[MPEG] reference picture question

Hi all,

I am reading MPEG's MSSG software implementation.

fullsearch() function in motion.c input includes "source reference
picture" and "reconstructed reference picture".

AFAIK, all the reference picture(including I frame and P frame) should
be reconstructed (by IDCT) before be used as reference. So what's these
two reference's difference?

Any suggestions will be appreciated!
All the best!
Davy

Motion Estimation is done between Current and Reference pictures
Reconstructed Reference is actually the reference picture which may be past frame in case of IPPP sequences and may be past or future frame in case of IBBP sequences..........
I think source reference is actually the current frame.

Best Regards

Hi,
MPEG uses forward pediction and Bi-Prediction to encode picture.
I - No prediction and involves only 8x8 DCT
P - Uses the encoded I frame as its reference frame.( the succesive P i.e in seq. like IP1P2P3...P1 refers I, P2 refers P2 and so on..)
B - requires two reference frames for it - the forward and the backward reference picture i.e in a seq. like I1B1B2P1B3B4P2....B1and B2 requires both I1 and P1...B3 and B4 requires both P1 and P2 as reference..and so on...
This way the maximum number of reference frames for encoding a picture is two and they need to be stored. One must be the forward reconstructed frame and the other backward reconstructed frame for Bi-prediction.
Since while encoding P pictures and since they need to be stored for further reference they must have named it as current reference frame.

Hope this was useful,
KJN