gilles/go2rtc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Rewrite AnnexB/AVCC parsers

Browse Source
This commit is contained in:
Alexey Khit
2023-08-16 16:50:55 +03:00
parent e1be2d9e48
commit 1dd3dbbcd8
28 changed files with 513 additions and 176 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkg/h264/README.md
+1
View File
@@ -13,3 +13,4 @@ Payloader code taken from [pion](https://github.com/pion/rtp) library. And chang
- [AVC profiles table](https://developer.mozilla.org/ru/docs/Web/Media/Formats/codecs_parameter)
- [Supported Media for Google Cast](https://developers.google.com/cast/docs/media)
- [Two stream formats, Annex-B, AVCC (H.264) and HVCC (H.265)](https://www.programmersought.com/article/3901815022/)
- https://docs.aws.amazon.com/kinesisvideostreams/latest/dg/producer-reference-nal.html
pkg/h264/annexb/annexb.go
+129
View File
@@ -0,0 +1,129 @@
// Package annexb - universal for H264 and H265
package annexb
import (
"bytes"
"encoding/binary"
)
const StartCode = "\x00\x00\x00\x01"
const startAUD = StartCode + "\x09\xF0" + StartCode
// EncodeToAVCC
// will change original slice data!
// safeAppend should be used if original slice has useful data after end (part of other slice)
//
// FFmpeg MPEG-TS: 00000001 AUD 00000001 SPS 00000001 PPS 000001 IFrame
// FFmpeg H264: 00000001 SPS 00000001 PPS 000001 IFrame 00000001 PFrame
func EncodeToAVCC(b []byte, safeAppend bool) []byte {
const minSize = len(StartCode) + 1
// 1. Check frist "start code"
if len(b) < len(startAUD) || string(b[:len(StartCode)]) != StartCode {
return nil
}
// 2. Skip Access unit delimiter (AUD) from FFmpeg
if string(b[:len(startAUD)]) == startAUD {
b = b[6:]
}
var start int
for i, n := minSize, len(b)-minSize; i < n; {
// 3. Check "start code" (first 2 bytes)
if b[i] != 0 || b[i+1] != 0 {
i++
continue
}
// 4. Check "start code" (3 bytes size or 4 bytes size)
if b[i+2] == 1 {
if safeAppend {
// protect original slice from "damage"
b = bytes.Clone(b)
safeAppend = false
}
// convert start code from 3 bytes to 4 bytes
b = append(b, 0)
copy(b[i+1:], b[i:])
n++
} else if b[i+2] != 0 || b[i+3] != 1 {
i++
continue
}
// 5. Set size for previous AU
size := uint32(i - start - len(StartCode))
binary.BigEndian.PutUint32(b[start:], size)
start = i
i += minSize
}
// 6. Set size for last AU
size := uint32(len(b) - start - len(StartCode))
binary.BigEndian.PutUint32(b[start:], size)
return b
}
func DecodeAVCC(b []byte) []byte {
b = bytes.Clone(b)
for i := 0; i < len(b); {
size := int(binary.BigEndian.Uint32(b[i:]))
b[i] = 0
b[i+1] = 0
b[i+2] = 0
b[i+3] = 1
i += 4 + size
}
return b
}
const (
h264PFrame = 1
h264IFrame = 5
h264SPS = 7
h264PPS = 8
h265VPS = 32
h265PFrame = 1
)
// IndexFrame - get new frame start position in the AnnexB stream
func IndexFrame(b []byte) int {
if len(b) < len(startAUD) {
return -1
}
for i := len(startAUD); ; {
if di := bytes.Index(b[i:], []byte(StartCode)); di < 0 {
break
} else {
i += di + 4 // move to NALU start
}
if i >= len(b) {
break
}
h264Type := b[i] & 0b1_1111
switch h264Type {
case h264PFrame, h264SPS:
return i - 4 // move to start code
case h264IFrame, h264PPS:
continue
}
h265Type := (b[i] >> 1) & 0b11_1111
switch h265Type {
case h265PFrame, h265VPS:
return i - 4 // move to start code
}
}
return -1
}
pkg/h264/avc.go
+1 -102
View File
@@ -3,46 +3,12 @@ package h264
import (
"bytes"
"encoding/binary"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/pion/rtp"
)
func AnnexB2AVC(b []byte) []byte {
for i := 0; i < len(b); {
if i+4 >= len(b) {
break
}
size := bytes.Index(b[i+4:], []byte{0, 0, 0, 1})
if size < 0 {
size = len(b) - (i + 4)
}
binary.BigEndian.PutUint32(b[i:], uint32(size))
i += size + 4
}
return b
}
func AVCtoAnnexB(b []byte) []byte {
b = bytes.Clone(b)
for i := 0; i < len(b); {
size := int(binary.BigEndian.Uint32(b[i:]))
b[i] = 0
b[i+1] = 0
b[i+2] = 0
b[i+3] = 1
i += 4 + size
}
return b
}
const forbiddenZeroBit = 0x80
const nalUnitType = 0x1F
// DecodeStream - find and return first AU in AVC format
// Deprecated: DecodeStream - find and return first AU in AVC format
// useful for processing live streams with unknown separator size
func DecodeStream(annexb []byte) ([]byte, int) {
startPos := -1
@@ -154,70 +120,3 @@ func IndexFrom(b []byte, sep []byte, from int) int {
return bytes.Index(b, sep)
}
func EncodeAVC(nals ...[]byte) (avc []byte) {
var i, n int
for _, nal := range nals {
if i = len(nal); i > 0 {
n += 4 + i
}
}
avc = make([]byte, n)
n = 0
for _, nal := range nals {
if i = len(nal); i > 0 {
binary.BigEndian.PutUint32(avc[n:], uint32(i))
n += 4 + copy(avc[n+4:], nal)
}
}
return
}
func RepairAVC(codec *core.Codec, handler core.HandlerFunc) core.HandlerFunc {
sps, pps := GetParameterSet(codec.FmtpLine)
ps := EncodeAVC(sps, pps)
return func(packet *rtp.Packet) {
if NALUType(packet.Payload) == NALUTypeIFrame {
packet.Payload = Join(ps, packet.Payload)
}
handler(packet)
}
}
func SplitAVC(data []byte) [][]byte {
var nals [][]byte
for {
// get AVC length
size := int(binary.BigEndian.Uint32(data)) + 4
// check if multiple items in one packet
if size < len(data) {
nals = append(nals, data[:size])
data = data[size:]
} else {
nals = append(nals, data)
break
}
}
return nals
}
func Types(data []byte) []byte {
var types []byte
for {
types = append(types, NALUType(data))
size := 4 + int(binary.BigEndian.Uint32(data))
if size < len(data) {
data = data[size:]
} else {
break
}
}
return types
}
pkg/h264/avc/avc_test.go
-22
View File
@@ -1,22 +0,0 @@
package avc
import (
"encoding/hex"
"testing"
"github.com/stretchr/testify/require"
)
func TestDecodeConfig(t *testing.T) {
s := "01640033ffe1000c67640033ac1514a02800f19001000468ee3cb0"
src, err := hex.DecodeString(s)
require.Nil(t, err)
profile, sps, pps := DecodeConfig(src)
require.NotNil(t, profile)
require.NotNil(t, sps)
require.NotNil(t, pps)
dst := EncodeConfig(sps, pps)
require.Equal(t, src, dst)
}
pkg/h264/avcc.go
+111
View File
@@ -0,0 +1,111 @@
// Package h264 - AVCC format related functions
package h264
import (
"bytes"
"encoding/base64"
"encoding/binary"
"encoding/hex"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/pion/rtp"
)
func RepairAVCC(codec *core.Codec, handler core.HandlerFunc) core.HandlerFunc {
sps, pps := GetParameterSet(codec.FmtpLine)
ps := JoinNALU(sps, pps)
return func(packet *rtp.Packet) {
if NALUType(packet.Payload) == NALUTypeIFrame {
packet.Payload = Join(ps, packet.Payload)
}
handler(packet)
}
}
func JoinNALU(nalus ...[]byte) (avcc []byte) {
var i, n int
for _, nalu := range nalus {
if i = len(nalu); i > 0 {
n += 4 + i
}
}
avcc = make([]byte, n)
n = 0
for _, nal := range nalus {
if i = len(nal); i > 0 {
binary.BigEndian.PutUint32(avcc[n:], uint32(i))
n += 4 + copy(avcc[n+4:], nal)
}
}
return
}
func SplitNALU(avcc []byte) [][]byte {
var nals [][]byte
for {
// get AVC length
size := int(binary.BigEndian.Uint32(avcc)) + 4
// check if multiple items in one packet
if size < len(avcc) {
nals = append(nals, avcc[:size])
avcc = avcc[size:]
} else {
nals = append(nals, avcc)
break
}
}
return nals
}
func NALUTypes(avcc []byte) []byte {
var types []byte
for {
types = append(types, NALUType(avcc))
size := 4 + int(binary.BigEndian.Uint32(avcc))
if size < len(avcc) {
avcc = avcc[size:]
} else {
break
}
}
return types
}
func AVCCToCodec(avcc []byte) *core.Codec {
buf := bytes.NewBufferString("packetization-mode=1")
for {
size := 4 + int(binary.BigEndian.Uint32(avcc))
switch NALUType(avcc) {
case NALUTypeSPS:
buf.WriteString(";profile-level-id=")
buf.WriteString(hex.EncodeToString(avcc[5:8]))
buf.WriteString(";sprop-parameter-sets=")
buf.WriteString(base64.StdEncoding.EncodeToString(avcc[4:size]))
case NALUTypePPS:
buf.WriteString(",")
buf.WriteString(base64.StdEncoding.EncodeToString(avcc[4:size]))
}
if size < len(avcc) {
avcc = avcc[size:]
} else {
break
}
}
return &core.Codec{
Name: core.CodecH264,
ClockRate: 90000,
FmtpLine: buf.String(),
PayloadType: core.PayloadTypeRAW,
}
}
pkg/h264/helper.go → pkg/h264/h264.go
View File
pkg/h264/avc/sps_test.go → pkg/h264/h264_test.go
+18 -3
View File
@@ -1,12 +1,27 @@
package avc
package h264
import (
"encoding/base64"
"encoding/hex"
"testing"
"github.com/stretchr/testify/require"
)
func TestDecodeConfig(t *testing.T) {
s := "01640033ffe1000c67640033ac1514a02800f19001000468ee3cb0"
src, err := hex.DecodeString(s)
require.Nil(t, err)
profile, sps, pps := DecodeConfig(src)
require.NotNil(t, profile)
require.NotNil(t, sps)
require.NotNil(t, pps)
dst := EncodeConfig(sps, pps)
require.Equal(t, src, dst)
}
func TestDecodeSPS(t *testing.T) {
s := "Z0IAMukAUAHjQgAAB9IAAOqcCAA=" // Amcrest AD410
b, err := base64.StdEncoding.DecodeString(s)
@@ -14,7 +29,7 @@ func TestDecodeSPS(t *testing.T) {
sps := DecodeSPS(b)
require.Equal(t, uint16(2560), sps.Width())
require.Equal(t, uint16(1920), sps.Heigth())
require.Equal(t, uint16(1920), sps.Height())
s = "R00AKZmgHgCJ+WEAAAMD6AAATiCE" // Sonoff
b, err = base64.StdEncoding.DecodeString(s)
@@ -22,5 +37,5 @@ func TestDecodeSPS(t *testing.T) {
sps = DecodeSPS(b)
require.Equal(t, uint16(1920), sps.Width())
require.Equal(t, uint16(1080), sps.Heigth())
require.Equal(t, uint16(1080), sps.Height())
}
pkg/h264/avc/avc.go → pkg/h264/mpeg4.go
+3 -1
View File
@@ -1,4 +1,5 @@
package avc
// Package h264 - MPEG4 format related functions
package h264
import (
"bytes"
@@ -9,6 +10,7 @@ import (
"github.com/AlexxIT/go2rtc/pkg/core"
)
// DecodeConfig - extract profile, SPS and PPS from MPEG4 config
func DecodeConfig(conf []byte) (profile []byte, sps []byte, pps []byte) {
if len(conf) < 6 || conf[0] != 1 {
return
pkg/h264/rtp.go
+4 -2
View File
@@ -2,7 +2,9 @@ package h264
import (
"encoding/binary"
"github.com/AlexxIT/go2rtc/pkg/core"
"github.com/AlexxIT/go2rtc/pkg/h264/annexb"
"github.com/pion/rtp"
"github.com/pion/rtp/codecs"
)
@@ -15,7 +17,7 @@ func RTPDepay(codec *core.Codec, handler core.HandlerFunc) core.HandlerFunc {
depack := &codecs.H264Packet{IsAVC: true}
sps, pps := GetParameterSet(codec.FmtpLine)
ps := EncodeAVC(sps, pps)
ps := JoinNALU(sps, pps)
buf := make([]byte, 0, 512*1024) // 512K
@@ -81,7 +83,7 @@ func RTPDepay(codec *core.Codec, handler core.HandlerFunc) core.HandlerFunc {
// some Chinese buggy cameras has single packet with SPS+PPS+IFrame separated by 00 00 00 01
// https://github.com/AlexxIT/WebRTC/issues/391
// https://github.com/AlexxIT/WebRTC/issues/392
AnnexB2AVC(payload)
payload = annexb.EncodeToAVCC(payload, false)
}
//log.Printf("[AVC] %v, len: %d, ts: %10d, seq: %d", Types(payload), len(payload), packet.Timestamp, packet.SequenceNumber)
pkg/h264/avc/sps.go → pkg/h264/sps.go
+17 -17
View File
@@ -1,4 +1,4 @@
package avc
package h264
import "github.com/AlexxIT/go2rtc/pkg/bits"
@@ -49,11 +49,26 @@ type SPS struct {
sar_height uint32
}
func (s *SPS) Width() uint16 {
width := 16 * (s.pic_width_in_mbs_minus_1 + 1)
crop := 2 * (s.frame_crop_left_offset + s.frame_crop_right_offset)
return uint16(width - crop)
}
func (s *SPS) Height() uint16 {
height := 16 * (s.pic_height_in_map_units_minus_1 + 1)
crop := 2 * (s.frame_crop_top_offset + s.frame_crop_bottom_offset)
if s.frame_mbs_only_flag == 0 {
height *= 2
}
return uint16(height - crop)
}
func DecodeSPS(sps []byte) *SPS {
r := bits.NewReader(sps)
hdr := r.ReadByte()
if hdr&0x1F != 7 {
if hdr&0x1F != NALUTypeSPS {
return nil
}
@@ -147,18 +162,3 @@ func DecodeSPS(sps []byte) *SPS {
return s
}
func (s *SPS) Width() uint16 {
width := 16 * (s.pic_width_in_mbs_minus_1 + 1)
crop := 2 * (s.frame_crop_left_offset + s.frame_crop_right_offset)
return uint16(width - crop)
}
func (s *SPS) Heigth() uint16 {
height := 16 * (s.pic_height_in_map_units_minus_1 + 1)
crop := 2 * (s.frame_crop_top_offset + s.frame_crop_bottom_offset)
if s.frame_mbs_only_flag == 0 {
height *= 2
}
return uint16(height - crop)
}