CHAPTER 7 HOW LIGHT BECOMES DATA^117
Of course, few things in the real world
are consistently the same color. Vagaries
of light and shade cause a spot in the
background to be a slightly different
shade. MPEG makes a decision:
Will anyone notice if it makes that
errant shadow the same color as the
rest of the wall? If the decision is
“no,” that’s a few thousand more
bits saved.(^3) One compression technique assumes,
rightly, that much video changes little
from one frame to another. To save space
recording the video—and broadcasting it,
for that matter—the technique combines
part of one frame with part of the next one.
In the case where there is little change, the
economy works.
4
But in this case, for example,
where the scene changes
abruptly from one frame to the
other, the technique results in
a fleeting visual glitch.
The same premise is behind motion estimation, which assumes that
consecutive frames will be the same except for changes caused by
objects moving between frames. By examining the changes in two of the
frames, MPEG predicts the position of objects in a third frame.
(^5) Key frame compression records only
certain crucial frames and deduces
from them what the missing intervening
frames should look like.
6
A similar technique makes a
record of one frame that is
largely unchanged in subse-
quent frames. Then the only
data recorded is the delta—
the differences from that
original reference frame that
occur in the frames trailing it.
7