Page 175 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

152 Chapter 6. Multiple-Reference motion Estimation Techniques

Table 6.1: VLC codewords for encoding the temporal displacement d t . Reproduced from Ref. 140
d t Bits Codeword
0 1 1
“x 0 ” + 1 (1:2) 3 0x 0 0
“x 1 x 0 ” + 3 (3:6) 5 0x 1 1x 0 0
“x 2 x 1 x 0 ” + 7 (7:14) 7 0x 2 1x 1 1x 0 0
“x 3 x 2 x 1 x 0 ” + 15 (15:30) 9 0x 3 1x 2 1x 1 1x 0 0
“x 4 x 3 x 2 x 1 x 0 ” + 31 (31:62) 11 0x 4 1x 3 1x 2 1x 1 1x 0 0

QSIF Akiyo @ 10 f.p.s., QP = 5, 10, 15, 20, 25, 30
40

38

36
PSNR Y (dB) 34

32

30
SR
SR-RC
MR
MR-RC
28
0 5 10 15 20 25 30
Bit rate (kbits/s)
Figure 6.6: R-D performance of di&erent single- and multiple-reference (with M = 50) encoders
when encoding QSIF AKIYO at 10 frames=s

multiple-reference encoders bene)t from the use of rate-constrained motion es-
timation and mode decision. Those bene)ts are more evident in high-movement
sequences, where the use of more bits to encode the longer motion vec-
tors has to be justi)ed and controlled. It should be pointed out, however,
that such bene)ts are achieved at the expense of increased computational
complexity.