Sequ: IPPPBBBBPPBBBBPPBBBBPPBBBB

[gm901]

Hello,

Trying to understand a few things…
I have a VOBU with this IBP sequence: [Go]IPPPBBBBPPBBBBPPBBBBPPBBBB, 26frames,[Go]=GOP open flag
According to the times (BCD Elapsed + PTMs) it should be only 13 frames.
The temporal sequence numbers are 003311226644559………

Any help would be very welcome as to understand this

Thanks

[mpucoder]

You have 26 pictures, not frames. Mpeg-2 encodes video as pictures, which may represent a field or a frame. It takes 2 field pictures, which will have the same TSN, to make one frame.

[gm901]

Thank you for the response.

So the safe way to distinguish field from frame is the Picture_structure value of Picture Coding Extension.
Therefore:
The TSNs should be equal (if not is error).
The fields Must be concecutive (same or next coming packs)
The calculation for VOBU 1st, 2nd and 3rd References will take into account that this 2 fields is a frame, so the end of the 2nd field is the end of the frame (or the start of the 1st the start of the frame).

Can I use the above assumptions as rules ??

As far as progressive and interlaced video is concerned

Progressive video ==> picture = frame
Interlaced video ==> picture = frame OR 2 fields = Frame
so
2 fields Frame ==> interlace (rule)

and

Sequence extension’s
progressive_sequence =1 then video is only progressive
progressive_sequence =0 then video is either progressive or interlaced (framepicture or field-picture or both)

therefore as rules

when progressive_sequence =1 then there is only progressive video
and
when there are 2 fields there is only interlaced.

so
progressive_sequence =1 AND 2 fields ==> error

Is that true???

Thanks

[mpucoder]

Your being too restrictive with your rules. All the allowable variations are described in ISO 13818-2. Basically in a non-progressive sequence anything is allowed in any combination, as the decoder sees fit. If a field picture is used it must have a mate for the other field in the same GOP, and the field order cannot change throughout the sequence. Frame pictures may be either progressive or non-progressive, and each macrocell within may use field or frame DCT - all in the same sequence, whatever the encoder wants to use.
You cannot determine if the source was interlaced or progressive by analyzing the encode, not even rff is a determining factor (it can, and is, used to pulldown other than film source)

Reference frames are a little odd, the first I picture of course, but then only the opposite fields. For example, if the I frame is a top field, then the 2 P bottom field pictures are references.