field/frame DCT & intra/inter coding?

[dattrax]

All,

Just a few questions, regarding what xvid does in certain situations.

Does xvid ever use intra coded blocks within a predicted frame? Presumably occasionally the coding of the prediction error and motion vector will be larger than just intra coding the block. Are the algorithms in place to make these kinds of decisions.

Does xvid use field/frame DCT at a macroblock level?
At first glance this may look like an interlaced/progressive issue, but it is not. It can impact the amount of blocks required to be coded within a macroblock. It can also effect the complexity of the DCT's and reduce the number of codes required for VLC.

Just some thoughts to make xvid better.

Jim

[sysKin]

Originally posted by dattrax
Does xvid ever use intra coded blocks within a predicted frame? Presumably occasionally the coding of the prediction error and motion vector will be larger than just intra coding the block. Are the algorithms in place to make these kinds of decisions.

Yes it does - exactly in the situation you say. There is no "algorithm", just a simple comparsion between prediction error versus complexity of the macroblock. Look for "inter/intra" decison part of MotionEstimation() function, in motion_est.c

Does xvid use field/frame DCT at a macroblock level?
At first glance this may look like an interlaced/progressive issue, but it is not. It can impact the amount of blocks required to be coded within a macroblock. It can also effect the complexity of the DCT's and reduce the number of codes required for VLC.
Yes it does - there is a function "MBDecideFieldDCT()" in mbtransquant.c which decides if the macroblock should be coded as interlaced or progressive.

Radek