Page 177 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

154 Chapter 6. Multiple-Reference motion Estimation Techniques

QSIF Table Tennis @ 10 f.p.s., QP = 5, 10, 15, 20, 25, 30
38

36

34
PSNR Y (dB) 32

30

28 SR
SR-RC
MR
MR-RC
26
10 20 30 40 50 60 70 80 90 100
Bit rate (kbits/s)
Figure 6.8: R-D performance of di&erent single- and multiple-reference (with M = 50) encoders
when encoding QSIF TABLE TENNIS at 10 frames=s

6.4 Discussion

Higher coding e ciency is one of the main requirements for mobile video
communication. One way to achieve higher coding e ciency is to use ad-
vanced motion estimation techniques. One of the promising advanced tech-
niques is multiple-reference motion-compensated prediction (MR-MCP).
This chapter reviewed the main e&orts in the )eld of MR-MCP. It then in-
vestigated the performance of the long-term memory motion-compensated pre-
diction (LTM-MCP) technique. It was found that this technique provides sig-
ni)cant prediction gains compared to the single-reference case. It was
realized, however, that such prediction gains are achieved at the expense of
an additional bit rate to transmit one extra temporal component per motion
vector. This additional bit rate has to be justi)ed in terms of an improved
rate-distortion (R-D) performance. An investigation into the R-D performance
of LTM-MCP codecs revealed that the use of rate-constrained motion esti-
mation and mode decision is important for the success of such techniques.
Without rate constraints, the R-D performance of the LTM-MCP technique