MPEG reference picture order?

Hi all,
I have 7 frame, their display order is I1 B2 B3 P4 B5 B6 P7.

What's B3's reference pictures?
(I1 and P4) or (B2 and P4)?

What's P7's reference pictures?
(I1) or (P4)?

Does the picture use the nearest encoded picture as the reference
picture?
Shall I store two nearest encoded picture?

Any suggestions will be appreciated!
Best regards,
Davy

"Davy" <zhushenli@gmail.com> a &#4294967295;crit dans le message de news: 1125973809.994256.192550@g14g2000cwa.googlegroups.com...
| Hi all,
| I have 7 frame, their display order is I1 B2 B3 P4 B5 B6 P7.
|
| What's B3's reference pictures?
| (I1 and P4) or (B2 and P4)?

I1 and P4 for mpeg 1 and 2

It seems that mpeg4 standard allow B2 too. To be confirmed

|
| What's P7's reference pictures?
| (I1) or (P4)?

P4 only

|
| Does the picture use the nearest encoded picture as the reference
| picture?
| Shall I store two nearest encoded picture?

|
| Any suggestions will be appreciated!
| Best regards,
| Davy
| 

Hi ast,

Happy to talk with you :-)

I am confused with the encoding loop.

The I1 and P4 stored for future reference usage is original frame or
reconstructed frame??

Best regards,
Davy

"Davy" <zhushenli@gmail.com> a &#4294967295;crit dans le message de news: 1125996590.427371.130170@g44g2000cwa.googlegroups.com...
| Hi ast,
|
| Happy to talk with you :-)
|
| I am confused with the encoding loop.
|
| The I1 and P4 stored for future reference usage is original frame or
| reconstructed frame??
|

Reconstructed frame, because the coder and the decoder shall
reconstruct the picture in exactly same way, and of course the
decoder only has the reconstructed picture.

But, the coder is free to search motion vector as it wants, so sometime
the coder use source picture for motion vector search. This was one
of your previous question. In that case, the encoder stores both source and
reconstructed picture.

Motion compensation, which is calculating the difference between a
source marcobloc and the prediction macrobloc (which depending on the vector)
is ALWAYS done with the prediction macrobloc taken in a reconstructed picture.

| Best regards,
| Davy