h.264 SODB RBSP EBSP的区别

SODB　数据比特串－－＞最原始的编码数据

RBSP　原始字节序列载荷－－＞在SODB的后面填加了结尾比特（RBSP trailing bits　一个bit“1”）若干比特“0”,以便字节对齐。

EBSP　扩展字节序列载荷-->在RBSP基础上填加了仿校验字节（0X03）它的原因是：　在NALU加到Annexb上时，需要填加每组NALU之前的开始码StartCodePrefix,如果该NALU对应的slice为一帧的开始则用4位字节表示，ox00000001,否则用3位字节表示ox000001.为了使NALU主体中不包括与开始码相冲突的，在编码时，每遇到两个字节连续为0，就插入一个字节的0x03。解码时将0x03去掉。也称为脱壳操作。

 

网上查询的区别：

在对整帧图像的数据比特串(SODB)添加原始字节序列载荷(RBSP)结尾比特(RBSP trailing bits,添加一比特的“1”和若干比特“0”,以便字节对齐)后,再检查RBSP 中是否存在连续的三字节“00000000 00000000 000000xx”;若存在这种连续的三字节码,在第三字节前插入一字节的“0×03”,以免与起始码竞争,形成EBSP码流,这需要将近两倍的整帧图像码流大小。为了减小存储器需求,在每个宏块编码结束后即检查该宏块SODB中的起始码竞争问题,并保留SODB最后两字节的零字节个数,以便与下一宏块的SODB的开始字节形成连续的起始码竞争检测;对一帧图像的最后一个宏块,先添加结尾停止比特,再检测起始码竞争。

 

程序：

 

typedef struct

{

  int             byte_pos;           //!< current position in bitstream;

  int             bits_to_go;         //!< current bitcounter

  byte            byte_buf;           //!< current buffer for last written byte

  int             stored_byte_pos;    //!< storage for position in bitstream;

  int             stored_bits_to_go;  //!< storage for bitcounter

  byte            stored_byte_buf;    //!< storage for buffer of last written byte

 

  byte            byte_buf_skip;      //!< current buffer for last written byte

  int             byte_pos_skip;      //!< storage for position in bitstream;

  int             bits_to_go_skip;    //!< storage for bitcounter

 

  byte            *streamBuffer;      //!< actual buffer for written bytes

  int             write_flag;         //!< Bitstream contains data and needs to be written

 

} Bitstream;　定义比特流结构

 

 

static byte *NAL_Payload_buffer;

void SODBtoRBSP(Bitstream *currStream)
{
  currStream->byte_buf <<= 1;　　//左移1bit
  currStream->byte_buf |= 1;　　　//在尾部填一个“1”占1bit
  currStream->bits_to_go--;
  currStream->byte_buf <<= currStream->bits_to_go;
  currStream->streamBuffer[currStream->byte_pos++] = currStream->byte_buf;
  currStream->bits_to_go = 8;
  currStream->byte_buf = 0;
}

 

 

int RBSPtoEBSP(byte *streamBuffer, int begin_bytepos, int end_bytepos, int min_num_bytes)
{
 
  int i, j, count;

  for(i = begin_bytepos; i < end_bytepos; i++)
    NAL_Payload_buffer[i] = streamBuffer[i];

  count = 0;
  j = begin_bytepos;
  for(i = begin_bytepos; i < end_bytepos; i++)
  {
    if(count == ZEROBYTES_SHORTSTARTCODE && !(NAL_Payload_buffer[i] & 0xFC))
    {
      streamBuffer[j] = 0x03;
      j++;
      count = 0;  
    }
    streamBuffer[j] = NAL_Payload_buffer[i];
    if(NAL_Payload_buffer[i] == 0x00)     
      count++;
    else
      count = 0;
    j++;
  }
  while (j < begin_bytepos+min_num_bytes) {
    streamBuffer[j] = 0x00; // cabac stuffing word
    streamBuffer[j+1] = 0x00;
    streamBuffer[j+2] = 0x03;
    j += 3;
    stat->bit_use_stuffingBits[img->type]+=16;
  }
  return j;
}

 

本文来自CSDN博客，转载请标明出处：http://blog.csdn.net/threewells_14/archive/2007/02/12/1508657.aspx