.M2V File Format – Start codes

MPEG-2 video elementary stream

An MPEG video elementary stream is a series of headers and data, each started by the three bytes 00 00 01 and one code byte. Start codes begin on a byte boundary. The standard is ITU-T H.262, also ISO/IEC 13818-2; MPEG-1 video, ISO/IEC 11172-2, uses the same codes. The first edition of H.262 is dated July 1995, with later editions in 2000 and 2012.

Code values

CodeNameMeaning
00picture_start_codea picture header
01–AFslice_start_codea slice; the code gives the slice’s row
B0, B1reserved
B2user_data_start_codeuser data, running to the next start code
B3sequence_header_codea sequence header
B4sequence_error_codemarks data lost in transmission
B5extension_start_codean extension
B6reserved
B7sequence_end_codethe end of a sequence
B8group_start_codea group of pictures header
B9–FFsystem start codesused by the MPEG program and transport layers, not inside a video elementary stream

The system codes include B9 program end, BA pack header, BB system header, BC program stream map, BD and BF private streams, BE padding, C0–DF audio streams and E0–EF video streams. An .m2v file has none of these, and a file that starts with 00 00 01 BA is a program stream.

Order of headers

A coded sequence in the MPEG-2 syntax goes:

  1. Sequence header (B3), and straight after it a sequence extension (B5). Optional sequence display and sequence scalable extensions and user data may follow.
  2. Optionally a group of pictures header (B8) and its user data.
  3. For each picture: a picture header (00), a picture coding extension (B5), any further picture extensions and user data, then the slices (01–AF).
  4. More group headers and pictures. The sequence header may be repeated.
  5. A sequence end (B7).

In a stream, pictures come in coded order, which differs from display order when B pictures are used. The temporal_reference of each picture gives its display position within its group.

MPEG-1 streams

A file is MPEG-2 if a sequence extension (B5 with identifier 1) directly follows the first sequence header. Otherwise it is MPEG-1.

  • MPEG-1 streams have no B5 extensions at all. They are always progressive 4:2:0.
  • The picture header’s motion vector f_code fields are used in MPEG-1; MPEG-2 sets them to 7 and carries the real values in the picture coding extension.
  • The D picture type (4), DC coefficients only, exists only in MPEG-1.
  • The aspect_ratio_information field means something different: see the sequence header.
  • In MPEG-1 the slice code is the row number with no extension, and the constrained_parameters_flag marks streams of limited size that every decoder must play.

Finding the start codes

Scan the bytes for 00 00 01 and read the next byte. The syntax is designed so that the pattern cannot occur inside the headers and data; unlike H.264 there are no emulation prevention bytes.