Page 169 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

146 Chapter 6. Multiple-Reference motion Estimation Techniques

QSIF Foreman
0.5
M=50, Skip=1
M=50, Skip=4
0.45
0.4

0.35
0.3
p(d t ) 0.25

0.2

0.15
0.1

0.05
0
0 5 10 15 20 25 30 35 40 45 50
d
t
Figure 6.3: Zero-biased distribution of the long-term memory temporal displacement d t

in memory) has the highest frequency of occurrence; and as the temporal
displacement increases, its frequency of occurrence decreases. Note that this
distribution becomes more spread at higher frame skips, which indicates that
the selection of older reference frames becomes slightly more probable.

Property 6.3.1.3 The long-term memory block-motion eld is smooth and
varies slowly. In other words, there is high correlation between the motion
vectors of adjacent blocks. This is evident from Figure 6.4, which shows the
distribution of the di&erence between the current vector C and its left neighbor
L. This is shown for the three components: d x (Figure 6.4(a)), d y (Figure
6.4(b)), and d t (Figure 6.4(c)). All three distributions are biased toward a
zero di&erence, which indicates high correlation. Note that this correlation is
slightly less in the multiple-reference case (M =50; skip = 1), compared to the
single-reference case (M =1; skip = 1). This correlation is further reduced at
higher frame skips, (M =50; skip = 4).

In general, it can be concluded that moving from a single-reference system
to a multiple-reference system does not signi)cantly change the properties of
the block-motion )eld.