPSNALUnit parses SPS NAL unit starting with NAL unit header
 func ParseSPSNALUnit(data []byte) (*SPS, error) {
 
 	sps := &SPS{}
@@ -176,9 +176,9 @@ func ParseSPSNALUnit(data []byte) (*SPS, error) {
 	sps.BitDepthChromaMinus8 = byte(r.ReadExpGolomb())
 	sps.Log2MaxPicOrderCntLsbMinus4 = byte(r.ReadExpGolomb())
 	sps.SubLayerOrderingInfoPresentFlag = r.ReadFlag()
-	startValue := byte(0)
+	startValue := sps.MaxSubLayersMinus1
 	if sps.SubLayerOrderingInfoPresentFlag {
-		startValue = sps.MaxSubLayersMinus1
+		startValue = 0
 	}
 	for i := startValue; i <= sps.MaxSubLayersMinus1; i++ {
 		sps.SubLayeringOrderingInfos = append(
@@ -197,7 +197,10 @@ func ParseSPSNALUnit(data []byte) (*SPS, error) {
 	sps.MaxTransformHierarchyDepthIntra = byte(r.ReadExpGolomb())
 	sps.ScalingListEnabledFlag = r.ReadFlag()
 	if sps.ScalingListEnabledFlag {
-		readPastScalingList(r)
+		sps.ScalingListDataPresentFlag = r.ReadFlag()
+		if sps.ScalingListDataPresentFlag {
+			readPastScalingListData(r)
+		}
 	}
 	sps.AmpEnabledFlag = r.ReadFlag()
 	sps.SampleAdaptiveOffsetEnabledFlag = r.ReadFlag()
@@ -407,8 +410,8 @@ func parseShortTermRPS(r *bits.AccErrEBSPReader, idx, numSTRefPicSets byte, sps
 	return stps
 }
 
-// readPastScalingList - read and parse all bits of scaling list, without storing values
-func readPastScalingList(r *bits.AccErrEBSPReader) {
+// readPastScalingListData - read and parse all bits of scaling list, without storing values
+func readPastScalingListData(r *bits.AccErrEBSPReader) {
 	for sizeId := 0; sizeId < 4; sizeId++ {
 		nrMatrixIds := 6
 		if sizeId == 3 {

diff --git a/hevc/sps_test.go b/hevc/sps_test.go
@@ -8,7 +8,9 @@ import (
 )
 
 const (
-	spsNalu = "420101022000000300b0000003000003007ba0078200887db6718b92448053888892cf24a69272c9124922dc91aa48fca223ff000100016a02020201"
+	spsNalu = ("420101022000000300b0000003000003007ba0078200887db6718b92448053888892" +
+		"cf24a69272c9124922dc91aa48fca223ff000100016a02020201")
+	spsNaluHdr10 = "420101022000000300b0000003000003009ca001e020021c4d8815ee4595602d4244024020"
 )
 
 func TestSPSParser1(t *testing.T) {
@@ -178,3 +180,88 @@ func TestSPSParser1(t *testing.T) {
 		t.Errorf("Got %dx%d instead of %dx%d", gotWidth, gotHeight, expWidth, expHeight)
 	}
 }
+
+func TestSPSParser2(t *testing.T) {
+	byteData, _ := hex.DecodeString(spsNaluHdr10)
+
+	wantedVUI := VUIParameters{
+		SampleAspectRatioWidth:     1,
+		SampleAspectRatioHeight:    1,
+		VideoSignalTypePresentFlag: true,
+		VideoFormat:                5,
+		ColourDescriptionFlag:      true,
+		ColourPrimaries:            9,
+		TransferCharacteristics:    16,
+		MatrixCoefficients:         9,
+	}
+	wanted := SPS{
+		VpsID:                 0,
+		MaxSubLayersMinus1:    0,
+		TemporalIDNestingFlag: true,
+		ProfileTierLevel: ProfileTierLevel{
+			GeneralProfileSpace:              0,
+			GeneralTierFlag:                  false,
+			GeneralProfileIDC:                2,
+			GeneralProfileCompatibilityFlags: 536870912,
+			GeneralConstraintIndicatorFlags:  193514046488576,
+			GeneralProgressiveSourceFlag:     false,
+			GeneralInterlacedSourceFlag:      false,
+			GeneralNonPackedConstraintFlag:   false,
+			GeneralFrameOnlyConstraintFlag:   false,
+			GeneralLevelIDC:                  156,
+		},
+		SpsID:                   0,
+		ChromaFormatIDC:         1,
+		SeparateColourPlaneFlag: false,
+		ConformanceWindowFlag:   false,
+		PicWidthInLumaSamples:   3840,
+		PicHeightInLumaSamples:  2160,
+		ConformanceWindow: ConformanceWindow{
+			LeftOffset:   0,
+			RightOffset:  0,
+			TopOffset:    0,
+			BottomOffset: 0,
+		},
+		BitDepthLumaMinus8:              2,
+		BitDepthChromaMinus8:            2,
+		Log2MaxPicOrderCntLsbMinus4:     7,
+		SubLayerOrderingInfoPresentFlag: false,
+		SubLayeringOrderingInfos: []SubLayerOrderingInfo{
+			{
+				MaxDecPicBufferingMinus1: 4,
+				MaxNumReorderPics:        2,
+				MaxLatencyIncreasePlus1:  0,
+			},
+		},
+		Log2MinLumaCodingBlockSizeMinus3:     0,
+		Log2DiffMaxMinLumaCodingBlockSize:    2,
+		Log2MinLumaTransformBlockSizeMinus2:  0,
+		Log2DiffMaxMinLumaTransformBlockSize: 3,
+		MaxTransformHierarchyDepthInter:      1,
+		MaxTransformHierarchyDepthIntra:      0,
+		ScalingListEnabledFlag:               true,
+		ScalingListDataPresentFlag:           false,
+		AmpEnabledFlag:                       false,
+		SampleAdaptiveOffsetEnabledFlag:      true,
+		PCMEnabledFlag:                       false,
+		NumShortTermRefPicSets:               0,
+		LongTermRefPicsPresentFlag:           false,
+		SpsTemporalMvpEnabledFlag:            true,
+		StrongIntraSmoothingEnabledFlag:      false,
+		VUIParametersPresentFlag:             true,
+		VUI:                                  &wantedVUI,
+	}
+	got, err := ParseSPSNALUnit(byteData)
+	if err != nil {
+		t.Error("Error parsing SPS")
+	}
+
+	if diff := deep.Equal(*got, wanted); diff != nil {
+		t.Error(diff)
+	}
+	gotWidth, gotHeight := got.ImageSize()
+	var expWidth, expHeight uint32 = 3840, 2160
+	if gotWidth != expWidth || gotHeight != expHeight {
+		t.Errorf("Got %dx%d instead of %dx%d", gotWidth, gotHeight, expWidth, expHeight)
+	}
+}