原文地址:https://blog.csdn.net/zero_sama/article/details/69802783
FLV 格式解析
FLV是由一个FLV Header 和 若干tag(Video Tag, Audio Tag, Script Tag三种,分别代表视频流,音频流和脚本流)组成的二进制文件。
FLV Header示意图
FLV Header:
- 文件类型: 固定为 “FLV” (3 bytes)
- 版本信息: 一般为0x01 (1 byte)
- 流信息: 0000 0101 此flv文件包含视音频, 0000 0001 此flv文件包含视频 0000 0100 包含音频 (1 byte)
- 头长度: FLV文件头长度,一般为 3+1+1+4=9 bytes (4bytes)
FLV Body:
Body由一系列pre tag length 和 tag组成。
+----------------------------------------------------------------------+
| Pre Tag Length | Tag Header | Tag Data | .... | Pre Tag Length | ... |
+----------------------------------------------------------------------+
- Pre Tag Length: 前一个tag的长度 4 bytes
- Tag Header: 1 + 3 + 3 +1 + 3 = 11 bytes
Tag:
tag header (11 bytes)
+-------------------------------------------------------------------------------------------------------------+
| Tag Type(1 byte) | Tag Data Length(3 bytes) | Timestap(3 bytes) | TimestapExt(1 byte) | StreamID(3 bytes) |
+-------------------------------------------------------------------------------------------------------------+
字段名 | 长度 | 取值 |
---|---|---|
Tag Type Tag 类型 | 1 byte | 0x08 音频 0x09 视频 0x12 脚本 |
Tag Data Length: Tag Data 长度 | 3 bytes | |
Timestamp 时间戳(单位ms) | 3 bytes | |
TimestampExt 扩展时间戳 | 1 byte | |
StreamID 流ID 总是0 | 3 bytes |
tag data
tag data如果是音频数据
第一个byte记录audio信息。前4bits表示音频格式(全部格式请看官方文档):
hex | comment |
---|---|
0 | 未压缩 |
1 | ADPCM |
2 | MP3 |
4 | Nellymoser 16-kHz mono |
5 | Nellymoser 8-kHz mono |
10 | AAC |
下面两个bits表示samplerate:
hex | comment |
---|---|
0 | 5.5KHz |
1 | 11kHz |
2 | 22kHz |
3 | 44kHz |
下面1bit表示采样长度:
hex | comment |
---|---|
0 | snd8Bit |
1 | snd16Bit |
下面1bit表示类型:
hex | comment |
---|---|
0 | sndMomo |
1 | sndStereo |
之后是数据。
tag data如果是视频数据
第一个byte记录video信息。前4bits表示视频帧类型:
hex | comment |
---|---|
1 | keyframe |
2 | inner frame |
3 | disposable inner frame (h.263 only) |
4 | generated keyframe |
后4bits表示解码器ID:
hex | comment |
---|---|
2 | seronson h.263 |
3 | screen video |
4 | On2 VP6 |
5 | On2 VP6 with alpha channel |
6 | Screen video version 2 |
7 | AVC (h.264) |
FLV文件解析
数据 | 说明 |
---|---|
00 00 00 00 | 前一个tag的长度,由于是第一个tag所以全为0. |
12 | 表示tag类型为脚本 |
00 00 F6 | 表示tag data长度为0xF6个字节 |
00 00 00 | 时间戳 |
00 | 扩展时间戳 |
00 00 00 | 流ID |
Video Tag
Audio Tag
RTMP抓包的视频流:
RTMP视频流格式与flv很相似,就是video tag 和 audio tag的tag data一个接一个的发送(不含tag header 和 pre tag length)。
- audio /video信息:1 字节 ,这里0x17 表示I帧 AVC
- AVC packet type :1字节
- 0x00:AVC Sequence Header
- 0x01:AVC NALU
- composition time : 3字节, AVC时无意义,全为0
当AVC packet type为AVC Sequence Header时:
接下来就是AVCDecoderConfigurationRecord的内容:
type | length(byte) | value | comment |
---|---|---|---|
configurationVersion | 1 | 0x01 | 版本 |
AVCProfileIndication | 1 | 0x4d | sps[1] |
profile_compatibility | 1 | 0x00 | sps[2] |
AVCLevelIndication | 1 | 0x2a | sps[3] |
lengthSizeMinusOne | 1 | 0xff | FLV中NALU包长数据所使用的字节数, 包长= (lengthSizeMinusOne & 3) + 1 |
numOfSequenceParameterSets | 1 | 0xe1 | SPS个数,通常为0xe1 个数= numOfSequenceParameterSets & 01F |
sequenceParameterSetLength | 2 | 0x0014 | SPS长度 |
sequenceParameterSetNALUnits | SPS内容 | ||
numOfPictureParameterSets | 1 | 0x01 | PPS个数,通常为0x01 |
pictureParameterSetLength | 2 | 0x0004 | PPS长度 |
pictureParameterSetNALUnits | PPS内容 |
当AVC packet type为AVC NALU(0x01)时:
接下来就是NALU的格式,如下:
NALU length:(lengthSizeMinusOne & 3) + 1字节 NALU长度
NALU Data:
NALU length:
NALU Data:
…
NALU Data包含H264编码数据,详解如下:
H264 SPS(Sequence Parameter Set) 和PPS(Picture Parameter Set) 数据结构,有如下H264bitstream
/* h.264 bitstreams */
const uint8_t sps[] =
{0x00, 0x00, 0x00, 0x01, 0x67, 0x42, 0x00, 0x0a, 0xf8, 0x41, 0xa2}; //0x00 00 00 01 或者 0x00 00 01 是分隔符
const uint8_t pps[] =
{0x00, 0x00, 0x00, 0x01, 0x68, 0xce, 0x38, 0x80};
代码中SPS 2进制如下:
SPS各位详述如下:
Parameter Name | Type u:unsigned bit ue:指数哥伦布码 | Value | Comments |
---|---|---|---|
forbidden_zero_bit | u(1) | 0 | Despite being forbidden, it must be set to 0! |
nal_ref_idc | u(2) | 3 | 3 means it is “important” (this is an SPS) |
nal_unit_type | u(5) | 7 | Indicates this is a sequence parameter set |
profile_idc | u(8) | 66 | Baseline profile |
constraint_set0_flag | u(1) | 0 | We’re not going to honor constraints |
constraint_set1_flag | u(1) | 0 | We’re not going to honor constraints |
constraint_set2_flag | u(1) | 0 | We’re not going to honor constraints |
constraint_set3_flag | u(1) | 0 | We’re not going to honor constraints |
reserved_zero_4bits | u(4) | 0 | Better set them to zero |
level_idc | u(8) | 10 | Level 1, sec A.3.1 |
seq_parameter_set_id | ue(v) | 0 | We’ll just use id 0. |
log2_max_frame_num_minus4 | ue(v) | 0 | Let’s have as few frame numbers as possible |
pic_order_cnt_type | ue(v) | 0 | Keep things simple |
log2_max_pic_order_cnt_lsb_minus4 | ue(v) | 0 | Fewer is better. |
num_ref_frames | ue(v) | 0 | We will only send I slices |
gaps_in_frame_num_value_allowed_flag | u(1) | 0 | We will have no gaps |
pic_width_in_mbs_minus_1 | ue(v) | 7 | SQCIF is 8 macroblocks wide |
pic_height_in_map_units_minus_1 | ue(v) | 5 | SQCIF is 6 macroblocks high |
frame_mbs_only_flag | u(1) | 1 | We will not to field/frame encoding |
direct_8x8_inference_flag | u(1) | 0 | Used for B slices. We will not send B slices |
frame_cropping_flag | u(1) | 0 | We will not do frame cropping |
vui_prameters_present_flag | u(1) | 0 | We will not send VUI data |
rbsp_stop_one_bit | u(1) | 1 | Stop bit. I missed this at first and it caused me much trouble. |
贴几段代码:将RTSP回调的H264裸流转换并通过RTMP协议发送到客户端.
// pData: H264裸流数据
// 将H264裸流转换成适用于RTMP协议的流并发送
void CRTMP::_DataCallBack( LONG nChannel, char* pData, LONG nSize, RTP_HEAD* pHead, RTP_FRAME_TYPE FrameType )
{
char* p = pData;
char* pEnd = pData + nSize;
if (m_bNeedUpdateMetadata)
{ // 第一次收到I帧时,获取sps pps并保存
if (FrameType != RTP_FRAME_TYPE_I)
{
return;
}
Nalu sps, pps;
p = ReadOneNalu(pData, nSize, sps); ASSERT(p != NULL);
p = ReadOneNalu(p, pEnd - p, pps); ASSERT(p != NULL);
h264_decode_sps((BYTE*)sps.data, sps.size, m_nWitdh, m_nHeight, m_nFps);
m_MetaData.nSpsLen = sps.size;
m_MetaData.Sps = (unsigned char*)calloc(sps.size, 1); memcpy(m_MetaData.Sps, sps.data, sps.size);
m_MetaData.nPpsLen = pps.size;
m_MetaData.Pps = (unsigned char *)calloc(pps.size, 1); memcpy(m_MetaData.Pps, pps.data, pps.size);
m_MetaData.nWidth = m_nWitdh;
m_MetaData.nHeight = m_nHeight;
m_MetaData.nFrameRate = m_nFps;
m_bNeedUpdateMetadata = false;
}
unsigned int tick_gap = 1000 / m_MetaData.nFrameRate; // 1000ms/fps
Nalu idr;
while ((p = ReadOneNalu(p, pEnd - p, idr)) != NULL)
{
if (idr.type == 0x07 || idr.type == 0x08)
{ // 忽略重复的sps pps
continue;
}
if (!SendH264Packet(idr, m_tick))
{
cout << "SendH264Packet Failed" << WSAGetLastError() << endl;
}
}
m_tick += tick_gap; // 更新时间戳
}
// 从缓存读取一个nalu
// 返回指向下一个nalu起始位置的指针或者NULL
static char* ReadOneNalu(char* pBuf, int nSize, Nalu& NaluUnit)
{
char Sep1[3], Sep2[4];
Sep1[0] = (char)(0x1 >> 16 & 0xFF);
Sep1[1] = (char)(0x1 >> 8 & 0xFF);
Sep1[2] = (char)(0x1 & 0xFF);
Sep2[0] = (char)(0x1 >> 24 & 0xFF);
Sep2[1] = (char)(0x1 >> 16 & 0xFF);
Sep2[2] = (char)(0x1 >> 8 & 0xFF);
Sep2[3] = (char)(0x1 & 0xFF);
bool bFindHead = false;
bool bFindTail = false;
char* pStart = pBuf;
char* pEnd = pBuf + nSize;
_FIND:
while (pStart <= pEnd - 4)
{ // nalu 以00 00 01 或者00 00 00 01作为分隔符
// look for head
if (!bFindHead
&&(::memcmp(Sep1, pStart, 3) == 0))
{
pStart += 3;
NaluUnit.data = pStart;
NaluUnit.type = NaluUnit.data[0] & 0x1F;
bFindHead = true;
}
else if (!bFindHead
&& (::memcmp(Sep2, pStart, 4) == 0))
{
pStart += 4;
NaluUnit.data = pStart;
NaluUnit.type = NaluUnit.data[0] & 0x1F;
bFindHead = true;
}
// look for tail
if (bFindHead)
{
if (pEnd - pStart < 3)
{
break;
}
if (::memcmp(Sep1, pStart, 3) == 0)
{
NaluUnit.size = pStart - NaluUnit.data;
bFindTail = true;
break;
}
else if (::memcmp(Sep2, pStart, 4) == 0)
{
NaluUnit.size = pStart - NaluUnit.data;
bFindTail = true;
break;
}
}
++pStart;
}
if (!bFindTail)
{
NaluUnit.size = pEnd - NaluUnit.data;
}
if ((NaluUnit.type & 0x1F) == 0x06)
{ // sei 跳过
bFindHead = false;
bFindTail = false;
goto _FIND;
}
return bFindHead ? NaluUnit.data + NaluUnit.size : NULL;
}
// 将nalu封包并发送
// 大端法存储
bool CRTMP::SendH264Packet(Nalu nalunit, unsigned int timestamp)
{
CAMFBuffer buf(nalunit.size + 9);
RTMP_Packet p;
if (nalunit.type == 5)
{
buf.WriteByte(0x17); // I 帧
buf.WriteByte(0x1); // nalu
buf.WriteInt24(0x0);
buf.WriteInt32(nalunit.size);
buf.WriteBuffer(nalunit.data, nalunit.size);
SendSpsPpsInfo(m_MetaData.Pps, m_MetaData.nPpsLen, m_MetaData.Sps, m_MetaData.nSpsLen);
}
else
{
buf.WriteByte(0x27); // p、b 帧
buf.WriteByte(0x1); // nalu
buf.WriteInt24(0x0);
buf.WriteInt32(nalunit.size);
buf.WriteBuffer(nalunit.data, nalunit.size);
}
return SendPacket(buf, 0x9, timestamp);
}
// 发送sps,pps信息
bool CRTMP::SendSpsPpsInfo(unsigned char *pps,int pps_len,unsigned char * sps,int sps_len)
{
CAMFBuffer buffer(pps_len + sps_len + 16);
buffer.WriteByte(0x17);
buffer.WriteByte(0x0); // AVCSequenceHeader
buffer.WriteInt24(0x0); // composition time
buffer.WriteByte(0x01);
buffer.WriteByte(sps[1]);
buffer.WriteByte(sps[2]);
buffer.WriteByte(sps[3]);
buffer.WriteByte((char)0xFF); // lengthSizeMinusOne NALU包长所用字节数=(lengthSizeMinusOne & 3) + 1
buffer.WriteByte((char)0xE1); // numOfSequenceParameterSet SPS个数=numOfSequenceParameterSet & 0x1F
buffer.WriteInt16(sps_len); // SPS 长度
buffer.WriteBuffer((char*)sps, sps_len);
buffer.WriteByte((char)0x01); // numOfPictureParameterSet PPS个数=numOfPictureParameterSet & 0x1F
buffer.WriteInt16(pps_len); // PPS 长度
buffer.WriteBuffer((char*)pps, pps_len);
return SendPacket(buffer, 0x9, 0);
}