h264: fix DTS extractor with B-frames and remove PTS-DTS offset

examples/client-read-h264-save-to-disk/mpegtsencoder.go

@@ -11,10 +11,6 @@ import (
 	"github.com/asticode/go-astits"
 )
 
-const (
-	ptsDTSOffset = 400 * time.Millisecond
-)
-
 // mpegtsEncoder allows to encode H264 NALUs into MPEG-TS.
 type mpegtsEncoder struct {
 	sps []byte
@@ -25,6 +21,7 @@ type mpegtsEncoder struct {
 	mux              *astits.Muxer
 	dtsExtractor     *h264.DTSExtractor
 	firstIDRReceived bool
+	startDTS         time.Duration
 }
 
 // newMPEGTSEncoder allocates a mpegtsEncoder.
@@ -92,6 +89,8 @@ func (e *mpegtsEncoder) encode(nalus [][]byte, pts time.Duration) error {
 		filteredNALUs = append(filteredNALUs, nalu)
 	}
 
+	var dts time.Duration
+
 	if !e.firstIDRReceived {
 		// skip samples silently until we find one with a IDR
 		if !idrPresent {
@@ -100,14 +99,27 @@ func (e *mpegtsEncoder) encode(nalus [][]byte, pts time.Duration) error {
 
 		e.firstIDRReceived = true
 		e.dtsExtractor = h264.NewDTSExtractor()
-	}
 
-	dts, err := e.dtsExtractor.Extract(filteredNALUs, pts)
+		var err error
-	if err != nil {
+		dts, err = e.dtsExtractor.Extract(filteredNALUs, pts)
-		return err
+		if err != nil {
-	}
+			return err
+		}
 
-	pts += ptsDTSOffset
+		e.startDTS = dts
+		dts = 0
+		pts -= e.startDTS
+
+	} else {
+		var err error
+		dts, err = e.dtsExtractor.Extract(filteredNALUs, pts)
+		if err != nil {
+			return err
+		}
+
+		dts -= e.startDTS
+		pts -= e.startDTS
+	}
 
 	oh := &astits.PESOptionalHeader{
 		MarkerBits: 2,

pkg/h264/dtsextractor.go

@@ -96,12 +96,13 @@ func getPOCDiff(poc1 uint32, poc2 uint32, sps *SPS) int32 {
 
 // DTSExtractor is a utility that allows to extract NALU DTS from PTS.
 type DTSExtractor struct {
-	sps         []byte
+	sps          []byte
-	spsp        *SPS
+	spsp         *SPS
-	prevPTS     time.Duration
+	prevPTS      time.Duration
-	prevDTS     time.Duration
+	prevDTS      time.Duration
-	prevPOCDiff int32
+	prevPOCDiff  int32
-	expectedPOC uint32
+	expectedPOC  uint32
+	ptsDTSOffset time.Duration
 }
 
 // NewDTSExtractor allocates a DTSExtractor.
@@ -128,6 +129,14 @@ func (d *DTSExtractor) extractInner(
 				}
 				d.sps = append([]byte(nil), nalu...)
 				d.spsp = &spsp
+
+				// in case of B-frames, we have to subtract from DTS the maximum number of reordered frames
+				if d.spsp.VUI != nil && d.spsp.VUI.TimingInfo != nil {
+					d.ptsDTSOffset = time.Duration(math.Round(float64(time.Duration(d.spsp.VUI.MaxNumReorderFrames)*time.Second*
+						time.Duration(d.spsp.VUI.TimingInfo.NumUnitsInTick)*2) / float64(d.spsp.VUI.TimingInfo.TimeScale)))
+				} else {
+					d.ptsDTSOffset = 0
+				}
 			}
 
 		// set IDR present flag
@@ -142,7 +151,7 @@ func (d *DTSExtractor) extractInner(
 
 	if idrPresent || d.spsp.PicOrderCntType == 2 {
 		d.expectedPOC = 0
-		return pts, 0, nil
+		return pts - d.ptsDTSOffset, 0, nil
 	}
 
 	// compute expectedPOC immediately in order to store it even in case of errors
@@ -157,6 +166,11 @@ func (d *DTSExtractor) extractInner(
 	pocDiff := getPOCDiff(poc, d.expectedPOC, d.spsp)
 
 	if pocDiff == 0 {
+		return pts - d.ptsDTSOffset, 0, nil
+	}
+
+	// special case to eliminate errors near 0
+	if d.spsp.VUI != nil && d.spsp.VUI.TimingInfo != nil && pocDiff == -int32(d.spsp.VUI.MaxNumReorderFrames)*2 {
 		return pts, pocDiff, nil
 	}
 
@@ -165,12 +179,13 @@ func (d *DTSExtractor) extractInner(
 			return 0, 0, fmt.Errorf("invalid frame POC")
 		}
 
-		return d.prevPTS + time.Duration(math.Round(float64(pts-d.prevPTS)/float64(pocDiff/2+1))), pocDiff, nil
+		return d.prevPTS - d.ptsDTSOffset +
+			time.Duration(math.Round(float64(pts-d.prevPTS)/float64(pocDiff/2+1))), pocDiff, nil
 	}
 
-	// pocDiff : prevPOCDiff = (pts - dts) : (prevPTS - prevDTS)
+	// pocDiff : prevPOCDiff = (pts - dts - ptsDTSOffset) : (prevPTS - prevDTS - ptsDTSOffset)
-	return pts + time.Duration(math.Round(float64(d.prevDTS-d.prevPTS)*float64(pocDiff)/float64(d.prevPOCDiff))),
+	return pts - d.ptsDTSOffset + time.Duration(math.Round(float64(d.prevDTS-d.prevPTS+d.ptsDTSOffset)*
-		pocDiff, nil
+		float64(pocDiff)/float64(d.prevPOCDiff))), pocDiff, nil
 }
 
 // Extract extracts the DTS of a NALU group.

pkg/h264/dtsextractor_test.go

@@ -80,7 +80,7 @@ func TestDTSExtractor(t *testing.T) {
 				},
 			},
 			0,
-			0,
+			-400 * time.Millisecond,
 		},
 		{
 			[][]byte{
@@ -97,7 +97,7 @@ func TestDTSExtractor(t *testing.T) {
 				},
 			},
 			800 * time.Millisecond,
-			200 * time.Millisecond,
+			-200 * time.Millisecond,
 		},
 		{
 			[][]byte{
@@ -114,7 +114,7 @@ func TestDTSExtractor(t *testing.T) {
 				},
 			},
 			400 * time.Millisecond,
-			400 * time.Millisecond,
+			0,
 		},
 		{
 			[][]byte{
@@ -130,7 +130,7 @@ func TestDTSExtractor(t *testing.T) {
 				},
 			},
 			200 * time.Millisecond,
-			600 * time.Millisecond,
+			200 * time.Millisecond,
 		},
 	}