add MPEG-1/2 video decoder and encoder (#415)

This commit is contained in:
Alessandro Ros
2023-09-13 13:11:49 +02:00
committed by GitHub
parent 73ba46591e
commit 3cdd680f06
16 changed files with 694 additions and 2 deletions

View File

@@ -109,7 +109,7 @@ In RTSP, media streams are routed between server and clients by using RTP packet
|H265||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#H265)|:heavy_check_mark:|
|H264||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#H264)|:heavy_check_mark:|
|MPEG-4 Video (H263, Xvid)||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#MPEG4VideoES)|:heavy_check_mark:|
|MPEG-1/2 Video||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#MPEG1Video)||
|MPEG-1/2 Video||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#MPEG1Video)|:heavy_check_mark:|
|M-JPEG||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#MJPEG)|:heavy_check_mark:|
### Audio
@@ -127,7 +127,7 @@ In RTSP, media streams are routed between server and clients by using RTP packet
|G711 (PCMA, PCMU)||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#G711)|:heavy_check_mark:|
|LPCM||[link](https://pkg.go.dev/github.com/bluenviron/gortsplib/v4/pkg/format#LPCM)|:heavy_check_mark:|
### Mixed
### Other
|format / codec|variant|documentation|encoder and decoder available|
|--------------|-------|-------------|-----------------------------|

View File

@@ -2,6 +2,8 @@ package format //nolint:dupl
import (
"github.com/pion/rtp"
"github.com/bluenviron/gortsplib/v4/pkg/format/rtpmpeg1video"
)
// MPEG1Video is a RTP format for a MPEG-1/2 Video codec.
@@ -41,3 +43,27 @@ func (f *MPEG1Video) FMTP() map[string]string {
func (f *MPEG1Video) PTSEqualsDTS(*rtp.Packet) bool {
return true
}
// CreateDecoder creates a decoder able to decode the content of the format.
func (f *MPEG1Video) CreateDecoder() (*rtpmpeg1video.Decoder, error) {
d := &rtpmpeg1video.Decoder{}
err := d.Init()
if err != nil {
return nil, err
}
return d, nil
}
// CreateEncoder creates an encoder able to encode the content of the format.
func (f *MPEG1Video) CreateEncoder() (*rtpmpeg1video.Encoder, error) {
e := &rtpmpeg1video.Encoder{}
err := e.Init()
if err != nil {
return nil, err
}
return e, nil
}

View File

@@ -13,3 +13,21 @@ func TestMPEG1VideoAttributes(t *testing.T) {
require.Equal(t, 90000, format.ClockRate())
require.Equal(t, true, format.PTSEqualsDTS(&rtp.Packet{}))
}
func TestMPEG1VideoDecEncoder(t *testing.T) {
format := &MPEG1Video{}
enc, err := format.CreateEncoder()
require.NoError(t, err)
pkts, err := enc.Encode([]byte{1, 2, 3, 4})
require.NoError(t, err)
require.Equal(t, format.PayloadType(), pkts[0].PayloadType)
dec, err := format.CreateDecoder()
require.NoError(t, err)
byts, err := dec.Decode(pkts[0])
require.NoError(t, err)
require.Equal(t, []byte{1, 2, 3, 4}, byts)
}

View File

@@ -0,0 +1,150 @@
package rtpmpeg1video
import (
"errors"
"fmt"
"github.com/pion/rtp"
)
const (
maxFrameSize = 1 * 1024 * 1024
)
// ErrMorePacketsNeeded is returned when more packets are needed.
var ErrMorePacketsNeeded = errors.New("need more packets")
// ErrNonStartingPacketAndNoPrevious is returned when we received a non-starting
// packet of a fragmented frame and we didn't received anything before.
// It's normal to receive this when decoding a stream that has been already
// running for some time.
var ErrNonStartingPacketAndNoPrevious = errors.New(
"received a non-starting fragment without any previous starting fragment")
func joinFragments(fragments [][]byte, size int) []byte {
ret := make([]byte, size)
n := 0
for _, p := range fragments {
n += copy(ret[n:], p)
}
return ret
}
// Decoder is a RTP/MPEG-1/2 Video decoder.
// Specification: https://datatracker.ietf.org/doc/html/rfc2250
type Decoder struct {
fragments [][]byte
fragmentsSize int
sliceBuffer [][]byte
sliceBufferSize int
}
// Init initializes the decoder.
func (d *Decoder) Init() error {
return nil
}
func (d *Decoder) decodeSlice(pkt *rtp.Packet) ([]byte, error) {
if len(pkt.Payload) < 4 {
d.fragments = d.fragments[:0] // discard pending fragments
d.fragmentsSize = 0
return nil, fmt.Errorf("payload is too short")
}
mbz := pkt.Payload[0] >> 3
if mbz != 0 {
d.fragments = d.fragments[:0] // discard pending fragments
d.fragmentsSize = 0
return nil, fmt.Errorf("invalid MBZ: %v", mbz)
}
t := (pkt.Payload[0] >> 2) & 0x01
if t != 0 {
d.fragments = d.fragments[:0] // discard pending fragments
d.fragmentsSize = 0
return nil, fmt.Errorf("MPEG-2 video-specific header extension is not supported yet")
}
an := pkt.Payload[2] >> 7
if an != 0 {
d.fragments = d.fragments[:0] // discard pending fragments
d.fragmentsSize = 0
return nil, fmt.Errorf("AN not supported yet")
}
n := (pkt.Payload[2] >> 6) & 0x01
if n != 0 {
d.fragments = d.fragments[:0] // discard pending fragments
d.fragmentsSize = 0
return nil, fmt.Errorf("N not supported yet")
}
b := (pkt.Payload[2] >> 4) & 0x01
e := (pkt.Payload[2] >> 3) & 0x01
switch {
case b == 1 && e == 1:
return pkt.Payload[4:], nil
case b == 1:
d.fragments = d.fragments[:0] // discard pending fragments
d.fragments = append(d.fragments, pkt.Payload[4:])
d.fragmentsSize = len(pkt.Payload[4:])
return nil, ErrMorePacketsNeeded
case e == 1:
if d.fragmentsSize == 0 {
return nil, ErrNonStartingPacketAndNoPrevious
}
d.fragments = append(d.fragments, pkt.Payload[4:])
d.fragmentsSize += len(pkt.Payload[4:])
slice := joinFragments(d.fragments, d.fragmentsSize)
d.fragments = d.fragments[:0]
d.fragmentsSize = 0
return slice, nil
default:
if d.fragmentsSize == 0 {
return nil, ErrNonStartingPacketAndNoPrevious
}
d.fragments = append(d.fragments, pkt.Payload[4:])
d.fragmentsSize += len(pkt.Payload[4:])
return nil, ErrMorePacketsNeeded
}
}
// Decode decodes frames from a RTP packet.
func (d *Decoder) Decode(pkt *rtp.Packet) ([]byte, error) {
slice, err := d.decodeSlice(pkt)
if err != nil {
return nil, err
}
addSize := len(slice)
if (d.sliceBufferSize + addSize) > maxFrameSize {
d.sliceBuffer = nil
d.sliceBufferSize = 0
return nil, fmt.Errorf("frame size (%d) is too big, maximum is %d",
d.sliceBufferSize+addSize, maxFrameSize)
}
d.sliceBuffer = append(d.sliceBuffer, slice)
d.sliceBufferSize += addSize
if !pkt.Marker {
return nil, ErrMorePacketsNeeded
}
ret := joinFragments(d.sliceBuffer, d.sliceBufferSize)
// do not reuse sliceBuffer to avoid race conditions
d.sliceBuffer = nil
d.sliceBufferSize = 0
return ret, nil
}

View File

@@ -0,0 +1,54 @@
package rtpmpeg1video
import (
"testing"
"github.com/pion/rtp"
"github.com/stretchr/testify/require"
)
func TestDecode(t *testing.T) {
for _, ca := range cases {
t.Run(ca.name, func(t *testing.T) {
d := &Decoder{}
err := d.Init()
require.NoError(t, err)
var frame []byte
for _, pkt := range ca.pkts {
frame, err = d.Decode(pkt)
}
require.NoError(t, err)
require.Equal(t, ca.frame, frame)
})
}
}
func FuzzDecoder(f *testing.F) {
f.Fuzz(func(t *testing.T, a []byte, b []byte) {
d := &Decoder{}
d.Init() //nolint:errcheck
d.Decode(&rtp.Packet{ //nolint:errcheck
Header: rtp.Header{
Version: 2,
SequenceNumber: 17645,
Timestamp: 2289527317,
SSRC: 0x9dbb7812,
},
Payload: a,
})
d.Decode(&rtp.Packet{ //nolint:errcheck
Header: rtp.Header{
Version: 2,
SequenceNumber: 17646,
Timestamp: 2289527317,
SSRC: 0x9dbb7812,
},
Payload: b,
})
})
}

View File

@@ -0,0 +1,239 @@
package rtpmpeg1video
import (
"bytes"
"crypto/rand"
"github.com/pion/rtp"
)
const (
rtpVersion = 2
defaultPayloadMaxSize = 1460 // 1500 (UDP MTU) - 20 (IP header) - 8 (UDP header) - 12 (RTP header)
)
func randUint32() (uint32, error) {
var b [4]byte
_, err := rand.Read(b[:])
if err != nil {
return 0, err
}
return uint32(b[0])<<24 | uint32(b[1])<<16 | uint32(b[2])<<8 | uint32(b[3]), nil
}
func lenAggregated(slices [][]byte, slice []byte) int {
l := 4 + len(slice)
for _, fr := range slices {
l += len(fr)
}
return l
}
func packetCount(avail, le int) int {
n := le / avail
if (le % avail) != 0 {
n++
}
return n
}
// Encoder is a RTP/MPEG-1/2 Video encoder.
// Specification: https://datatracker.ietf.org/doc/html/rfc2250
type Encoder struct {
// SSRC of packets (optional).
// It defaults to a random value.
SSRC *uint32
// initial sequence number of packets (optional).
// It defaults to a random value.
InitialSequenceNumber *uint16
// maximum size of packet payloads (optional).
// It defaults to 1460.
PayloadMaxSize int
sequenceNumber uint16
}
// Init initializes the encoder.
func (e *Encoder) Init() error {
if e.SSRC == nil {
v, err := randUint32()
if err != nil {
return err
}
e.SSRC = &v
}
if e.InitialSequenceNumber == nil {
v, err := randUint32()
if err != nil {
return err
}
v2 := uint16(v)
e.InitialSequenceNumber = &v2
}
if e.PayloadMaxSize == 0 {
e.PayloadMaxSize = defaultPayloadMaxSize
}
e.sequenceNumber = *e.InitialSequenceNumber
return nil
}
// Encode encodes frames into RTP packets.
func (e *Encoder) Encode(frame []byte) ([]*rtp.Packet, error) {
var rets []*rtp.Packet
var batch [][]byte
var temporalReference uint16
beginOfSequence := uint8(0)
var frameType uint8
for {
var slice []byte
end := bytes.Index(frame[4:], []byte{0, 0, 1})
if end >= 0 {
slice, frame = frame[:end+4], frame[end+4:]
} else {
slice, frame = frame, nil
}
if lenAggregated(batch, slice) <= e.PayloadMaxSize {
batch = append(batch, slice)
} else {
// write current batch
if batch != nil {
pkts, err := e.writeBatch(batch,
temporalReference,
beginOfSequence,
frameType)
if err != nil {
return nil, err
}
rets = append(rets, pkts...)
beginOfSequence = 0
}
// initialize new batch
batch = [][]byte{slice}
}
switch slice[3] {
case 0:
temporalReference = uint16(slice[4])<<2 | uint16(slice[5])>>6
frameType = (slice[5] >> 3) & 0b111
case 0xB8:
beginOfSequence = 1
}
if frame == nil {
break
}
}
// write last batch
pkts, err := e.writeBatch(batch,
temporalReference,
beginOfSequence,
frameType)
if err != nil {
return nil, err
}
rets = append(rets, pkts...)
rets[len(rets)-1].Marker = true
return rets, nil
}
func (e *Encoder) writeBatch(
slices [][]byte,
temporalReference uint16,
beginOfSequence uint8,
frameType uint8,
) ([]*rtp.Packet, error) {
if len(slices) != 1 || lenAggregated(slices, nil) < e.PayloadMaxSize {
return e.writeAggregated(slices, temporalReference, beginOfSequence, frameType)
}
return e.writeFragmented(slices[0], temporalReference, beginOfSequence, frameType)
}
func (e *Encoder) writeFragmented(
slice []byte,
temporalReference uint16,
beginOfSequence uint8,
frameType uint8,
) ([]*rtp.Packet, error) {
avail := e.PayloadMaxSize - 4
le := len(slice)
packetCount := packetCount(avail, le)
ret := make([]*rtp.Packet, packetCount)
le = avail
start := uint8(1)
end := uint8(0)
for i := range ret {
if i == (packetCount - 1) {
le = len(slice)
end = 1
}
payload := make([]byte, 4+le)
payload[0] = byte(temporalReference >> 8)
payload[1] = byte(temporalReference)
payload[2] = beginOfSequence<<5 | start<<4 | end<<3 | frameType
copy(payload[4:], slice)
slice = slice[le:]
ret[i] = &rtp.Packet{
Header: rtp.Header{
Version: rtpVersion,
PayloadType: 32,
SequenceNumber: e.sequenceNumber,
SSRC: *e.SSRC,
},
Payload: payload,
}
e.sequenceNumber++
start = 0
beginOfSequence = 0
}
return ret, nil
}
func (e *Encoder) writeAggregated(
slices [][]byte,
temporalReference uint16,
beginOfSequence uint8,
frameType uint8,
) ([]*rtp.Packet, error) {
payload := make([]byte, lenAggregated(slices, nil))
payload[0] = byte(temporalReference >> 8)
payload[1] = byte(temporalReference)
payload[2] = beginOfSequence<<5 | 1<<4 | 1<<3 | frameType
n := 4
for _, slice := range slices {
n += copy(payload[n:], slice)
}
pkt := &rtp.Packet{
Header: rtp.Header{
Version: rtpVersion,
PayloadType: 32,
SequenceNumber: e.sequenceNumber,
SSRC: *e.SSRC,
},
Payload: payload,
}
e.sequenceNumber++
return []*rtp.Packet{pkt}, nil
}

View File

@@ -0,0 +1,179 @@
package rtpmpeg1video
import (
"bytes"
"testing"
"github.com/pion/rtp"
"github.com/stretchr/testify/require"
)
func uint16Ptr(v uint16) *uint16 {
return &v
}
func uint32Ptr(v uint32) *uint32 {
return &v
}
func mergeBytes(vals ...[]byte) []byte {
size := 0
for _, v := range vals {
size += len(v)
}
res := make([]byte, size)
pos := 0
for _, v := range vals {
n := copy(res[pos:], v)
pos += n
}
return res
}
var cases = []struct {
name string
frame []byte
pkts []*rtp.Packet
}{
{
"single",
bytes.Repeat([]byte{1, 2, 3, 4}, 240/4),
[]*rtp.Packet{{
Header: rtp.Header{
Version: 2,
Marker: true,
PayloadType: 32,
SequenceNumber: 17645,
SSRC: 0x9dbb7812,
},
Payload: mergeBytes(
[]byte{0, 0, 0x18, 0},
bytes.Repeat([]byte{1, 2, 3, 4}, 240/4),
),
}},
},
{
"aggregated",
mergeBytes(
[]byte{0, 0, 1},
bytes.Repeat([]byte{1, 2, 3, 4}, 128/4),
[]byte{0, 0, 1},
bytes.Repeat([]byte{5, 6, 7, 8}, 128/4),
),
[]*rtp.Packet{{
Header: rtp.Header{
Version: 2,
Marker: true,
PayloadType: 32,
SequenceNumber: 17645,
SSRC: 0x9dbb7812,
},
Payload: mergeBytes(
[]byte{0, 0, 0x18, 0},
[]byte{0, 0, 1},
bytes.Repeat([]byte{1, 2, 3, 4}, 128/4),
[]byte{0, 0, 1},
bytes.Repeat([]byte{5, 6, 7, 8}, 128/4),
),
}},
},
{
"fragmented",
mergeBytes(
[]byte{0, 0, 1},
bytes.Repeat([]byte{1}, 2000),
),
[]*rtp.Packet{
{
Header: rtp.Header{
Version: 2,
Marker: false,
PayloadType: 32,
SequenceNumber: 17645,
SSRC: 0x9dbb7812,
},
Payload: mergeBytes(
[]byte{0, 0, 0x10, 0},
[]byte{0, 0, 1},
bytes.Repeat([]byte{1}, 1453),
),
},
{
Header: rtp.Header{
Version: 2,
Marker: true,
PayloadType: 32,
SequenceNumber: 17646,
SSRC: 0x9dbb7812,
},
Payload: mergeBytes(
[]byte{0, 0, 0x08, 0},
bytes.Repeat([]byte{1}, 547),
),
},
},
},
{
"fragmented to the limit",
mergeBytes(
[]byte{0, 0, 1},
bytes.Repeat([]byte{1}, 2909),
),
[]*rtp.Packet{
{
Header: rtp.Header{
Version: 2,
Marker: false,
PayloadType: 32,
SequenceNumber: 17645,
SSRC: 2646308882,
},
Payload: mergeBytes(
[]byte{0, 0, 0x10, 0},
[]byte{0, 0, 1},
bytes.Repeat([]byte{1}, 1453),
),
},
{
Header: rtp.Header{
Version: 2,
Marker: true,
PayloadType: 32,
SequenceNumber: 17646,
SSRC: 2646308882,
},
Payload: mergeBytes(
[]byte{0, 0, 0x08, 0},
bytes.Repeat([]byte{1}, 1456),
),
},
},
},
}
func TestEncode(t *testing.T) {
for _, ca := range cases {
t.Run(ca.name, func(t *testing.T) {
e := &Encoder{
SSRC: uint32Ptr(0x9dbb7812),
InitialSequenceNumber: uint16Ptr(0x44ed),
}
err := e.Init()
require.NoError(t, err)
pkts, err := e.Encode(ca.frame)
require.NoError(t, err)
require.Equal(t, ca.pkts, pkts)
})
}
}
func TestEncodeRandomInitialState(t *testing.T) {
e := &Encoder{}
err := e.Init()
require.NoError(t, err)
require.NotEqual(t, nil, e.SSRC)
require.NotEqual(t, nil, e.InitialSequenceNumber)
}

View File

@@ -0,0 +1,2 @@
// Package rtpmpeg1video contains a RTP/MPEG-1/2 Video decoder and encoder.
package rtpmpeg1video

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("\x000A0")
[]byte("0")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("0")
[]byte("\x000(0")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("0")
[]byte("\x000 0")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("0")
[]byte("\x06000")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("\x000\x800")
[]byte("0")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("0")
[]byte("")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("0")
[]byte("0000")

View File

@@ -0,0 +1,3 @@
go test fuzz v1
[]byte("\x000000")
[]byte("\x000 0")