"Davy" <zhushenli@gmail.com> a &#4294967295;crit dans le message de news: 1126599875.788898.289810@g44g2000cwa.googlegroups.com...
| Hi all,
|
| I want to implement a encoder deal with I and P frame.
| So only one reference frame is stored for motion search.
| Is it the reconstructed forward frame?

Yes (In fact it may be)

|
| But the MSSG (http://www.mpeg.org/MPEG/MSSG/) encoder's full search
| function require 2 frames.
|
| Its integer full search is based on the "original source reference".
| And its half pel search (9 point related to the best integer) is based
| on the "reconstructed reference".

They are using source picture to search the integer motion vector and
they are using the reconstructed picture to refine it to 1/2 pixel. That's
their choice. Mpeg standard doesnt say anything about motion estimation.


| Do you think "original source reference" can be replaced with
| "reconstructed reference"?

yes, you are free to do motion estimation as you want.

But motion compensation, so the process to get the predictor macrobloc
according to the vector found SHALL use the reconstructed picture.

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

Hi all,

I want to implement a encoder deal with I and P frame.
So only one reference frame is stored for motion search.
Is it the reconstructed forward frame?

But the MSSG (http://www.mpeg.org/MPEG/MSSG/) encoder's full search
function require 2 frames.

Its integer full search is based on the "original source reference".
And its half pel search (9 point related to the best integer) is based
on the "reconstructed reference".

Do you think "original source reference" can be replaced with
"reconstructed reference"?

Any suggestions will be appreciated!

Best regards,
Davy