In section 6.1, a coding structure with different
picture modes is introduced. Within each picture, macroblocks
may be coded in several ways, thus aiming at high coding efficiency.
The MB modes for intra, predicted and interpolated pictures are
shown in 6.2 to 6.4.
Pictures are coded in several modes as a trade-off between coding efficiency and random accessibility. There are basically three picture coding modes, or picture types:
- I-pictures: intra coded pictures.
- P-pictures: forward motion-compensated prediction pictures.
- B-pictures: motion compensated interpolation pictures.
Although, in principle, freedom could be allowed
for choosing one of these methods for a certain picture, for the
Test model a fixed, periodic structure is used depending on the
respective picture.
Every N-th picture of a sequence starting with the
first picture is coded as intra picture i.e. pictures 1, N+1,
etc. (see Test Model Fig. 10.1). Following every M-th picture in between (within
a Group of Pictures) is a predicted picture coded relative to
the previous predicted or intra picture. The interpolated pictures
are coded with reference to the two closest previous and next
predicted or intra pictures. In this TM, M=3 and N=15 for 29.97
Hz and M=3 and N=12 for 25 Hz.
The following parameters are currently to be used
for most of the core experiments:
|
|
|
Coding modes available for predicted and interpolated pictures
are described in detail in the following paragraphs.
Use the MSE critium to select the best Macroblock
mode.
In order to take advantage of the Special prediction modes, the decision rule must be modified for Field-based prediction.
It has been noted that quality is improved by choosing Field-based prediction less often, to the benefit of the Special prediction mode, particularly in B-Pictures.
For example, even in cases where Field-based prediction has an MSE slightly better than any of the other prediction modes, it may cost a significant overhead to transmit two field-vectors (four in B-Frames).
Until further improvement, we propose to use the following decision rule in core experiments involving the Special prediction modes :
- Field-based prediction chosen
in B-pictures: if (MSE_field + 8 < MSE_best_of_other_modes)
in P-Pictures: if (MSE_field < MSE_best_of_other_modes)
where MSE = Mean Square Error PER PEL of predicted
MB