基于RTP的H264视频数据打包解包类
睿丰德科技 专注RFID识别技术和条码识别技术与管理软件的集成项目。质量追溯系统、MES系统、金蝶与条码系统对接、用友与条码系统对接
from:http://blog.csdn.net/dengzikun/article/details/5807694
最近考虑使用RTP替换原有的高清视频传输协议,遂上网查找有关H264视频RTP打包、解包的文档和代码。功夫不负有心人,找到不少有价值的文档和代码。参考这些资料,写了H264 RTP打包类、解包类,实现了单个NAL单元包和FU_A分片单元包。对于丢包处理,采用简单的策略:丢弃随后的所有数据包,直到收到关键帧。测试效果还不错,代码贴上来,若能为同道中人借鉴一二,足矣。两个类的使用说明如下(省略了错误处理过程):
DWORD H264SSRC ;
CH264_RTP_PACK pack ( H264SSRC ) ;
BYTE *pVideoData ;
DWORD Size, ts ;
bool IsEndOfFrame ;
WORD wLen ;
pack.Set ( pVideoData, Size, ts, IsEndOfFrame ) ;
BYTE *pPacket ;
while ( pPacket = pack.Get ( &wLen ) )
{
// rtp packet process
// ...
}
HRESULT hr ;
CH264_RTP_UNPACK unpack ( hr ) ;
BYTE *pRtpData ;
WORD inSize;
int outSize ;
BYTE *pFrame = unpack.Parse_RTP_Packet ( pRtpData, inSize, &outSize ) ;
if ( pFrame != NULL )
{
// frame process
// ...
}
[cpp] view plaincopy
- //////////////////////////////////////////////////////////////////////////////////////////
- // class CH264_RTP_PACK start
- class CH264_RTP_PACK
- {
- #define RTP_VERSION 2
- typedef struct NAL_msg_s
- {
- bool eoFrame ;
- unsigned char type; // NAL type
- unsigned char *start; // pointer to first location in the send buffer
- unsigned char *end; // pointer to last location in send buffer
- unsigned long size ;
- } NAL_MSG_t;
- typedef struct
- {
- //LITTLE_ENDIAN
- unsigned short cc:4; /* CSRC count */
- unsigned short x:1; /* header extension flag */
- unsigned short p:1; /* padding flag */
- unsigned short v:2; /* packet type */
- unsigned short pt:7; /* payload type */
- unsigned short m:1; /* marker bit */
- unsigned short seq; /* sequence number */
- unsigned long ts; /* timestamp */
- unsigned long ssrc; /* synchronization source */
- } rtp_hdr_t;
- typedef struct tagRTP_INFO
- {
- NAL_MSG_t nal; // NAL information
- rtp_hdr_t rtp_hdr; // RTP header is assembled here
- int hdr_len; // length of RTP header
- unsigned char *pRTP; // pointer to where RTP packet has beem assembled
- unsigned char *start; // pointer to start of payload
- unsigned char *end; // pointer to end of payload
- unsigned int s_bit; // bit in the FU header
- unsigned int e_bit; // bit in the FU header
- bool FU_flag; // fragmented NAL Unit flag
- } RTP_INFO;
- public:
- CH264_RTP_PACK(unsigned long H264SSRC, unsigned char H264PAYLOADTYPE=96, unsigned short MAXRTPPACKSIZE=1472 )
- {
- m_MAXRTPPACKSIZE = MAXRTPPACKSIZE ;
- if ( m_MAXRTPPACKSIZE > 10000 )
- {
- m_MAXRTPPACKSIZE = 10000 ;
- }
- if ( m_MAXRTPPACKSIZE < 50 )
- {
- m_MAXRTPPACKSIZE = 50 ;
- }
- memset ( &m_RTP_Info, 0, sizeof(m_RTP_Info) ) ;
- m_RTP_Info.rtp_hdr.pt = H264PAYLOADTYPE ;
- m_RTP_Info.rtp_hdr.ssrc = H264SSRC ;
- m_RTP_Info.rtp_hdr.v = RTP_VERSION ;
- m_RTP_Info.rtp_hdr.seq = 0 ;
- }
- ~CH264_RTP_PACK(void)
- {
- }
- //传入Set的数据必须是一个完整的NAL,起始码为0x00000001。
- //起始码之前至少预留10个字节,以避免内存COPY操作。
- //打包完成后,原缓冲区内的数据被破坏。
- bool Set ( unsigned char *NAL_Buf, unsigned long NAL_Size, unsigned long Time_Stamp, bool End_Of_Frame )
- {
- unsigned long startcode = StartCode(NAL_Buf) ;
- if ( startcode != 0x01000000 )
- {
- return false ;
- }
- int type = NAL_Buf[4] & 0x1f ;
- if ( type < 1 || type > 12 )
- {
- return false ;
- }
- m_RTP_Info.nal.start = NAL_Buf ;
- m_RTP_Info.nal.size = NAL_Size ;
- m_RTP_Info.nal.eoFrame = End_Of_Frame ;
- m_RTP_Info.nal.type = m_RTP_Info.nal.start[4] ;
- m_RTP_Info.nal.end = m_RTP_Info.nal.start + m_RTP_Info.nal.size ;
- m_RTP_Info.rtp_hdr.ts = Time_Stamp ;
- m_RTP_Info.nal.start += 4 ; // skip the syncword
- if ( (m_RTP_Info.nal.size + 7) > m_MAXRTPPACKSIZE )
- {
- m_RTP_Info.FU_flag = true ;
- m_RTP_Info.s_bit = 1 ;
- m_RTP_Info.e_bit = 0 ;
- m_RTP_Info.nal.start += 1 ; // skip NAL header
- }
- else
- {
- m_RTP_Info.FU_flag = false ;
- m_RTP_Info.s_bit = m_RTP_Info.e_bit = 0 ;
- }
- m_RTP_Info.start = m_RTP_Info.end = m_RTP_Info.nal.start ;
- m_bBeginNAL = true ;
- return true ;
- }
- //循环调用Get获取RTP包,直到返回值为NULL
- unsigned char* Get ( unsigned short *pPacketSize )
- {
- if ( m_RTP_Info.end == m_RTP_Info.nal.end )
- {
- *pPacketSize = 0 ;
- return NULL ;
- }
- if ( m_bBeginNAL )
- {
- m_bBeginNAL = false ;
- }
- else
- {
- m_RTP_Info.start = m_RTP_Info.end; // continue with the next RTP-FU packet
- }
- int bytesLeft = m_RTP_Info.nal.end - m_RTP_Info.start ;
- int maxSize = m_MAXRTPPACKSIZE - 12 ; // sizeof(basic rtp header) == 12 bytes
- if ( m_RTP_Info.FU_flag )
- maxSize -= 2 ;
- if ( bytesLeft > maxSize )
- {
- m_RTP_Info.end = m_RTP_Info.start + maxSize ; // limit RTP packetsize to 1472 bytes
- }
- else
- {
- m_RTP_Info.end = m_RTP_Info.start + bytesLeft ;
- }
- if ( m_RTP_Info.FU_flag )
- { // multiple packet NAL slice
- if ( m_RTP_Info.end == m_RTP_Info.nal.end )
- {
- m_RTP_Info.e_bit = 1 ;
- }
- }
- m_RTP_Info.rtp_hdr.m = m_RTP_Info.nal.eoFrame ? 1 : 0 ; // should be set at EofFrame
- if ( m_RTP_Info.FU_flag && !m_RTP_Info.e_bit )
- {
- m_RTP_Info.rtp_hdr.m = 0 ;
- }
- m_RTP_Info.rtp_hdr.seq++ ;
- unsigned char *cp = m_RTP_Info.start ;
- cp -= ( m_RTP_Info.FU_flag ? 14 : 12 ) ;
- m_RTP_Info.pRTP = cp ;
- unsigned char *cp2 = (unsigned char *)&m_RTP_Info.rtp_hdr ;
- cp[0] = cp2[0] ;
- cp[1] = cp2[1] ;
- cp[2] = ( m_RTP_Info.rtp_hdr.seq >> 8 ) & 0xff ;
- cp[3] = m_RTP_Info.rtp_hdr.seq & 0xff ;
- cp[4] = ( m_RTP_Info.rtp_hdr.ts >> 24 ) & 0xff ;
- cp[5] = ( m_RTP_Info.rtp_hdr.ts >> 16 ) & 0xff ;
- cp[6] = ( m_RTP_Info.rtp_hdr.ts >> 8 ) & 0xff ;
- cp[7] = m_RTP_Info.rtp_hdr.ts & 0xff ;
- cp[8] = ( m_RTP_Info.rtp_hdr.ssrc >> 24 ) & 0xff ;
- cp[9] = ( m_RTP_Info.rtp_hdr.ssrc >> 16 ) & 0xff ;
- cp[10] = ( m_RTP_Info.rtp_hdr.ssrc >> 8 ) & 0xff ;
- cp[11] = m_RTP_Info.rtp_hdr.ssrc & 0xff ;
- m_RTP_Info.hdr_len = 12 ;
- /*!
- * /n The FU indicator octet has the following format:
- * /n
- * /n +---------------+
- * /n MSB |0|1|2|3|4|5|6|7| LSB
- * /n +-+-+-+-+-+-+-+-+
- * /n |F|NRI| Type |
- * /n +---------------+
- * /n
- * /n The FU header has the following format:
- * /n
- * /n +---------------+
- * /n |0|1|2|3|4|5|6|7|
- * /n +-+-+-+-+-+-+-+-+
- * /n |S|E|R| Type |
- * /n +---------------+
- */
- if ( m_RTP_Info.FU_flag )
- {
- // FU indicator F|NRI|Type
- cp[12] = ( m_RTP_Info.nal.type & 0xe0 ) | 28 ; //Type is 28 for FU_A
- //FU header S|E|R|Type
- cp[13] = ( m_RTP_Info.s_bit << 7 ) | ( m_RTP_Info.e_bit << 6 ) | ( m_RTP_Info.nal.type & 0x1f ) ; //R = 0, must be ignored by receiver
- m_RTP_Info.s_bit = m_RTP_Info.e_bit= 0 ;
- m_RTP_Info.hdr_len = 14 ;
- }
- m_RTP_Info.start = &cp[m_RTP_Info.hdr_len] ; // new start of payload
- *pPacketSize = m_RTP_Info.hdr_len + ( m_RTP_Info.end - m_RTP_Info.start ) ;
- return m_RTP_Info.pRTP ;
- }
- private:
- unsigned int StartCode( unsigned char *cp )
- {
- unsigned int d32 ;
- d32 = cp[3] ;
- d32 <<= 8 ;
- d32 |= cp[2] ;
- d32 <<= 8 ;
- d32 |= cp[1] ;
- d32 <<= 8 ;
- d32 |= cp[0] ;
- return d32 ;
- }
- private:
- RTP_INFO m_RTP_Info ;
- bool m_bBeginNAL ;
- unsigned short m_MAXRTPPACKSIZE ;
- };
- // class CH264_RTP_PACK end
- //////////////////////////////////////////////////////////////////////////////////////////
- //////////////////////////////////////////////////////////////////////////////////////////
- // class CH264_RTP_UNPACK start
- class CH264_RTP_UNPACK
- {
- #define RTP_VERSION 2
- #define BUF_SIZE (1024 * 500)
- typedef struct
- {
- //LITTLE_ENDIAN
- unsigned short cc:4; /* CSRC count */
- unsigned short x:1; /* header extension flag */
- unsigned short p:1; /* padding flag */
- unsigned short v:2; /* packet type */
- unsigned short pt:7; /* payload type */
- unsigned short m:1; /* marker bit */
- unsigned short seq; /* sequence number */
- unsigned long ts; /* timestamp */
- unsigned long ssrc; /* synchronization source */
- } rtp_hdr_t;
- public:
- CH264_RTP_UNPACK ( HRESULT &hr, unsigned char H264PAYLOADTYPE = 96 )
- : m_bSPSFound(false)
- , m_bWaitKeyFrame(true)
- , m_bPrevFrameEnd(false)
- , m_bAssemblingFrame(false)
- , m_wSeq(1234)
- , m_ssrc(0)
- {
- m_pBuf = new BYTE[BUF_SIZE] ;
- if ( m_pBuf == NULL )
- {
- hr = E_OUTOFMEMORY ;
- return ;
- }
- m_H264PAYLOADTYPE = H264PAYLOADTYPE ;
- m_pEnd = m_pBuf + BUF_SIZE ;
- m_pStart = m_pBuf ;
- m_dwSize = 0 ;
- hr = S_OK ;
- }
- ~CH264_RTP_UNPACK(void)
- {
- delete [] m_pBuf ;
- }
- //pBuf为H264 RTP视频数据包,nSize为RTP视频数据包字节长度,outSize为输出视频数据帧字节长度。
- //返回值为指向视频数据帧的指针。输入数据可能被破坏。
- BYTE* Parse_RTP_Packet ( BYTE *pBuf, unsigned short nSize, int *outSize )
- {
- if ( nSize <= 12 )
- {
- return NULL ;
- }
- BYTE *cp = (BYTE*)&m_RTP_Header ;
- cp[0] = pBuf[0] ;
- cp[1] = pBuf[1] ;
- m_RTP_Header.seq = pBuf[2] ;
- m_RTP_Header.seq <<= 8 ;
- m_RTP_Header.seq |= pBuf[3] ;
- m_RTP_Header.ts = pBuf[4] ;
- m_RTP_Header.ts <<= 8 ;
- m_RTP_Header.ts |= pBuf[5] ;
- m_RTP_Header.ts <<= 8 ;
- m_RTP_Header.ts |= pBuf[6] ;
- m_RTP_Header.ts <<= 8 ;
- m_RTP_Header.ts |= pBuf[7] ;
- m_RTP_Header.ssrc = pBuf[8] ;
- m_RTP_Header.ssrc <<= 8 ;
- m_RTP_Header.ssrc |= pBuf[9] ;
- m_RTP_Header.ssrc <<= 8 ;
- m_RTP_Header.ssrc |= pBuf[10] ;
- m_RTP_Header.ssrc <<= 8 ;
- m_RTP_Header.ssrc |= pBuf[11] ;
- BYTE *pPayload = pBuf + 12 ;
- DWORD PayloadSize = nSize - 12 ;
- // Check the RTP version number (it should be 2):
- if ( m_RTP_Header.v != RTP_VERSION )
- {
- return NULL ;
- }
- /*
- // Skip over any CSRC identifiers in the header:
- if ( m_RTP_Header.cc )
- {
- long cc = m_RTP_Header.cc * 4 ;
- if ( Size < cc )
- {
- return NULL ;
- }
- Size -= cc ;
- p += cc ;
- }
- // Check for (& ignore) any RTP header extension
- if ( m_RTP_Header.x )
- {
- if ( Size < 4 )
- {
- return NULL ;
- }
- Size -= 4 ;
- p += 2 ;
- long l = p[0] ;
- l <<= 8 ;
- l |= p[1] ;
- p += 2 ;
- l *= 4 ;
- if ( Size < l ) ;
- {
- return NULL ;
- }
- Size -= l ;
- p += l ;
- }
- // Discard any padding bytes:
- if ( m_RTP_Header.p )
- {
- if ( Size == 0 )
- {
- return NULL ;
- }
- long Padding = p[Size-1] ;
- if ( Size < Padding )
- {
- return NULL ;
- }
- Size -= Padding ;
- }*/
- // Check the Payload Type.
- if ( m_RTP_Header.pt != m_H264PAYLOADTYPE )
- {
- return NULL ;
- }
- int PayloadType = pPayload[0] & 0x1f ;
- int NALType = PayloadType ;
- if ( NALType == 28 ) // FU_A
- {
- if ( PayloadSize < 2 )
- {
- return NULL ;
- }
- NALType = pPayload[1] & 0x1f ;
- }
- if ( m_ssrc != m_RTP_Header.ssrc )
- {
- m_ssrc = m_RTP_Header.ssrc ;
- SetLostPacket () ;
- }
- if ( NALType == 0x07 ) // SPS
- {
- m_bSPSFound = true ;
- }
- if ( !m_bSPSFound )
- {
- return NULL ;
- }
- if ( NALType == 0x07 || NALType == 0x08 ) // SPS PPS
- {
- m_wSeq = m_RTP_Header.seq ;
- m_bPrevFrameEnd = true ;
- pPayload -= 4 ;
- *((DWORD*)(pPayload)) = 0x01000000 ;
- *outSize = PayloadSize + 4 ;
- return pPayload ;
- }
- if ( m_bWaitKeyFrame )
- {
- if ( m_RTP_Header.m ) // frame end
- {
- m_bPrevFrameEnd = true ;
- if ( !m_bAssemblingFrame )
- {
- m_wSeq = m_RTP_Header.seq ;
- return NULL ;
- }
- }
- if ( !m_bPrevFrameEnd )
- {
- m_wSeq = m_RTP_Header.seq ;
- return NULL ;
- }
- else
- {
- if ( NALType != 0x05 ) // KEY FRAME
- {
- m_wSeq = m_RTP_Header.seq ;
- m_bPrevFrameEnd = false ;
- return NULL ;
- }
- }
- }
- ///////////////////////////////////////////////////////////////
- if ( m_RTP_Header.seq != (WORD)( m_wSeq + 1 ) ) // lost packet
- {
- m_wSeq = m_RTP_Header.seq ;
- SetLostPacket () ;
- return NULL ;
- }
- else
- {
- // 码流正常
- m_wSeq = m_RTP_Header.seq ;
- m_bAssemblingFrame = true ;
- if ( PayloadType != 28 ) // whole NAL
- {
- *((DWORD*)(m_pStart)) = 0x01000000 ;
- m_pStart += 4 ;
- m_dwSize += 4 ;
- }
- else // FU_A
- {
- if ( pPayload[1] & 0x80 ) // FU_A start
- {
- *((DWORD*)(m_pStart)) = 0x01000000 ;
- m_pStart += 4 ;
- m_dwSize += 4 ;
- pPayload[1] = ( pPayload[0] & 0xE0 ) | NALType ;
- pPayload += 1 ;
- PayloadSize -= 1 ;
- }
- else
- {
- pPayload += 2 ;
- PayloadSize -= 2 ;
- }
- }
- if ( m_pStart + PayloadSize < m_pEnd )
- {
- CopyMemory ( m_pStart, pPayload, PayloadSize ) ;
- m_dwSize += PayloadSize ;
- m_pStart += PayloadSize ;
- }
- else // memory overflow
- {
- SetLostPacket () ;
- return NULL ;
- }
- if ( m_RTP_Header.m ) // frame end
- {
- *outSize = m_dwSize ;
- m_pStart = m_pBuf ;
- m_dwSize = 0 ;
- if ( NALType == 0x05 ) // KEY FRAME
- {
- m_bWaitKeyFrame = false ;
- }
- return m_pBuf ;
- }
- else
- {
- return NULL ;
- }
- }
- }
- void SetLostPacket()
- {
- m_bSPSFound = false ;
- m_bWaitKeyFrame = true ;
- m_bPrevFrameEnd = false ;
- m_bAssemblingFrame = false ;
- m_pStart = m_pBuf ;
- m_dwSize = 0 ;
- }
- private:
- rtp_hdr_t m_RTP_Header ;
- BYTE *m_pBuf ;
- bool m_bSPSFound ;
- bool m_bWaitKeyFrame ;
- bool m_bAssemblingFrame ;
- bool m_bPrevFrameEnd ;
- BYTE *m_pStart ;
- BYTE *m_pEnd ;
- DWORD m_dwSize ;
- WORD m_wSeq ;
- BYTE m_H264PAYLOADTYPE ;
- DWORD m_ssrc ;
- };
- // class CH264_RTP_UNPACK end
- //////////////////////////////////////////////////////////////////////////////////////////
- 猜你在找
- 准备好了么? 更多职位尽在
- 数据工程师(外包) 工银瑞信基金管理有限公司 9-18K/月 我要跳槽
- Android视频工程师 国广星空视频科技(北京)有限公司 15-20K/月 我要跳槽
- H5高级攻城狮 五彩世界(北京)文化传媒有限公司 10-15K/月 我要跳槽
- H5前端开发工程师 猎聘网 10-15K/月 我要跳槽
- 23楼 _beginthreadex 2014-11-07 16:31发表 [回复]
- 调用博主的代码成功实现了对RTP包的解析和帧重组,直接保存成文件可以用VLC播放。另,博主的解包类里面似乎没有对SEI进行处理,所以在转换某个摄像机的视频时,我得到的数据全都是SPS与PPS,后查看直接保存的二进制文件,发现因为SEI也是有序列号的,不加1的话,下一次比较定会出错。已手动修改。
- 22楼 对牛乱弹琴 2014-07-09 09:40发表 [回复]
- 你好,非常感谢你的代码,我调试可以正常解包的。
但是我现在有个问题,UDP传输,我本机发送本机接收,总是断断续续的丢包,
if ( m_RTP_Header.seq != (WORD)( m_wSeq + 1 ) )//lost packet
会进入到这个判断里面?百思不得解,希望博主提示一下哈。
谢谢
- Re: dengzikun 2014-07-09 23:20发表 [回复]
- 回复chen495810242:本机收发丢包,可能是收发速率不匹配,可以增大 socket缓冲区试试。试试把接收数据后的处理逻辑去掉,只接收数据。
- Re: 对牛乱弹琴 2014-07-10 08:53发表 [回复]
- 回复dengzikun:额,我必须表示感谢,增加接收缓冲区大小就可以了,⊙╊⊙b汗。
在麻烦一下,你还有其他组包模式的解析方式吗?
FU-B,STAP-A,STAP-B等,还有你说的错误处理省略了,能否也加上,我觉得这个其实更重要,非常感谢
- Re: dengzikun 2014-07-11 14:26发表 [回复]
- 回复chen495810242:"省略错误处理"是指博文中的类使用示例代码。这两个类是可以放心使用的。H264 RTP包的其他解析方式你可以参考live555等开源库。
- Re: dengzikun 2014-07-16 11:32发表 [回复]
- 回复chen495810242:typedef struct {
#ifdef _BIG_ENDIAN
unsigned short v:2; /* packet type */
unsigned short p:1; /* padding flag */
unsigned short x:1; /* header extension flag */
unsigned short cc:4; /* CSRC count */
unsigned short m:1; /* marker bit */
unsigned short pt:7; /* payload type */
#else
unsigned short cc:4; /* CSRC count */
unsigned short x:1; /* header extension flag */
unsigned short p:1; /* padding flag */
unsigned short v:2; /* packet type */
unsigned short pt:7; /* payload type */
unsigned short m:1; /* marker bit */
#endif
uint16_t seq; /* sequence number */
uint32_t ts; /* timestamp */
uint32_t ssrc; /* synchronization source */
} rtp_hdr_t;
- 21楼 nanqingzhe 2014-06-09 17:33发表 [回复]
- 楼主,最近在做RTP包解析,发现用你这个解包不能成功啊。解包的时候,第一帧是I帧,SPS部分没有解析成功,然后每个包打包的多余字节没有去掉
- Re: nanqingzhe 2014-06-11 10:29发表 [回复]
- 回复dengzikun:打包解包均采用你给的代码。
一个完整的I帧塞进去打包,打包得到的数据交由解包的代码。
- Re: nanqingzhe 2014-06-11 10:27发表 [回复]
- 回复dengzikun:我用的就是你这个打包代码。将一个完整的I帧塞进去打包,打包后的数据交给你的这个解包代码。
- 19楼 chinapacs 2012-06-29 14:16发表 [回复]
- 非常感谢谢!!!基于时间戳这一块,困扰我大半个月。。目前圆满解决。我采用的方式跟你的类似,再次谢谢您。
memcpy(dst, nal[0].p_payload, i_frame_size);
if (count > 1) {
struct timeval now;
gettimeofday(&now, NULL);
double val = (now.tv_sec - firstTime.tv_sec) +
(now.tv_usec - firstTime.tv_usec) / 1000000.0;
ts_current= ts_current + val * 90000.0 + 0.5;
firstTime = now;
}
else {
ts_current= 0;
gettimeofday(&firstTime, NULL);
}
count ++;
- Re: dengzikun 2012-06-29 09:12发表 [回复] [引用] [举报]
- 回复chinapacs:class RTP_Timestamp
{
public:
RTP_Timestamp(DWORD unit)
: m_dwUnit(unit)
{
QueryPerformanceFrequency ( (LARGE_INTEGER*)&m_Freq ) ;
}
~RTP_Timestamp(void)
{
}
DWORD GetTime ()
{
__int64 current ;
QueryPerformanceCounter ( (LARGE_INTEGER*)¤t ) ;
DWORD ts = current * m_dwUnit / m_Freq ;
return ts ;
}
private:
DWORD m_dwUnit ;
__int64 m_Freq ;
};
RTP_Timestamp TS ( 90000 ) ;
每一帧调用TS.GetTime()获取时间戳。