Image Processing Reference
referred to as field dominance. Sometimes these are referred to as the top and bottom
While the physical implementation of odd/even first/second fields is the same
for U.S. and European broadcasting, the signaling of the field dominance is not always
the same and a weird twittering effect is evident if the fields are played in the wrong
Frame and field prediction is modified in MPEG-2. Instead of 16
16 arrays of pix-
els, there are 16
8 arrays (because of the interlace effect). Prediction is only operated on
corresponding fields in another frame. The division of frames into pairs of fields contin-
ues on down into the macroblock and the DCT coding mechanisms, and all of this must
be re-examined in the context of a field-based model rather than a frame-based model. The
ordering of the DCT coefficients into a bit stream and the subsequent entropy coding is
also modified to favor the vertical components.
The internals of the compression codec work similarly whether they are processing
frames or fields. The complexity is about the same but the specifics of how the mac-
roblocks are organized for luma and chroma coding is slightly different. The overall tech-
niques are very similar to those already discussed for MPEG-1. You should consult the
topics on video-compression theory if you want to delve further into these more complex
issues. All of this will be hidden deep inside your encoding software. It is likely to be acti-
vated by a single check box that selects interlaced processing.
The color model is modified a little in MPEG-2 when compared with MPEG-1. Additional
color coding is supported. The 4:2:0, 4:2:2, and 4:4:4 sampling schemes are all available.
Profiles and levels must be understood in order to encode correctly, and those are dis-
cussed in a little while. Your encoding software should provide help with this.
Figure 11-1 shows the sampling windows for each of those sampling schemes:
When field-based coding is used, the color macroblocks are related to the upper one
of the two fields. Color information is averaged both spatially and temporally. Some
ghosting of the color may be evident in fast-moving objects having a saturated color.
A common example is red-shirted soccer players against a green pitch.
MPEG-2 introduces a restriction on the slice organization. MPEG-1 allows slices to extend
to several rows of macroblocks but MPEG-2 only allows those slices to extend to a full row.
Slices can be smaller but must not span a right edge and wrap around as they could with
MPEG-1. Figure 11-2 compares the two.
The benefit here is that dropouts in the signal cause much less damage and the
reconstruction process stands a better chance of hiding the dropout if the decoder is smart