mirror
/
av
mirror of https://bitbucket.org/ausocean/av.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

codec/h264/h264dec: merged in master and fixed resultant problems

This commit is contained in:
Saxon 2019-08-05 14:41:26 +09:30
parent 95a5ce0ac0 2473b9ee0a
commit 150492a5bd
25 changed files with 1708 additions and 984 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
cmd/revid-cli/main.go
View File

@ -8,6 +8,7 @@ DESCRIPTION
AUTHORS
Saxon A. Nelson-Milton <saxon@ausocean.org>
Jack Richardson <jack@ausocean.org>
Trek Hopton <trek@ausocean.org>
LICENSE
revid-cli is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)
@ -26,6 +27,7 @@ LICENSE
along with revid in gpl.txt. If not, see http://www.gnu.org/licenses.
*/
// revid-cli is a command line interface for revid.
package main
import (
@ -36,6 +38,7 @@ import (
"strings"
"time"
"bitbucket.org/ausocean/av/codec/codecutil"
"bitbucket.org/ausocean/av/container/mts"
"bitbucket.org/ausocean/av/container/mts/meta"
"bitbucket.org/ausocean/av/revid"
@ -105,9 +108,9 @@ func handleFlags() revid.Config {
var (
cpuprofile = flag.String("cpuprofile", "", "write cpu profile to `file`")
inputPtr = flag.String("Input", "", "The input type: Raspivid, File, Webcam, RTSP")
inputCodecPtr = flag.String("InputCodec", "", "The codec of the input: H264, Mjpeg, PCM, ADPCM")
inputPtr = flag.String("Input", "", "The input type: Raspivid, File, v4l, Audio, RTSP")
rtspURLPtr = flag.String("RTSPURL", "", "The URL for an RTSP server.")
inputCodecPtr = flag.String("InputCodec", "", "The codec of the input: H264, Mjpeg")
quantizePtr = flag.Bool("Quantize", false, "Quantize input (non-variable bitrate)")
verbosityPtr = flag.String("Verbosity", "Info", "Verbosity: Debug, Info, Warning, Error, Fatal")
rtpAddrPtr = flag.String("RtpAddr", "", "Rtp destination address: <IP>:<port> (port is generally 6970-6999)")
@ -131,6 +134,12 @@ func handleFlags() revid.Config {
saturationPtr = flag.Int("Saturation", 0, "Set Saturation. (100-100)")
exposurePtr = flag.String("Exposure", "auto", "Set exposure mode. ("+strings.Join(revid.ExposureModes[:], ",")+")")
autoWhiteBalancePtr = flag.String("Awb", "auto", "Set automatic white balance mode. ("+strings.Join(revid.AutoWhiteBalanceModes[:], ",")+")")
// Audio specific flags.
sampleRatePtr = flag.Int("SampleRate", 48000, "Sample rate of recorded audio")
channelsPtr = flag.Int("Channels", 1, "Record in Mono or Stereo (1 or 2)")
recPeriodPtr = flag.Float64("recPeriod", 1, "How many seconds to record at a time")
bitDepthPtr = flag.Int("bitDepth", 16, "Bit Depth to record audio at.")
)
var outputs flagStrings
@ -179,6 +188,8 @@ func handleFlags() revid.Config {
cfg.Input = revid.V4L
case "File":
cfg.Input = revid.File
case "Audio":
cfg.Input = revid.Audio
case "RTSP":
cfg.Input = revid.RTSP
case "":
@ -188,12 +199,23 @@ func handleFlags() revid.Config {
switch *inputCodecPtr {
case "H264":
cfg.InputCodec = revid.H264
cfg.InputCodec = codecutil.H264
case "PCM":
cfg.InputCodec = codecutil.PCM
case "ADPCM":
cfg.InputCodec = codecutil.ADPCM
case "":
default:
log.Log(logger.Error, pkg+"bad input codec argument")
}
switch *inputPtr {
case "Audio":
cfg.WriteRate = 1.0 / (*recPeriodPtr)
default:
cfg.WriteRate = float64(*frameRatePtr)
}
for _, o := range outputs {
switch o {
case "File":
@ -235,6 +257,10 @@ func handleFlags() revid.Config {
cfg.Saturation = *saturationPtr
cfg.Exposure = *exposurePtr
cfg.AutoWhiteBalance = *autoWhiteBalancePtr
cfg.SampleRate = *sampleRatePtr
cfg.Channels = *channelsPtr
cfg.RecPeriod = *recPeriodPtr
cfg.BitDepth = *bitDepthPtr
return cfg
}

84
codec/codecutil/lex.go Normal file
View File

@ -0,0 +1,84 @@
/*
NAME
lex.go
AUTHOR
Trek Hopton <trek@ausocean.org>
LICENSE
This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
It is free software: you can redistribute it and/or modify them
under the terms of the GNU General Public License as published by the
Free Software Foundation, either version 3 of the License, or (at your
option) any later version.
It is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License in gpl.txt.
If not, see [GNU licenses](http://www.gnu.org/licenses).
*/
package codecutil
import (
"fmt"
"io"
"time"
)
// ByteLexer is used to lex bytes using a buffer size which is configured upon construction.
type ByteLexer struct {
bufSize *int
}
// NewByteLexer returns a pointer to a ByteLexer with the given buffer size.
func NewByteLexer(bufSize *int) *ByteLexer {
return &ByteLexer{bufSize: bufSize}
}
// zeroTicks can be used to create an instant ticker.
var zeroTicks chan time.Time
func init() {
zeroTicks = make(chan time.Time)
close(zeroTicks)
}
// Lex reads *l.bufSize bytes from src and writes them to dst every d seconds.
func (l *ByteLexer) Lex(dst io.Writer, src io.Reader, d time.Duration) error {
if l.bufSize == nil {
return fmt.Errorf("buffer size has not been set")
}
bufSize := *l.bufSize
if bufSize <= 0 {
return fmt.Errorf("invalid buffer size: %v", bufSize)
}
if d < 0 {
return fmt.Errorf("invalid delay: %v", d)
}
var ticker *time.Ticker
if d == 0 {
ticker = &time.Ticker{C: zeroTicks}
} else {
ticker = time.NewTicker(d)
defer ticker.Stop()
}
buf := make([]byte, bufSize)
for {
<-ticker.C
off, err := src.Read(buf)
if err != nil {
return err
}
_, err = dst.Write(buf[:off])
if err != nil {
return err
}
}
}

65
codec/codecutil/lex_test.go Normal file
View File

@ -0,0 +1,65 @@
/*
NAME
lex_test.go
AUTHOR
Trek Hopton <trek@ausocean.org>
LICENSE
This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
It is free software: you can redistribute it and/or modify them
under the terms of the GNU General Public License as published by the
Free Software Foundation, either version 3 of the License, or (at your
option) any later version.
It is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License in gpl.txt.
If not, see [GNU licenses](http://www.gnu.org/licenses).
*/
package codecutil
import (
"bytes"
"io"
"strconv"
"testing"
"time"
)
var lexTests = []struct {
data []byte
t time.Duration
n int
isValid bool // Whether or not this test should fail.
}{
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, 4, true},
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, 3, true},
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, 0, 2, true},
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, 0, 1, true},
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Nanosecond, 0, false},
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, -1, false},
{[]byte{0x10, 0x00, 0xf3, 0x45, 0xfe, 0xd2, 0xaa, 0x4e}, time.Millisecond, 15, true},
}
func TestByteLexer(t *testing.T) {
for i, tt := range lexTests {
t.Run(strconv.Itoa(i), func(t *testing.T) {
dst := bytes.NewBuffer([]byte{})
l := NewByteLexer(&tt.n)
err := l.Lex(dst, bytes.NewReader(tt.data), tt.t)
if err != nil && err != io.EOF {
if tt.isValid {
t.Errorf("unexpected error: %v", err)
}
} else if !bytes.Equal(dst.Bytes(), tt.data) {
t.Errorf("data before and after lex are not equal: want %v, got %v", tt.data, dst.Bytes())
}
})
}
}

43
codec/codecutil/list.go Normal file
View File

@ -0,0 +1,43 @@
/*
NAME
list.go
AUTHOR
Trek Hopton <trek@ausocean.org>
LICENSE
This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
It is free software: you can redistribute it and/or modify them
under the terms of the GNU General Public License as published by the
Free Software Foundation, either version 3 of the License, or (at your
option) any later version.
It is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License in gpl.txt.
If not, see [GNU licenses](http://www.gnu.org/licenses).
*/
package codecutil
// numCodecs is the number of entries in the list of codecs.
const numCodecs = 5
// A global list containing all available codecs for reference in any application.
// When adding or removing a codec from this list, the numCodecs const must be updated.
const (
PCM = iota
ADPCM
H264
H265
MJPEG
)
// IsValid recieves an int representing a codec and checks if it is valid.
func IsValid(codec uint8) bool {
return 0 <= codec && codec < numCodecs
}

4
codec/h264/h264dec/cabac.go
View File

@ -35,14 +35,14 @@ func YOffset(yRefMin16, refMbH int) int {
}
func MbWidthC(sps *SPS) int {
mbWidthC := 16 / SubWidthC(sps)
if sps.ChromaFormat == chromaMonochrome || sps.UseSeparateColorPlane {
if sps.ChromaFormatIDC == chromaMonochrome || sps.SeparateColorPlaneFlag {
mbWidthC = 0
}
return mbWidthC
}
func MbHeightC(sps *SPS) int {
mbHeightC := 16 / SubHeightC(sps)
if sps.ChromaFormat == chromaMonochrome || sps.UseSeparateColorPlane {
if sps.ChromaFormatIDC == chromaMonochrome || sps.SeparateColorPlaneFlag {
mbHeightC = 0
}
return mbHeightC

21
codec/h264/h264dec/parse.go
View File

@ -64,11 +64,11 @@ func (r fieldReader) readBits(n int) uint64 {
// Exp-Golomb-coded element using method as specified in section 9.1 of ITU-T
// H.264 and return as an int. The read does not happen if the fieldReader
// has a non-nil error.
func (r fieldReader) readUe() int {
func (r fieldReader) readUe() uint64 {
if r.e != nil {
return 0
}
var i int
var i uint64
i, r.e = readUe(r.br)
return i
}
@ -77,11 +77,11 @@ func (r fieldReader) readUe() int {
// Exp-Golomb-coded syntax element using method as specified in section 9.1
// and returns as an int. The read does not happen if the fieldReader
// has a non-nil error.
func (r fieldReader) readTe(x uint) int {
func (r fieldReader) readTe(x uint) int64 {
if r.e != nil {
return 0
}
var i int
var i int64
i, r.e = readTe(r.br, x)
return i
}
@ -122,7 +122,7 @@ func (r fieldReader) err() error {
//
// TODO: this should return uint, but rest of code needs to be changed for this
// to happen.
func readUe(r *bits.BitReader) (int, error) {
func readUe(r *bits.BitReader) (uint64, error) {
nZeros := -1
var err error
for b := uint64(0); b == 0; nZeros++ {
@ -135,7 +135,7 @@ func readUe(r *bits.BitReader) (int, error) {
if err != nil {
return 0, err
}
return int(math.Pow(float64(2), float64(nZeros)) - 1 + float64(rem)), nil
return uint64(math.Pow(float64(2), float64(nZeros)) - 1 + float64(rem)), nil
}
// readTe parses a syntax element of te(v) descriptor i.e, truncated
@ -143,9 +143,10 @@ func readUe(r *bits.BitReader) (int, error) {
// Rec. ITU-T H.264 (04/2017).
//
// TODO: this should also return uint.
func readTe(r *bits.BitReader, x uint) (int, error) {
func readTe(r *bits.BitReader, x uint) (int64, error) {
if x > 1 {
return readUe(r)
ue, err := readUe(r)
return int64(ue), err
}
if x == 1 {
@ -181,7 +182,7 @@ func readSe(r *bits.BitReader) (int, error) {
// in Rec. ITU-T H.264 (04/2017).
func readMe(r *bits.BitReader, chromaArrayType uint, mpm mbPartPredMode) (uint, error) {
// Indexes to codedBlockPattern map.
var i1, i2, i3 int
var i1, i2, i3 uint64
// ChromaArrayType selects first index.
switch chromaArrayType {
@ -200,7 +201,7 @@ func readMe(r *bits.BitReader, chromaArrayType uint, mpm mbPartPredMode) (uint,
}
// Need to check that we won't go out of bounds with this index.
if i2 >= len(codedBlockPattern[i1]) {
if int(i2) >= len(codedBlockPattern[i1]) {
return 0, errInvalidCodeNum
}

165
codec/h264/h264dec/pps.go
View File

@ -4,7 +4,6 @@ import (
"math"
"bitbucket.org/ausocean/av/codec/h264/h264dec/bits"
"github.com/pkg/errors"
)
// import "strings"
@ -42,152 +41,54 @@ type PPS struct {
func NewPPS(br *bits.BitReader, chromaFormat int) (*PPS, error) {
pps := PPS{}
var err error
r := newFieldReader(br)
pps.ID, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse ID")
}
pps.SPSID, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse SPS ID")
}
b, err := br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read EntropyCodingMode")
}
pps.EntropyCodingMode = int(b)
b, err = br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read BottomFieldPicOrderInFramePresent")
}
pps.BottomFieldPicOrderInFramePresent = b == 1
pps.NumSliceGroupsMinus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse NumSliceGroupsMinus1")
}
pps.ID = int(r.readUe())
pps.SPSID = int(r.readUe())
pps.EntropyCodingMode = int(r.readBits(1))
pps.BottomFieldPicOrderInFramePresent = r.readBits(1) == 1
pps.NumSliceGroupsMinus1 = int(r.readUe())
if pps.NumSliceGroupsMinus1 > 0 {
pps.SliceGroupMapType, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse SliceGroupMapType")
}
pps.SliceGroupMapType = int(r.readUe())
if pps.SliceGroupMapType == 0 {
for iGroup := 0; iGroup <= pps.NumSliceGroupsMinus1; iGroup++ {
b, err := readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse RunLengthMinus1")
}
pps.RunLengthMinus1 = append(pps.RunLengthMinus1, b)
pps.RunLengthMinus1 = append(pps.RunLengthMinus1, int(r.readUe()))
}
} else if pps.SliceGroupMapType == 2 {
for iGroup := 0; iGroup < pps.NumSliceGroupsMinus1; iGroup++ {
pps.TopLeft[iGroup], err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse TopLeft[iGroup]")
}
if err != nil {
return nil, errors.Wrap(err, "could not parse TopLeft[iGroup]")
}
pps.BottomRight[iGroup], err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse BottomRight[iGroup]")
}
pps.TopLeft[iGroup] = int(r.readUe())
pps.BottomRight[iGroup] = int(r.readUe())
}
} else if pps.SliceGroupMapType > 2 && pps.SliceGroupMapType < 6 {
b, err = br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read SliceGroupChangeDirection")
}
pps.SliceGroupChangeDirection = b == 1
pps.SliceGroupChangeRateMinus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse SliceGroupChangeRateMinus1")
}
pps.SliceGroupChangeDirection = r.readBits(1) == 1
pps.SliceGroupChangeRateMinus1 = int(r.readUe())
} else if pps.SliceGroupMapType == 6 {
pps.PicSizeInMapUnitsMinus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse PicSizeInMapUnitsMinus1")
}
pps.PicSizeInMapUnitsMinus1 = int(r.readUe())
for i := 0; i <= pps.PicSizeInMapUnitsMinus1; i++ {
b, err = br.ReadBits(int(math.Ceil(math.Log2(float64(pps.NumSliceGroupsMinus1 + 1)))))
if err != nil {
return nil, errors.Wrap(err, "coult not read SliceGroupId")
}
pps.SliceGroupId[i] = int(b)
pps.SliceGroupId[i] = int(r.readBits(int(math.Ceil(math.Log2(float64(pps.NumSliceGroupsMinus1 + 1))))))
}
}
}
pps.NumRefIdxL0DefaultActiveMinus1, err = readUe(br)
if err != nil {
return nil, errors.New("could not parse NumRefIdxL0DefaultActiveMinus1")
}
pps.NumRefIdxL1DefaultActiveMinus1, err = readUe(br)
if err != nil {
return nil, errors.New("could not parse NumRefIdxL1DefaultActiveMinus1")
}
b, err = br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read WeightedPred")
}
pps.WeightedPred = b == 1
b, err = br.ReadBits(2)
if err != nil {
return nil, errors.Wrap(err, "could not read WeightedBipred")
}
pps.WeightedBipred = int(b)
pps.PicInitQpMinus26, err = readSe(br)
if err != nil {
return nil, errors.New("could not parse PicInitQpMinus26")
}
pps.PicInitQsMinus26, err = readSe(br)
if err != nil {
return nil, errors.New("could not parse PicInitQsMinus26")
}
pps.ChromaQpIndexOffset, err = readSe(br)
if err != nil {
return nil, errors.New("could not parse ChromaQpIndexOffset")
}
err = readFlags(br, []flag{
{&pps.DeblockingFilterControlPresent, "DeblockingFilterControlPresent"},
{&pps.ConstrainedIntraPred, "ConstrainedIntraPred"},
{&pps.RedundantPicCntPresent, "RedundantPicCntPresent"},
})
if err != nil {
return nil, err
}
pps.NumRefIdxL0DefaultActiveMinus1 = int(r.readUe())
pps.NumRefIdxL1DefaultActiveMinus1 = int(r.readUe())
pps.WeightedPred = r.readBits(1) == 1
pps.WeightedBipred = int(r.readBits(2))
pps.PicInitQpMinus26 = int(r.readSe())
pps.PicInitQsMinus26 = int(r.readSe())
pps.ChromaQpIndexOffset = int(r.readSe())
pps.DeblockingFilterControlPresent = r.readBits(1) == 1
pps.ConstrainedIntraPred = r.readBits(1) == 1
pps.RedundantPicCntPresent = r.readBits(1) == 1
logger.Printf("debug: \tChecking for more PPS data")
if moreRBSPData(br) {
logger.Printf("debug: \tProcessing additional PPS data")
b, err = br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read Transform8x8Mode")
}
pps.Transform8x8Mode = int(b)
b, err = br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read PicScalingMatrixPresent")
}
pps.PicScalingMatrixPresent = b == 1
pps.Transform8x8Mode = int(r.readBits(1))
pps.PicScalingMatrixPresent = r.readBits(1) == 1
if pps.PicScalingMatrixPresent {
v := 6
@ -195,11 +96,7 @@ func NewPPS(br *bits.BitReader, chromaFormat int) (*PPS, error) {
v = 2
}
for i := 0; i < 6+(v*pps.Transform8x8Mode); i++ {
b, err = br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read PicScalingListPresent")
}
pps.PicScalingListPresent[i] = b == 1
pps.PicScalingListPresent[i] = r.readBits(1) == 1
if pps.PicScalingListPresent[i] {
if i < 6 {
scalingList(
@ -219,11 +116,9 @@ func NewPPS(br *bits.BitReader, chromaFormat int) (*PPS, error) {
}
}
}
pps.SecondChromaQpIndexOffset, err = readSe(br)
if err != nil {
return nil, errors.New("could not parse SecondChromaQpIndexOffset")
}
pps.SecondChromaQpIndexOffset = r.readSe()
moreRBSPData(br)
// rbspTrailingBits()
}
moreRBSPData(br)
return &pps, nil
}

3
codec/h264/h264dec/read.go
View File

@ -71,7 +71,8 @@ func (h *H264Reader) Start() {
case naluTypePPS:
videoStream := h.VideoStreams[len(h.VideoStreams)-1]
// TODO: handle this error
videoStream.PPS, _ = NewPPS(nil, videoStream.SPS.ChromaFormat)
// TODO: fix chromaFormat
videoStream.PPS, _ = NewPPS(nil, 0)
case naluTypeSliceIDRPicture:
fallthrough
case naluTypeSliceNonIDRPicture:

234
codec/h264/h264dec/slice.go
View File

@ -81,12 +81,12 @@ func NewRefPicListModification(br *bits.BitReader, p *PPS, s *SliceHeader) (*Ref
if r.RefPicListModificationFlag[0] {
for i := 0; ; i++ {
r.ModificationOfPicNums[0][i] = fr.readUe()
r.ModificationOfPicNums[0][i] = int(fr.readUe())
if r.ModificationOfPicNums[0][i] == 0 || r.ModificationOfPicNums[0][i] == 1 {
r.AbsDiffPicNumMinus1[0][i] = fr.readUe()
r.AbsDiffPicNumMinus1[0][i] = int(fr.readUe())
} else if r.ModificationOfPicNums[0][i] == 2 {
r.LongTermPicNum[0][i] = fr.readUe()
r.LongTermPicNum[0][i] = int(fr.readUe())
}
if r.ModificationOfPicNums[0][i] == 3 {
@ -101,12 +101,12 @@ func NewRefPicListModification(br *bits.BitReader, p *PPS, s *SliceHeader) (*Ref
if r.RefPicListModificationFlag[1] {
for i := 0; ; i++ {
r.ModificationOfPicNums[1][i] = fr.readUe()
r.ModificationOfPicNums[1][i] = int(fr.readUe())
if r.ModificationOfPicNums[1][i] == 0 || r.ModificationOfPicNums[1][i] == 1 {
r.AbsDiffPicNumMinus1[1][i] = fr.readUe()
r.AbsDiffPicNumMinus1[1][i] = int(fr.readUe())
} else if r.ModificationOfPicNums[1][i] == 2 {
r.LongTermPicNum[1][i] = fr.readUe()
r.LongTermPicNum[1][i] = int(fr.readUe())
}
if r.ModificationOfPicNums[1][i] == 3 {
@ -142,25 +142,15 @@ type PredWeightTable struct {
// PredWeightTable.
func NewPredWeightTable(br *bits.BitReader, h *SliceHeader, chromaArrayType int) (*PredWeightTable, error) {
p := &PredWeightTable{}
var err error
r := newFieldReader(br)
p.LumaLog2WeightDenom, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse LumaLog2WeightDenom")
}
p.LumaLog2WeightDenom = int(r.readUe())
if chromaArrayType != 0 {
p.ChromaLog2WeightDenom, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse ChromaLog2WeightDenom")
}
p.ChromaLog2WeightDenom = int(r.readUe())
}
for i := 0; i <= h.NumRefIdxL0ActiveMinus1; i++ {
b, err := br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read LumaWeightL0Flag")
}
p.LumaWeightL0Flag = b == 1
p.LumaWeightL0Flag = r.readBits(1) == 1
if p.LumaWeightL0Flag {
se, err := readSe(br)
@ -255,10 +245,10 @@ func NewPredWeightTable(br *bits.BitReader, h *SliceHeader, chromaArrayType int)
// DecRefPicMarking provides elements of a dec_ref_pic_marking syntax structure
// as defined in section 7.3.3.3 of the specifications.
type DecRefPicMarking struct {
NoOutputOfPriorPicsFlag bool
LongTermReferenceFlag bool
AdaptiveRefPicMarkingModeFlag bool
elements []drpmElement
NoOutputOfPriorPicsFlag bool
LongTermReferenceFlag bool
AdaptiveRefPicMarkingModeFlag bool
elements []drpmElement
}
type drpmElement struct {
@ -274,6 +264,7 @@ type drpmElement struct {
// DecRefPicMarking.
func NewDecRefPicMarking(br *bits.BitReader, idrPic bool) (*DecRefPicMarking, error) {
d := &DecRefPicMarking{}
r := newFieldReader(br)
if idrPic {
b, err := br.ReadBits(1)
if err != nil {
@ -295,36 +286,21 @@ func NewDecRefPicMarking(br *bits.BitReader, idrPic bool) (*DecRefPicMarking, er
if d.AdaptiveRefPicMarkingModeFlag {
for i := 0; ; i++ {
d.elements = append(d.elements,drpmElement{})
d.elements = append(d.elements, drpmElement{})
d.elements[i].MemoryManagementControlOperation, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse MemoryManagementControlOperation")
}
d.elements[i].MemoryManagementControlOperation = int(r.readUe())
if d.elements[i].MemoryManagementControlOperation == 1 || d.elements[i].MemoryManagementControlOperation == 3 {
d.elements[i].DifferenceOfPicNumsMinus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse MemoryManagementControlOperation")
}
d.elements[i].DifferenceOfPicNumsMinus1 = int(r.readUe())
}
if d.elements[i].MemoryManagementControlOperation == 2 {
d.elements[i].LongTermPicNum, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse LongTermPicNum")
}
d.elements[i].LongTermPicNum = int(r.readUe())
}
if d.elements[i].MemoryManagementControlOperation == 3 || d.elements[i].MemoryManagementControlOperation == 6 {
d.elements[i].LongTermFrameIdx, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse LongTermFrameIdx")
}
d.elements[i].LongTermFrameIdx = int(r.readUe())
}
if d.elements[i].MemoryManagementControlOperation == 4 {
d.elements[i].MaxLongTermFrameIdxPlus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse MaxLongTermFrameIdxPlus1")
}
d.elements[i].MaxLongTermFrameIdxPlus1 = int(r.readUe())
}
if d.elements[i].MemoryManagementControlOperation == 0 {
@ -425,13 +401,13 @@ func (d SliceData) ae(v int) int {
// 8.2.2
func MbToSliceGroupMap(sps *SPS, pps *PPS, header *SliceHeader) []int {
mbaffFrameFlag := 0
if sps.MBAdaptiveFrameField && !header.FieldPic {
if sps.MBAdaptiveFrameFieldFlag && !header.FieldPic {
mbaffFrameFlag = 1
}
mapUnitToSliceGroupMap := MapUnitToSliceGroupMap(sps, pps, header)
mbToSliceGroupMap := []int{}
for i := 0; i <= PicSizeInMbs(sps, header)-1; i++ {
if sps.FrameMbsOnly || header.FieldPic {
if sps.FrameMBSOnlyFlag || header.FieldPic {
mbToSliceGroupMap = append(mbToSliceGroupMap, mapUnitToSliceGroupMap[i])
continue
}
@ -439,7 +415,7 @@ func MbToSliceGroupMap(sps *SPS, pps *PPS, header *SliceHeader) []int {
mbToSliceGroupMap = append(mbToSliceGroupMap, mapUnitToSliceGroupMap[i/2])
continue
}
if !sps.FrameMbsOnly && !sps.MBAdaptiveFrameField && !header.FieldPic {
if !sps.FrameMBSOnlyFlag && !sps.MBAdaptiveFrameFieldFlag && !header.FieldPic {
mbToSliceGroupMap = append(
mbToSliceGroupMap,
mapUnitToSliceGroupMap[(i/(2*PicWidthInMbs(sps)))*PicWidthInMbs(sps)+(i%PicWidthInMbs(sps))])
@ -449,34 +425,34 @@ func MbToSliceGroupMap(sps *SPS, pps *PPS, header *SliceHeader) []int {
}
func PicWidthInMbs(sps *SPS) int {
return sps.PicWidthInMbsMinus1 + 1
return int(sps.PicWidthInMBSMinus1 + 1)
}
func PicHeightInMapUnits(sps *SPS) int {
return sps.PicHeightInMapUnitsMinus1 + 1
return int(sps.PicHeightInMapUnitsMinus1 + 1)
}
func PicSizeInMapUnits(sps *SPS) int {
return PicWidthInMbs(sps) * PicHeightInMapUnits(sps)
return int(PicWidthInMbs(sps) * PicHeightInMapUnits(sps))
}
func FrameHeightInMbs(sps *SPS) int {
return (2 - flagVal(sps.FrameMbsOnly)) * PicHeightInMapUnits(sps)
return int((2 - flagVal(sps.FrameMBSOnlyFlag)) * PicHeightInMapUnits(sps))
}
func PicHeightInMbs(sps *SPS, header *SliceHeader) int {
return FrameHeightInMbs(sps) / (1 + flagVal(header.FieldPic))
return int(FrameHeightInMbs(sps) / (1 + flagVal(header.FieldPic)))
}
func PicSizeInMbs(sps *SPS, header *SliceHeader) int {
return PicWidthInMbs(sps) * PicHeightInMbs(sps, header)
return int(PicWidthInMbs(sps) * PicHeightInMbs(sps, header))
}
// table 6-1
func SubWidthC(sps *SPS) int {
n := 17
if sps.UseSeparateColorPlane {
if sps.ChromaFormat == chroma444 {
if sps.SeparateColorPlaneFlag {
if sps.ChromaFormatIDC == chroma444 {
return n
}
}
switch sps.ChromaFormat {
switch sps.ChromaFormatIDC {
case chromaMonochrome:
return n
case chroma420:
@ -491,12 +467,12 @@ func SubWidthC(sps *SPS) int {
}
func SubHeightC(sps *SPS) int {
n := 17
if sps.UseSeparateColorPlane {
if sps.ChromaFormat == chroma444 {
if sps.SeparateColorPlaneFlag {
if sps.ChromaFormatIDC == chroma444 {
return n
}
}
switch sps.ChromaFormat {
switch sps.ChromaFormatIDC {
case chromaMonochrome:
return n
case chroma420:
@ -543,6 +519,8 @@ func NumMbPart(nalUnit *NALUnit, sps *SPS, header *SliceHeader, data *SliceData)
func MbPred(chromaArrayType int, sliceContext *SliceContext, br *bits.BitReader, rbsp []byte) error {
var cabac *CABAC
r := newFieldReader(br)
sliceType := sliceTypeMap[sliceContext.Slice.Header.SliceType]
mbPartPredMode, err := MbPartPredMode(sliceContext.Slice.Data, sliceType, sliceContext.Slice.Data.MbType, 0)
if err != nil {
@ -652,11 +630,7 @@ func MbPred(chromaArrayType int, sliceContext *SliceContext, br *bits.BitReader,
logger.Printf("TODO: ae for IntraChromaPredMode\n")
} else {
var err error
sliceContext.Slice.Data.IntraChromaPredMode, err = readUe(br)
if err != nil {
return errors.Wrap(err, "could not parse IntraChromaPredMode")
}
sliceContext.Slice.Data.IntraChromaPredMode = int(r.readUe())
}
}
@ -685,14 +659,10 @@ func MbPred(chromaArrayType int, sliceContext *SliceContext, br *bits.BitReader,
// TODO: Only one reference picture is used for inter-prediction,
// then the value should be 0
if MbaffFrameFlag(sliceContext.SPS, sliceContext.Slice.Header) == 0 || !sliceContext.Slice.Data.MbFieldDecodingFlag {
sliceContext.Slice.Data.RefIdxL0[mbPartIdx], _ = readTe(
br,
uint(sliceContext.Slice.Header.NumRefIdxL0ActiveMinus1))
sliceContext.Slice.Data.RefIdxL0[mbPartIdx] = int(r.readTe(uint(sliceContext.Slice.Header.NumRefIdxL0ActiveMinus1)))
} else {
rangeMax := 2*sliceContext.Slice.Header.NumRefIdxL0ActiveMinus1 + 1
sliceContext.Slice.Data.RefIdxL0[mbPartIdx], _ = readTe(
br,
uint(rangeMax))
sliceContext.Slice.Data.RefIdxL0[mbPartIdx] = int(r.readTe(uint(rangeMax)))
}
}
}
@ -858,19 +828,19 @@ func nextMbAddress(n int, sps *SPS, pps *PPS, header *SliceHeader) int {
i := n + 1
// picSizeInMbs is the number of macroblocks in picture 0
// 7-13
// PicWidthInMbs = sps.PicWidthInMbsMinus1 + 1
// PicWidthInMbs = sps.PicWidthInMBSMinus1 + 1
// PicHeightInMapUnits = sps.PicHeightInMapUnitsMinus1 + 1
// 7-29
// picSizeInMbs = PicWidthInMbs * PicHeightInMbs
// 7-26
// PicHeightInMbs = FrameHeightInMbs / (1 + header.fieldPicFlag)
// 7-18
// FrameHeightInMbs = (2 - ps.FrameMbsOnly) * PicHeightInMapUnits
picWidthInMbs := sps.PicWidthInMbsMinus1 + 1
// FrameHeightInMbs = (2 - ps.FrameMBSOnlyFlag) * PicHeightInMapUnits
picWidthInMbs := sps.PicWidthInMBSMinus1 + 1
picHeightInMapUnits := sps.PicHeightInMapUnitsMinus1 + 1
frameHeightInMbs := (2 - flagVal(sps.FrameMbsOnly)) * picHeightInMapUnits
frameHeightInMbs := (2 - flagVal(sps.FrameMBSOnlyFlag)) * int(picHeightInMapUnits)
picHeightInMbs := frameHeightInMbs / (1 + flagVal(header.FieldPic))
picSizeInMbs := picWidthInMbs * picHeightInMbs
picSizeInMbs := int(picWidthInMbs) * picHeightInMbs
mbToSliceGroupMap := MbToSliceGroupMap(sps, pps, header)
for i < picSizeInMbs && mbToSliceGroupMap[i] != mbToSliceGroupMap[i] {
i++
@ -880,7 +850,7 @@ func nextMbAddress(n int, sps *SPS, pps *PPS, header *SliceHeader) int {
func CurrMbAddr(sps *SPS, header *SliceHeader) int {
mbaffFrameFlag := 0
if sps.MBAdaptiveFrameField && !header.FieldPic {
if sps.MBAdaptiveFrameFieldFlag && !header.FieldPic {
mbaffFrameFlag = 1
}
@ -888,15 +858,15 @@ func CurrMbAddr(sps *SPS, header *SliceHeader) int {
}
func MbaffFrameFlag(sps *SPS, header *SliceHeader) int {
if sps.MBAdaptiveFrameField && !header.FieldPic {
if sps.MBAdaptiveFrameFieldFlag && !header.FieldPic {
return 1
}
return 0
}
func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitReader) (*SliceData, error) {
r := newFieldReader(br)
var cabac *CABAC
var err error
sliceContext.Slice.Data = &SliceData{BitReader: br}
// TODO: Why is this being initialized here?
// initCabac(sliceContext)
@ -910,7 +880,7 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
}
}
mbaffFrameFlag := 0
if sliceContext.SPS.MBAdaptiveFrameField && !sliceContext.Slice.Header.FieldPic {
if sliceContext.SPS.MBAdaptiveFrameFieldFlag && !sliceContext.Slice.Header.FieldPic {
mbaffFrameFlag = 1
}
currMbAddr := sliceContext.Slice.Header.FirstMbInSlice * (1 * mbaffFrameFlag)
@ -925,10 +895,7 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
if sliceContext.Slice.Data.SliceTypeName != "I" && sliceContext.Slice.Data.SliceTypeName != "SI" {
logger.Printf("debug: \tNonI/SI slice, processing moreData\n")
if sliceContext.PPS.EntropyCodingMode == 0 {
sliceContext.Slice.Data.MbSkipRun, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse MbSkipRun")
}
sliceContext.Slice.Data.MbSkipRun = int(r.readUe())
if sliceContext.Slice.Data.MbSkipRun > 0 {
prevMbSkipped = 1
@ -1042,10 +1009,7 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
logger.Printf("TODO: ae for MBType\n")
} else {
sliceContext.Slice.Data.MbType, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse MbType")
}
sliceContext.Slice.Data.MbType = int(r.readUe())
}
if sliceContext.Slice.Data.MbTypeName == "I_PCM" {
for !br.ByteAligned() {
@ -1057,7 +1021,7 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
// 7-3 p95
bitDepthY := 8 + sliceContext.SPS.BitDepthLumaMinus8
for i := 0; i < 256; i++ {
s, err := br.ReadBits(bitDepthY)
s, err := br.ReadBits(int(bitDepthY))
if err != nil {
return nil, errors.Wrap(err, fmt.Sprintf("could not read PcmSampleLuma[%d]", i))
}
@ -1071,14 +1035,14 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
mbWidthC := 16 / SubWidthC(sliceContext.SPS)
mbHeightC := 16 / SubHeightC(sliceContext.SPS)
// if monochrome
if sliceContext.SPS.ChromaFormat == chromaMonochrome || sliceContext.SPS.UseSeparateColorPlane {
if sliceContext.SPS.ChromaFormatIDC == chromaMonochrome || sliceContext.SPS.SeparateColorPlaneFlag {
mbWidthC = 0
mbHeightC = 0
}
bitDepthC := 8 + sliceContext.SPS.BitDepthChromaMinus8
for i := 0; i < 2*mbWidthC*mbHeightC; i++ {
s, err := br.ReadBits(bitDepthC)
s, err := br.ReadBits(int(bitDepthC))
if err != nil {
return nil, errors.Wrap(err, fmt.Sprintf("could not read PcmSampleChroma[%d]", i))
}
@ -1104,7 +1068,7 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
if NumbSubMbPart(subMbType[mbPartIdx]) > 1 {
noSubMbPartSizeLessThan8x8Flag = 0
}
} else if !sliceContext.SPS.Direct8x8Inference {
} else if !sliceContext.SPS.Direct8x8InferenceFlag {
noSubMbPartSizeLessThan8x8Flag = 0
}
}
@ -1156,7 +1120,7 @@ func NewSliceData(chromaArrayType int, sliceContext *SliceContext, br *bits.BitR
}
// sliceContext.Slice.Data.CodedBlockPattern = me(v) | ae(v)
if CodedBlockPatternLuma(sliceContext.Slice.Data) > 0 && sliceContext.PPS.Transform8x8Mode == 1 && sliceContext.Slice.Data.MbTypeName != "I_NxN" && noSubMbPartSizeLessThan8x8Flag == 1 && (sliceContext.Slice.Data.MbTypeName != "B_Direct_16x16" || sliceContext.SPS.Direct8x8Inference) {
if CodedBlockPatternLuma(sliceContext.Slice.Data) > 0 && sliceContext.PPS.Transform8x8Mode == 1 && sliceContext.Slice.Data.MbTypeName != "I_NxN" && noSubMbPartSizeLessThan8x8Flag == 1 && (sliceContext.Slice.Data.MbTypeName != "B_Direct_16x16" || sliceContext.SPS.Direct8x8InferenceFlag) {
// TODO: 1 bit or ae(v)
if sliceContext.PPS.EntropyCodingMode == 1 {
binarization := NewBinarization("Transform8x8Flag", sliceContext.Slice.Data)
@ -1236,31 +1200,21 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
idrPic = true
}
header := SliceHeader{}
if sps.UseSeparateColorPlane {
if sps.SeparateColorPlaneFlag {
vid.ChromaArrayType = 0
} else {
vid.ChromaArrayType = sps.ChromaFormat
vid.ChromaArrayType = int(sps.ChromaFormatIDC)
}
br := bits.NewBitReader(bytes.NewReader(rbsp))
r := newFieldReader(br)
header.FirstMbInSlice, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse FirstMbInSlice")
}
header.SliceType, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse SliceType")
}
header.FirstMbInSlice = int(r.readUe())
header.SliceType = int(r.readUe())
sliceType := sliceTypeMap[header.SliceType]
logger.Printf("debug: %s (%s) slice of %d bytes\n", NALUnitType[int(nalUnit.Type)], sliceType, len(rbsp))
header.PPSID, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse PPSID")
}
if sps.UseSeparateColorPlane {
header.PPSID = int(r.readUe())
if sps.SeparateColorPlaneFlag {
b, err := br.ReadBits(2)
if err != nil {
return nil, errors.Wrap(err, "could not read ColorPlaneID")
@ -1269,7 +1223,7 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
}
// TODO: See 7.4.3
// header.FrameNum = b.NextField("FrameNum", 0)
if !sps.FrameMbsOnly {
if !sps.FrameMBSOnlyFlag {
b, err := br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read FieldPic")
@ -1284,13 +1238,10 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
}
}
if idrPic {
header.IDRPicID, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse IDRPicID")
}
header.IDRPicID = int(r.readUe())
}
if sps.PicOrderCountType == 0 {
b, err := br.ReadBits(sps.Log2MaxPicOrderCntLSBMin4 + 4)
b, err := br.ReadBits(int(sps.Log2MaxPicOrderCntLSBMin4 + 4))
if err != nil {
return nil, errors.Wrap(err, "could not read PicOrderCntLsb")
}
@ -1303,7 +1254,7 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
}
}
}
if sps.PicOrderCountType == 1 && !sps.DeltaPicOrderAlwaysZero {
if sps.PicOrderCountType == 1 && !sps.DeltaPicOrderAlwaysZeroFlag {
header.DeltaPicOrderCnt[0], err = readSe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse DeltaPicOrderCnt")
@ -1317,10 +1268,7 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
}
}
if pps.RedundantPicCntPresent {
header.RedundantPicCnt, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse RedundantPicCnt")
}
header.RedundantPicCnt = int(r.readUe())
}
if sliceType == "B" {
b, err := br.ReadBits(1)
@ -1337,15 +1285,9 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
header.NumRefIdxActiveOverride = b == 1
if header.NumRefIdxActiveOverride {
header.NumRefIdxL0ActiveMinus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse NumRefIdxL0ActiveMinus1")
}
header.NumRefIdxL0ActiveMinus1 = int(r.readUe())
if sliceType == "B" {
header.NumRefIdxL1ActiveMinus1, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse NumRefIdxL1ActiveMinus1")
}
header.NumRefIdxL1ActiveMinus1 = int(r.readUe())
}
}
}
@ -1375,35 +1317,18 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
}
}
if pps.EntropyCodingMode == 1 && sliceType != "I" && sliceType != "SI" {
header.CabacInit, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse CabacInit")
}
}
header.SliceQpDelta, err = readSe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse SliceQpDelta")
header.CabacInit = int(r.readUe())
}
header.SliceQpDelta = int(r.readSe())
if sliceType == "SP" || sliceType == "SI" {
if sliceType == "SP" {
b, err := br.ReadBits(1)
if err != nil {
return nil, errors.Wrap(err, "could not read SpForSwitch")
}
header.SpForSwitch = b == 1
}
header.SliceQsDelta, err = readSe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse SliceQsDelta")
header.SpForSwitch = r.readBits(1) == 1
}
header.SliceQsDelta = int(r.readSe())
}
if pps.DeblockingFilterControlPresent {
header.DisableDeblockingFilter, err = readUe(br)
if err != nil {
return nil, errors.Wrap(err, "could not parse DisableDeblockingFilter")
}
header.DisableDeblockingFilter = int(r.readUe())
if header.DisableDeblockingFilter != 1 {
header.SliceAlphaC0OffsetDiv2, err = readSe(br)
if err != nil {
@ -1432,13 +1357,10 @@ func NewSliceContext(vid *VideoStream, nalUnit *NALUnit, rbsp []byte, showPacket
Header: &header,
},
}
sliceContext.Slice.Data, err = NewSliceData(vid.ChromaArrayType,sliceContext, br)
sliceContext.Slice.Data, err = NewSliceData(vid.ChromaArrayType, sliceContext, br)
if err != nil {
return nil, errors.Wrap(err, "could not create slice data")
}
if showPacket {
debugPacket("debug: Header", sliceContext.Slice.Header)
debugPacket("debug: Data", sliceContext.Slice.Data)
}
return sliceContext, nil
}

24
codec/h264/h264dec/slice_test.go
View File

@ -22,12 +22,12 @@ var subWidthCTests = []struct {
want int
}{
{SPS{}, 17},
{SPS{ChromaFormat: 0}, 17},
{SPS{ChromaFormat: 1}, 2},
{SPS{ChromaFormat: 2}, 2},
{SPS{ChromaFormat: 3}, 1},
{SPS{ChromaFormat: 3, UseSeparateColorPlane: true}, 17},
{SPS{ChromaFormat: 999}, 17},
{SPS{ChromaFormatIDC: 0}, 17},
{SPS{ChromaFormatIDC: 1}, 2},
{SPS{ChromaFormatIDC: 2}, 2},
{SPS{ChromaFormatIDC: 3}, 1},
{SPS{ChromaFormatIDC: 3, SeparateColorPlaneFlag: true}, 17},
{SPS{ChromaFormatIDC: 999}, 17},
}
// TestSubWidthC tests that the correct SubWidthC is returned given
@ -45,12 +45,12 @@ var subHeightCTests = []struct {
want int
}{
{SPS{}, 17},
{SPS{ChromaFormat: 0}, 17},
{SPS{ChromaFormat: 1}, 2},
{SPS{ChromaFormat: 2}, 1},
{SPS{ChromaFormat: 3}, 1},
{SPS{ChromaFormat: 3, UseSeparateColorPlane: true}, 17},
{SPS{ChromaFormat: 999}, 17},
{SPS{ChromaFormatIDC: 0}, 17},
{SPS{ChromaFormatIDC: 1}, 2},
{SPS{ChromaFormatIDC: 2}, 1},
{SPS{ChromaFormatIDC: 3}, 1},
{SPS{ChromaFormatIDC: 3, SeparateColorPlaneFlag: true}, 17},
{SPS{ChromaFormatIDC: 999}, 17},
}
// TestSubHeightC tests that the correct SubHeightC is returned given

1198
codec/h264/h264dec/sps.go
View File

File diff suppressed because it is too large Load Diff

65
codec/pcm/pcm.go
View File

@ -35,20 +35,20 @@ import (
"github.com/yobert/alsa"
)
// Resample takes an alsa.Buffer (b) and resamples the pcm audio data to 'rate' Hz and returns the resulting pcm.
// If an error occurs, an error will be returned along with the original b's data.
// Resample takes alsa.Buffer b and resamples the pcm audio data to 'rate' Hz and returns an alsa.Buffer with the resampled data.
// Notes:
// - Currently only downsampling is implemented and b's rate must be divisible by 'rate' or an error will occur.
// - If the number of bytes in b.Data is not divisible by the decimation factor (ratioFrom), the remaining bytes will
// not be included in the result. Eg. input of length 480002 downsampling 6:1 will result in output length 80000.
func Resample(b alsa.Buffer, rate int) ([]byte, error) {
fromRate := b.Format.Rate
if fromRate == rate {
return b.Data, nil
} else if fromRate < 0 {
return nil, fmt.Errorf("Unable to convert from: %v Hz", fromRate)
} else if rate < 0 {
return nil, fmt.Errorf("Unable to convert to: %v Hz", rate)
func Resample(b alsa.Buffer, rate int) (alsa.Buffer, error) {
if b.Format.Rate == rate {
return b, nil
}
if b.Format.Rate < 0 {
return alsa.Buffer{}, fmt.Errorf("Unable to convert from: %v Hz", b.Format.Rate)
}
if rate < 0 {
return alsa.Buffer{}, fmt.Errorf("Unable to convert to: %v Hz", rate)
}
// The number of bytes in a sample.
@ -59,22 +59,22 @@ func Resample(b alsa.Buffer, rate int) ([]byte, error) {
case alsa.S16_LE:
sampleLen = 2 * b.Format.Channels
default:
return nil, fmt.Errorf("Unhandled ALSA format: %v", b.Format.SampleFormat)
return alsa.Buffer{}, fmt.Errorf("Unhandled ALSA format: %v", b.Format.SampleFormat)
}
inPcmLen := len(b.Data)
// Calculate sample rate ratio ratioFrom:ratioTo.
rateGcd := gcd(rate, fromRate)
ratioFrom := fromRate / rateGcd
rateGcd := gcd(rate, b.Format.Rate)
ratioFrom := b.Format.Rate / rateGcd
ratioTo := rate / rateGcd
// ratioTo = 1 is the only number that will result in an even sampling.
if ratioTo != 1 {
return nil, fmt.Errorf("unhandled from:to rate ratio %v:%v: 'to' must be 1", ratioFrom, ratioTo)
return alsa.Buffer{}, fmt.Errorf("unhandled from:to rate ratio %v:%v: 'to' must be 1", ratioFrom, ratioTo)
}
newLen := inPcmLen / ratioFrom
result := make([]byte, 0, newLen)
resampled := make([]byte, 0, newLen)
// For each new sample to be generated, loop through the respective 'ratioFrom' samples in 'b.Data' to add them
// up and average them. The result is the new sample.
@ -96,19 +96,28 @@ func Resample(b alsa.Buffer, rate int) ([]byte, error) {
case alsa.S16_LE:
binary.LittleEndian.PutUint16(bAvg, uint16(avg))
}
result = append(result, bAvg...)
resampled = append(resampled, bAvg...)
}
return result, nil
// Return a new alsa.Buffer with resampled data.
return alsa.Buffer{
Format: alsa.BufferFormat{
Channels: b.Format.Channels,
SampleFormat: b.Format.SampleFormat,
Rate: rate,
},
Data: resampled,
}, nil
}
// StereoToMono returns raw mono audio data generated from only the left channel from
// the given stereo recording (ALSA buffer)
// if an error occurs, an error will be returned along with the original stereo data.
func StereoToMono(b alsa.Buffer) ([]byte, error) {
func StereoToMono(b alsa.Buffer) (alsa.Buffer, error) {
if b.Format.Channels == 1 {
return b.Data, nil
} else if b.Format.Channels != 2 {
return nil, fmt.Errorf("Audio is not stereo or mono, it has %v channels", b.Format.Channels)
return b, nil
}
if b.Format.Channels != 2 {
return alsa.Buffer{}, fmt.Errorf("Audio is not stereo or mono, it has %v channels", b.Format.Channels)
}
var stereoSampleBytes int
@ -118,7 +127,7 @@ func StereoToMono(b alsa.Buffer) ([]byte, error) {
case alsa.S16_LE:
stereoSampleBytes = 4
default:
return nil, fmt.Errorf("Unhandled ALSA format %v", b.Format.SampleFormat)
return alsa.Buffer{}, fmt.Errorf("Unhandled ALSA format %v", b.Format.SampleFormat)
}
recLength := len(b.Data)
@ -134,7 +143,15 @@ func StereoToMono(b alsa.Buffer) ([]byte, error) {
}
}
return mono, nil
// Return a new alsa.Buffer with resampled data.
return alsa.Buffer{
Format: alsa.BufferFormat{
Channels: 1,
SampleFormat: b.Format.SampleFormat,
Rate: b.Format.Rate,
},
Data: mono,
}, nil
}
// gcd is used for calculating the greatest common divisor of two positive integers, a and b.

4
codec/pcm/pcm_test.go
View File

@ -71,7 +71,7 @@ func TestResample(t *testing.T) {
}
// Compare result with expected.
if !bytes.Equal(resampled, exp) {
if !bytes.Equal(resampled.Data, exp) {
t.Error("Resampled data does not match expected result.")
}
}
@ -112,7 +112,7 @@ func TestStereoToMono(t *testing.T) {
}
// Compare result with expected.
if !bytes.Equal(mono, exp) {
if !bytes.Equal(mono.Data, exp) {
t.Error("Converted data does not match expected result.")
}
}

17
container/mts/encoder.go
View File

@ -89,11 +89,6 @@ var (
)
const (
sdtPid = 17
patPid = 0
pmtPid = 4096
videoPid = 256
audioPid = 210
H264ID = 27
H265ID = 36
audioStreamID = 0xc0 // First audio stream ID.
@ -151,13 +146,13 @@ func NewEncoder(dst io.WriteCloser, rate float64, mediaType int) *Encoder {
var sid byte
switch mediaType {
case EncodeAudio:
mPid = audioPid
mPid = AudioPid
sid = audioStreamID
case EncodeH265:
mPid = videoPid
mPid = VideoPid
sid = H265ID
case EncodeH264:
mPid = videoPid
mPid = VideoPid
sid = H264ID
}
@ -187,8 +182,8 @@ func NewEncoder(dst io.WriteCloser, rate float64, mediaType int) *Encoder {
streamID: sid,
continuity: map[int]byte{
patPid: 0,
pmtPid: 0,
PatPid: 0,
PmtPid: 0,
mPid: 0,
},
}
@ -214,7 +209,7 @@ func (e *Encoder) TimeBasedPsi(b bool, sendCount int) {
e.pktCount = e.psiSendCount
}
// Write implements io.Writer. Write takes raw h264 and encodes into MPEG-TS,
// Write implements io.Writer. Write takes raw video or audio data and encodes into MPEG-TS,
// then sending it to the encoder's io.Writer destination.
func (e *Encoder) Write(data []byte) (int, error) {
now := time.Now()

2
container/mts/encoder_test.go
View File

@ -199,7 +199,7 @@ func TestEncodePcm(t *testing.T) {
for i+PacketSize <= len(clip) {
// Check MTS packet
if !(pkt.PID() == audioPid) {
if pkt.PID() != AudioPid {
i += PacketSize
if i+PacketSize <= len(clip) {
copy(pkt[:], clip[i:i+PacketSize])

1
container/mts/mpegts.go
View File

@ -47,6 +47,7 @@ const (
PatPid = 0
PmtPid = 4096
VideoPid = 256
AudioPid = 210
)
// StreamID is the id of the first stream.

4
container/mts/mpegts_test.go
View File

@ -82,7 +82,7 @@ func TestGetPTSRange1(t *testing.T) {
curTime += interval
}
got, err := GetPTSRange(clip.Bytes(), videoPid)
got, err := GetPTSRange(clip.Bytes(), VideoPid)
if err != nil {
t.Fatalf("did not expect error getting PTS range: %v", err)
}
@ -142,7 +142,7 @@ func writeFrame(b *bytes.Buffer, frame []byte, pts uint64) error {
for len(buf) != 0 {
pkt := Packet{
PUSI: pusi,
PID: videoPid,
PID: VideoPid,
RAI: pusi,
CC: 0,
AFC: hasAdaptationField | hasPayload,

4
exp/pcm/resample/resample.go
View File

@ -81,9 +81,9 @@ func main() {
}
// Save resampled to file.
err = ioutil.WriteFile(outPath, resampled, 0644)
err = ioutil.WriteFile(outPath, resampled.Data, 0644)
if err != nil {
log.Fatal(err)
}
fmt.Println("Encoded and wrote", len(resampled), "bytes to file", outPath)
fmt.Println("Encoded and wrote", len(resampled.Data), "bytes to file", outPath)
}

4
exp/pcm/stereo-to-mono/stereo-to-mono.go
View File

@ -77,9 +77,9 @@ func main() {
}
// Save mono to file.
err = ioutil.WriteFile(outPath, mono, 0644)
err = ioutil.WriteFile(outPath, mono.Data, 0644)
if err != nil {
log.Fatal(err)
}
fmt.Println("Encoded and wrote", len(mono), "bytes to file", outPath)
fmt.Println("Encoded and wrote", len(mono.Data), "bytes to file", outPath)
}

464
input/audio/audio.go Normal file
View File

@ -0,0 +1,464 @@
/*
NAME
audio.go
AUTHOR
Alan Noble <alan@ausocean.org>
Trek Hopton <trek@ausocean.org>
LICENSE
This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
It is free software: you can redistribute it and/or modify them
under the terms of the GNU General Public License as published by the
Free Software Foundation, either version 3 of the License, or (at your
option) any later version.
It is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License in gpl.txt.
If not, see [GNU licenses](http://www.gnu.org/licenses).
*/
// Package audio provides access to input from audio devices.
package audio
import (
"bytes"
"errors"
"fmt"
"sync"
"time"
"github.com/yobert/alsa"
"bitbucket.org/ausocean/av/codec/adpcm"
"bitbucket.org/ausocean/av/codec/codecutil"
"bitbucket.org/ausocean/av/codec/pcm"
"bitbucket.org/ausocean/utils/logger"
"bitbucket.org/ausocean/utils/ring"
)
const (
pkg = "audio: "
rbTimeout = 100 * time.Millisecond
rbNextTimeout = 100 * time.Millisecond
rbLen = 200
defaultSampleRate = 48000
)
// "running" means the input goroutine is reading from the ALSA device and writing to the ringbuffer.
// "paused" means the input routine is sleeping until unpaused or stopped.
// "stopped" means the input routine is stopped and the ALSA device is closed.
const (
running = iota + 1
paused
stopped
)
// Device holds everything we need to know about the audio input stream and implements io.Reader.
type Device struct {
l Logger // Logger for device's routines to log to.
mode uint8 // Operating mode, either running, paused, or stopped.
mu sync.Mutex // Provides synchronisation when changing modes concurrently.
title string // Name of audio title, or empty for the default title.
dev *alsa.Device // ALSA's Audio input device.
ab alsa.Buffer // ALSA's buffer.
rb *ring.Buffer // Our buffer.
chunkSize int // This is the number of bytes that will be stored in rb at a time.
*Config // Configuration parameters for this device.
}
// Config provides parameters used by Device.
type Config struct {
SampleRate int
Channels int
BitDepth int
RecPeriod float64
Codec uint8
}
// Logger enables any implementation of a logger to be used.
// TODO: Make this part of the logger package.
type Logger interface {
SetLevel(int8)
Log(level int8, message string, params ...interface{})
}
// OpenError is used to determine whether an error has originated from attempting to open a device.
type OpenError error
// NewDevice initializes and returns an Device which can be started, read from, and stopped.
func NewDevice(cfg *Config, l Logger) (*Device, error) {
err := validate(cfg)
if err != nil {
return nil, err
}
d := &Device{
Config: cfg,
l: l,
}
// Open the requested audio device.
err = d.open()
if err != nil {
d.l.Log(logger.Error, pkg+"failed to open device")
return nil, err
}
// Setup the device to record with desired period.
d.ab = d.dev.NewBufferDuration(time.Duration(d.RecPeriod * float64(time.Second)))
// Account for channel conversion.
chunkSize := float64(len(d.ab.Data) / d.dev.BufferFormat().Channels * d.Channels)
// Account for resampling.
chunkSize = (chunkSize / float64(d.dev.BufferFormat().Rate)) * float64(d.SampleRate)
if chunkSize < 1 {
return nil, errors.New("given Config parameters are too small")
}
// Account for codec conversion.
if d.Codec == codecutil.ADPCM {
d.chunkSize = adpcm.EncBytes(int(chunkSize))
} else {
d.chunkSize = int(chunkSize)
}
// Create ring buffer with appropriate chunk size.
d.rb = ring.NewBuffer(rbLen, d.chunkSize, rbTimeout)
// Start device in paused mode.
d.mode = paused
go d.input()
return d, nil
}
// Start will start recording audio and writing to the ringbuffer.
// Once a Device has been stopped it cannot be started again. This is likely to change in future.
func (d *Device) Start() error {
d.mu.Lock()
mode := d.mode
d.mu.Unlock()
switch mode {
case paused:
d.mu.Lock()
d.mode = running
d.mu.Unlock()
return nil
case stopped:
// TODO(Trek): Make this reopen device and start recording.
return errors.New("device is stopped")
case running:
return nil
default:
return fmt.Errorf("invalid mode: %d", mode)
}
}
// Stop will stop recording audio and close the device.
// Once a Device has been stopped it cannot be started again. This is likely to change in future.
func (d *Device) Stop() {
d.mu.Lock()
d.mode = stopped
d.mu.Unlock()
}
// ChunkSize returns the number of bytes written to the ringbuffer per d.RecPeriod.
func (d *Device) ChunkSize() int {
return d.chunkSize
}
// validate checks if Config parameters are valid and returns an error if they are not.
func validate(c *Config) error {
if c.SampleRate <= 0 {
return fmt.Errorf("invalid sample rate: %v", c.SampleRate)
}
if c.Channels <= 0 {
return fmt.Errorf("invalid number of channels: %v", c.Channels)
}
if c.BitDepth <= 0 {
return fmt.Errorf("invalid bitdepth: %v", c.BitDepth)
}
if c.RecPeriod <= 0 {
return fmt.Errorf("invalid recording period: %v", c.RecPeriod)
}
if !codecutil.IsValid(c.Codec) {
return errors.New("invalid codec")
}
return nil
}
// open the recording device with the given name and prepare it to record.
// If name is empty, the first recording device is used.
func (d *Device) open() error {
// Close any existing device.
if d.dev != nil {
d.l.Log(logger.Debug, pkg+"closing device", "title", d.title)
d.dev.Close()
d.dev = nil
}
// Open sound card and open recording device.
d.l.Log(logger.Debug, pkg+"opening sound card")
cards, err := alsa.OpenCards()
if err != nil {
return OpenError(err)
}
defer alsa.CloseCards(cards)
d.l.Log(logger.Debug, pkg+"finding audio device")
for _, card := range cards {
devices, err := card.Devices()
if err != nil {
continue
}
for _, dev := range devices {
if dev.Type != alsa.PCM || !dev.Record {
continue
}
if dev.Title == d.title || d.title == "" {
d.dev = dev
break
}
}
}
if d.dev == nil {
return OpenError(errors.New("no audio device found"))
}
d.l.Log(logger.Debug, pkg+"opening audio device", "title", d.dev.Title)
err = d.dev.Open()
if err != nil {
return OpenError(err)
}
// 2 channels is what most devices need to record in. If mono is requested,
// the recording will be converted in formatBuffer().
channels, err := d.dev.NegotiateChannels(2)
if err != nil {
return OpenError(err)
}
d.l.Log(logger.Debug, pkg+"alsa device channels set", "channels", channels)
// Try to negotiate a rate to record in that is divisible by the wanted rate
// so that it can be easily downsampled to the wanted rate.
// rates is a slice of common sample rates including the standard for CD (44100Hz) and standard for professional audio recording (48000Hz).
// Note: if a card thinks it can record at a rate but can't actually, this can cause a failure.
// Eg. the audioinjector sound card is supposed to record at 8000Hz and 16000Hz but it can't due to a firmware issue,
// a fix for this is to remove 8000 and 16000 from the rates slice.
var rates = [8]int{8000, 16000, 32000, 44100, 48000, 88200, 96000, 192000}
var rate int
foundRate := false
for r := range rates {
if r < d.SampleRate {
continue
}
if r%d.SampleRate == 0 {
rate, err = d.dev.NegotiateRate(r)
if err == nil {
foundRate = true
d.l.Log(logger.Debug, pkg+"alsa device sample rate set", "rate", rate)
break
}
}
}
// If no easily divisible rate is found, then use the default rate.
if !foundRate {
d.l.Log(logger.Warning, pkg+"Unable to sample at requested rate, default used.", "rateRequested", d.SampleRate)
rate, err = d.dev.NegotiateRate(defaultSampleRate)
if err != nil {
return OpenError(err)
}
d.l.Log(logger.Debug, pkg+"alsa device sample rate set", "rate", rate)
}
var aFmt alsa.FormatType
switch d.BitDepth {
case 16:
aFmt = alsa.S16_LE
case 32:
aFmt = alsa.S32_LE
default:
return OpenError(fmt.Errorf("unsupported sample bits %v", d.BitDepth))
}
devFmt, err := d.dev.NegotiateFormat(aFmt)
if err != nil {
return err
}
var bitdepth int
switch devFmt {
case alsa.S16_LE:
bitdepth = 16
case alsa.S32_LE:
bitdepth = 32
default:
return OpenError(fmt.Errorf("unsupported sample bits %v", d.BitDepth))
}
d.l.Log(logger.Debug, pkg+"alsa device bit depth set", "bitdepth", bitdepth)
// A 50ms period is a sensible value for low-ish latency. (this could be made configurable if needed)
// Some devices only accept even period sizes while others want powers of 2.
// So we will find the closest power of 2 to the desired period size.
const wantPeriod = 0.05 //seconds
bytesPerSecond := rate * channels * (bitdepth / 8)
wantPeriodSize := int(float64(bytesPerSecond) * wantPeriod)
nearWantPeriodSize := nearestPowerOfTwo(wantPeriodSize)
// At least two period sizes should fit within the buffer.
bufSize, err := d.dev.NegotiateBufferSize(nearWantPeriodSize * 2)
if err != nil {
return OpenError(err)
}
d.l.Log(logger.Debug, pkg+"alsa device buffer size set", "buffersize", bufSize)
if err = d.dev.Prepare(); err != nil {
return OpenError(err)
}
d.l.Log(logger.Debug, pkg+"successfully negotiated ALSA params")
return nil
}
// input continously records audio and writes it to the ringbuffer.
// Re-opens the device and tries again if ASLA returns an error.
func (d *Device) input() {
for {
// Check mode.
d.mu.Lock()
mode := d.mode
d.mu.Unlock()
switch mode {
case paused:
time.Sleep(time.Duration(d.RecPeriod) * time.Second)
continue
case stopped:
if d.dev != nil {
d.l.Log(logger.Debug, pkg+"closing audio device", "title", d.title)
d.dev.Close()
d.dev = nil
}
return
}
// Read from audio device.
d.l.Log(logger.Debug, pkg+"recording audio for period", "seconds", d.RecPeriod)
err := d.dev.Read(d.ab.Data)
if err != nil {
d.l.Log(logger.Debug, pkg+"read failed", "error", err.Error())
err = d.open() // re-open
if err != nil {
d.l.Log(logger.Fatal, pkg+"reopening device failed", "error", err.Error())
return
}
continue
}
// Process audio.
d.l.Log(logger.Debug, pkg+"processing audio")
toWrite := d.formatBuffer()
// Write audio to ringbuffer.
n, err := d.rb.Write(toWrite.Data)
switch err {
case nil:
d.l.Log(logger.Debug, pkg+"wrote audio to ringbuffer", "length", n)
case ring.ErrDropped:
d.l.Log(logger.Warning, pkg+"old audio data overwritten")
default:
d.l.Log(logger.Error, pkg+"unexpected ringbuffer error", "error", err.Error())
return
}
}
}
// Read reads from the ringbuffer, returning the number of bytes read upon success.
func (d *Device) Read(p []byte) (int, error) {
// Ready ringbuffer for read.
_, err := d.rb.Next(rbNextTimeout)
if err != nil {
return 0, err
}
// Read from ring buffer.
return d.rb.Read(p)
}
// formatBuffer returns audio that has been converted to the desired format.
func (d *Device) formatBuffer() alsa.Buffer {
var err error
// If nothing needs to be changed, return the original.
if d.ab.Format.Channels == d.Channels && d.ab.Format.Rate == d.SampleRate {
return d.ab
}
var formatted alsa.Buffer
if d.ab.Format.Channels != d.Channels {
// Convert channels.
// TODO(Trek): Make this work for conversions other than stereo to mono.
if d.ab.Format.Channels == 2 && d.Channels == 1 {
formatted, err = pcm.StereoToMono(d.ab)
if err != nil {
d.l.Log(logger.Fatal, pkg+"channel conversion failed", "error", err.Error())
}
}
}
if d.ab.Format.Rate != d.SampleRate {
// Convert rate.
formatted, err = pcm.Resample(formatted, d.SampleRate)
if err != nil {
d.l.Log(logger.Fatal, pkg+"rate conversion failed", "error", err.Error())
}
}
switch d.Codec {
case codecutil.PCM:
case codecutil.ADPCM:
b := bytes.NewBuffer(make([]byte, 0, adpcm.EncBytes(len(formatted.Data))))
enc := adpcm.NewEncoder(b)
_, err = enc.Write(formatted.Data)
if err != nil {
d.l.Log(logger.Fatal, pkg+"unable to encode", "error", err.Error())
}
formatted.Data = b.Bytes()
default:
d.l.Log(logger.Error, pkg+"unhandled audio codec")
}
return formatted
}
// nearestPowerOfTwo finds and returns the nearest power of two to the given integer.
// If the lower and higher power of two are the same distance, it returns the higher power.
// For negative values, 1 is returned.
// Source: https://stackoverflow.com/a/45859570
func nearestPowerOfTwo(n int) int {
if n <= 0 {
return 1
}
if n == 1 {
return 2
}
v := n
v--
v |= v >> 1
v |= v >> 2
v |= v >> 4
v |= v >> 8
v |= v >> 16
v++ // higher power of 2
x := v >> 1 // lower power of 2
if (v - n) > (n - x) {
return x
}
return v
}

105
input/audio/audio_test.go Normal file
View File

@ -0,0 +1,105 @@
/*
NAME
audio_test.go
AUTHOR
Trek Hopton <trek@ausocean.org>
LICENSE
This file is Copyright (C) 2019 the Australian Ocean Lab (AusOcean)
It is free software: you can redistribute it and/or modify them
under the terms of the GNU General Public License as published by the
Free Software Foundation, either version 3 of the License, or (at your
option) any later version.
It is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License in gpl.txt.
If not, see [GNU licenses](http://www.gnu.org/licenses).
*/
package audio
import (
"io/ioutil"
"os"
"strconv"
"testing"
"time"
"bitbucket.org/ausocean/av/codec/codecutil"
"bitbucket.org/ausocean/utils/logger"
)
func TestDevice(t *testing.T) {
// We want to open a device with a standard configuration.
ac := &Config{
SampleRate: 8000,
Channels: 1,
RecPeriod: 0.3,
BitDepth: 16,
Codec: codecutil.ADPCM,
}
n := 2 // Number of periods to wait while recording.
// Create a new audio Device, start, read/lex, and then stop it.
l := logger.New(logger.Debug, os.Stderr)
ai, err := NewDevice(ac, l)
// If there was an error opening the device, skip this test.
if _, ok := err.(OpenError); ok {
t.Skip(err)
}
// For any other error, report it.
if err != nil {
t.Error(err)
}
err = ai.Start()
if err != nil {
t.Error(err)
}
chunkSize := ai.ChunkSize()
lexer := codecutil.NewByteLexer(&chunkSize)
go lexer.Lex(ioutil.Discard, ai, time.Duration(ac.RecPeriod*float64(time.Second)))
time.Sleep(time.Duration(ac.RecPeriod*float64(time.Second)) * time.Duration(n))
ai.Stop()
}
var powerTests = []struct {
in int
out int
}{
{36, 32},
{47, 32},
{3, 4},
{46, 32},
{7, 8},
{2, 2},
{36, 32},
{757, 512},
{2464, 2048},
{18980, 16384},
{70000, 65536},
{8192, 8192},
{2048, 2048},
{65536, 65536},
{-2048, 1},
{-127, 1},
{-1, 1},
{0, 1},
{1, 2},
}
func TestNearestPowerOfTwo(t *testing.T) {
for _, tt := range powerTests {
t.Run(strconv.Itoa(tt.in), func(t *testing.T) {
v := nearestPowerOfTwo(tt.in)
if v != tt.out {
t.Errorf("got %v, want %v", v, tt.out)
}
})
}
}

74
revid/config.go
View File

@ -2,11 +2,9 @@
NAME
Config.go
DESCRIPTION
See Readme.md
AUTHORS
Saxon A. Nelson-Milton <saxon@ausocean.org>
Trek Hopton <trek@ausocean.org>
LICENSE
Config.go is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)
@ -30,6 +28,7 @@ package revid
import (
"errors"
"bitbucket.org/ausocean/av/codec/codecutil"
"bitbucket.org/ausocean/utils/logger"
)
@ -75,6 +74,7 @@ const (
Raspivid
V4L
RTSP
Audio
// Outputs.
RTMP
@ -93,6 +93,7 @@ const (
defaultInput = Raspivid
defaultOutput = HTTP
defaultFrameRate = 25
defaultWriteRate = 25
defaultWidth = 1280
defaultHeight = 720
defaultIntraRefreshPeriod = 100
@ -101,7 +102,7 @@ const (
defaultBitrate = 400000
defaultFramesPerClip = 1
httpFramesPerClip = 560
defaultInputCodec = H264
defaultInputCodec = codecutil.H264
defaultVerbosity = logger.Error
defaultRtpAddr = "localhost:6970"
defaultBurstPeriod = 10 // Seconds
@ -109,6 +110,12 @@ const (
defaultBrightness = 50
defaultExposure = "auto"
defaultAutoWhiteBalance = "auto"
defaultAudioInputCodec = codecutil.ADPCM
defaultSampleRate = 48000
defaultBitDepth = 16
defaultChannels = 1
defaultRecPeriod = 1.0
)
// Config provides parameters relevant to a revid instance. A new config must
@ -183,6 +190,9 @@ type Config struct {
// Raspivid input supports custom framerate.
FrameRate uint
// WriteRate is how many times a second revid encoders will be written to.
WriteRate float64
// HTTPAddress defines a custom HTTP destination if we do not wish to use that
// defined in /etc/netsender.conf.
HTTPAddress string
@ -215,6 +225,13 @@ type Config struct {
// defined at the start of the file.
AutoWhiteBalance string
// Audio
SampleRate int // Samples a second (Hz).
RecPeriod float64 // How many seconds to record at a time.
Channels int // Number of audio channels, 1 for mono, 2 for stereo.
BitDepth int // Sample bit depth.
ChunkSize int // ChunkSize is the size of the chunks in the audio.Device's ringbuffer.
RTPAddress string // RTPAddress defines the RTP output destination.
BurstPeriod uint // BurstPeriod defines the revid burst period in seconds.
Rotation uint // Rotation defines the video rotation angle in degrees Raspivid input.
@ -240,7 +257,7 @@ func (c *Config) Validate(r *Revid) error {
}
switch c.Input {
case Raspivid, V4L, File, RTSP:
case Raspivid, V4L, File, Audio, RTSP:
case NothingDefined:
c.Logger.Log(logger.Info, pkg+"no input type defined, defaulting", "input", defaultInput)
c.Input = defaultInput
@ -249,7 +266,7 @@ func (c *Config) Validate(r *Revid) error {
}
switch c.InputCodec {
case H264:
case codecutil.H264:
// FIXME(kortschak): This is not really what we want.
// Configuration really needs to be rethought here.
if c.Quantize && c.Quantization == 0 {
@ -260,18 +277,22 @@ func (c *Config) Validate(r *Revid) error {
return errors.New("bad bitrate and quantization combination for H264 input")
}
case MJPEG:
case codecutil.MJPEG:
if c.Quantization > 0 || c.Bitrate == 0 {
return errors.New("bad bitrate or quantization for mjpeg input")
}
case NothingDefined:
c.Logger.Log(logger.Info, pkg+"no input codec defined, defaulting", "inputCodec", defaultInputCodec)
c.InputCodec = defaultInputCodec
c.Logger.Log(logger.Info, pkg+"defaulting quantization", "quantization", defaultQuantization)
c.Quantization = defaultQuantization
case codecutil.PCM, codecutil.ADPCM:
default:
return errors.New("bad input codec defined in config")
switch c.Input {
case Audio:
c.Logger.Log(logger.Info, pkg+"input is audio but no codec defined, defaulting", "inputCodec", defaultAudioInputCodec)
c.InputCodec = defaultAudioInputCodec
default:
c.Logger.Log(logger.Info, pkg+"no input codec defined, defaulting", "inputCodec", defaultInputCodec)
c.InputCodec = defaultInputCodec
c.Logger.Log(logger.Info, pkg+"defaulting quantization", "quantization", defaultQuantization)
c.Quantization = defaultQuantization
}
}
if c.Outputs == nil {
@ -330,6 +351,31 @@ func (c *Config) Validate(r *Revid) error {
c.FrameRate = defaultFrameRate
}
if c.SampleRate == 0 {
c.Logger.Log(logger.Info, pkg+"no sample rate defined, defaulting", "sampleRate", defaultSampleRate)
c.SampleRate = defaultSampleRate
}
if c.Channels == 0 {
c.Logger.Log(logger.Info, pkg+"no number of channels defined, defaulting", "Channels", defaultChannels)
c.Channels = defaultChannels
}
if c.BitDepth == 0 {
c.Logger.Log(logger.Info, pkg+"no bit depth defined, defaulting", "BitDepth", defaultBitDepth)
c.BitDepth = defaultBitDepth
}
if c.RecPeriod == 0 {
c.Logger.Log(logger.Info, pkg+"no record period defined, defaulting", "recPeriod", defaultRecPeriod)
c.RecPeriod = defaultRecPeriod
}
if c.WriteRate == 0 {
c.Logger.Log(logger.Info, pkg+"no write rate defined, defaulting", "writeRate", defaultWriteRate)
c.WriteRate = defaultWriteRate
}
if c.Bitrate == 0 {
c.Logger.Log(logger.Info, pkg+"no bitrate defined, defaulting", "bitrate", defaultBitrate)
c.Bitrate = defaultBitrate

69
revid/revid.go
View File

@ -2,13 +2,11 @@
NAME
revid.go
DESCRIPTION
See Readme.md
AUTHORS
Saxon A. Nelson-Milton <saxon@ausocean.org>
Alan Noble <alan@ausocean.org>
Dan Kortschak <dan@ausocean.org>
Trek Hopton <trek@ausocean.org>
LICENSE
revid is Copyright (C) 2017-2018 the Australian Ocean Lab (AusOcean)
@ -27,6 +25,7 @@ LICENSE
in gpl.txt. If not, see http://www.gnu.org/licenses.
*/
// Package revid provides an API for reading, transcoding, and writing audio/video streams and files.
package revid
import (
@ -41,10 +40,12 @@ import (
"sync"
"time"
"bitbucket.org/ausocean/av/codec/codecutil"
"bitbucket.org/ausocean/av/codec/h264"
"bitbucket.org/ausocean/av/codec/h265"
"bitbucket.org/ausocean/av/container/flv"
"bitbucket.org/ausocean/av/container/mts"
"bitbucket.org/ausocean/av/input/audio"
"bitbucket.org/ausocean/av/protocol/rtcp"
"bitbucket.org/ausocean/av/protocol/rtp"
"bitbucket.org/ausocean/av/protocol/rtsp"
@ -173,13 +174,15 @@ func (r *Revid) reset(config Config) error {
r.config.Logger.SetLevel(config.LogLevel)
err = r.setupPipeline(
func(dst io.WriteCloser, fps int) (io.WriteCloser, error) {
func(dst io.WriteCloser, fps float64) (io.WriteCloser, error) {
var st int
switch r.config.Input {
case Raspivid, File, V4L:
st = mts.EncodeH264
case RTSP:
st = mts.EncodeH265
case Audio:
st = mts.EncodeAudio
}
e := mts.NewEncoder(dst, float64(fps), st)
return e, nil
@ -215,7 +218,7 @@ func (r *Revid) setConfig(config Config) error {
// mtsEnc and flvEnc will be called to obtain an mts encoder and flv encoder
// respectively. multiWriter will be used to create an ioext.multiWriteCloser
// so that encoders can write to multiple senders.
func (r *Revid) setupPipeline(mtsEnc, flvEnc func(dst io.WriteCloser, rate int) (io.WriteCloser, error), multiWriter func(...io.WriteCloser) io.WriteCloser) error {
func (r *Revid) setupPipeline(mtsEnc func(dst io.WriteCloser, rate float64) (io.WriteCloser, error), flvEnc func(dst io.WriteCloser, rate int) (io.WriteCloser, error), multiWriter func(...io.WriteCloser) io.WriteCloser) error {
// encoders will hold the encoders that are required for revid's current
// configuration.
var encoders []io.WriteCloser
@ -259,7 +262,7 @@ func (r *Revid) setupPipeline(mtsEnc, flvEnc func(dst io.WriteCloser, rate int)
// as a destination.
if len(mtsSenders) != 0 {
mw := multiWriter(mtsSenders...)
e, _ := mtsEnc(mw, int(r.config.FrameRate))
e, _ := mtsEnc(mw, r.config.WriteRate)
encoders = append(encoders, e)
}
@ -289,6 +292,9 @@ func (r *Revid) setupPipeline(mtsEnc, flvEnc func(dst io.WriteCloser, rate int)
case RTSP:
r.setupInput = r.startRTSPCamera
r.lexTo = h265.NewLexer(false).Lex
case Audio:
r.setupInput = r.startAudioDevice
r.lexTo = codecutil.NewByteLexer(&r.config.ChunkSize).Lex
}
return nil
@ -533,7 +539,7 @@ func (r *Revid) startRaspivid() (func() error, error) {
switch r.config.InputCodec {
default:
return nil, fmt.Errorf("revid: invalid input codec: %v", r.config.InputCodec)
case H264:
case codecutil.H264:
args = append(args,
"--codec", "H264",
"--inline",
@ -542,7 +548,7 @@ func (r *Revid) startRaspivid() (func() error, error) {
if r.config.Quantize {
args = append(args, "-qp", fmt.Sprint(r.config.Quantization))
}
case MJPEG:
case codecutil.MJPEG:
args = append(args, "--codec", "MJPEG")
}
r.config.Logger.Log(logger.Info, pkg+"raspivid args", "raspividArgs", strings.Join(args, " "))
@ -615,10 +621,55 @@ func (r *Revid) setupInputForFile() (func() error, error) {
// TODO(kortschak): Maybe we want a context.Context-aware parser that we can stop.
r.wg.Add(1)
go r.processFrom(f, time.Second/time.Duration(r.config.FrameRate))
go r.processFrom(f, 0)
return func() error { return f.Close() }, nil
}
// startAudioDevice is used to start capturing audio from an audio device and processing it.
// It returns a function that can be used to stop the device and any errors that occur.
func (r *Revid) startAudioDevice() (func() error, error) {
// Create audio device.
ac := &audio.Config{
SampleRate: r.config.SampleRate,
Channels: r.config.Channels,
RecPeriod: r.config.RecPeriod,
BitDepth: r.config.BitDepth,
Codec: r.config.InputCodec,
}
mts.Meta.Add("sampleRate", strconv.Itoa(r.config.SampleRate))
mts.Meta.Add("channels", strconv.Itoa(r.config.Channels))
mts.Meta.Add("period", fmt.Sprintf("%.6f", r.config.RecPeriod))
mts.Meta.Add("bitDepth", strconv.Itoa(r.config.BitDepth))
switch r.config.InputCodec {
case codecutil.PCM:
mts.Meta.Add("codec", "pcm")
case codecutil.ADPCM:
mts.Meta.Add("codec", "adpcm")
default:
r.config.Logger.Log(logger.Fatal, pkg+"no audio codec set in config")
}
ai, err := audio.NewDevice(ac, r.config.Logger)
if err != nil {
r.config.Logger.Log(logger.Fatal, pkg+"failed to create audio device", "error", err.Error())
}
// Start audio device
err = ai.Start()
if err != nil {
r.config.Logger.Log(logger.Fatal, pkg+"failed to start audio device", "error", err.Error())
}
// Process output from audio device.
r.config.ChunkSize = ai.ChunkSize()
r.wg.Add(1)
go r.processFrom(ai, time.Duration(float64(time.Second)/r.config.WriteRate))
return func() error {
ai.Stop()
return nil
}, nil
}
// startRTSPCamera uses RTSP to request an RTP stream from an IP camera. An RTP
// client is created from which RTP packets containing either h264/h265 can read
// by the selected lexer.

4
revid/revid_test.go
View File

@ -41,7 +41,7 @@ import (
const raspividPath = "/usr/local/bin/raspivid"
// Suppress all test logging, except for t.Errorf output.
var silent bool
var silent = true
// TestRaspivid tests that raspivid starts correctly.
// It is intended to be run on a Raspberry Pi.
@ -232,7 +232,7 @@ func TestResetEncoderSenderSetup(t *testing.T) {
// This logic is what we want to check.
err = rv.setupPipeline(
func(dst io.WriteCloser, rate int) (io.WriteCloser, error) {
func(dst io.WriteCloser, rate float64) (io.WriteCloser, error) {
return &tstMtsEncoder{dst: dst}, nil
},
func(dst io.WriteCloser, rate int) (io.WriteCloser, error) {

2
revid/senders.go
View File

@ -57,7 +57,7 @@ type httpSender struct {
log func(lvl int8, msg string, args ...interface{})
}
// newMinimalHttpSender returns a pointer to a new minimalHttpSender.
// newHttpSender returns a pointer to a new httpSender.
func newHttpSender(ns *netsender.Sender, log func(lvl int8, msg string, args ...interface{})) *httpSender {
return &httpSender{
client: ns,