Page 274 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 10.5. Temporal Error Concealment for Multiple-Reference 251

Spatial-components recovery, Foreman, M=10, QP=10, Skip=3 Spatial-components recovery, Foreman, M=10, QP=10, Macroblock error rate=20%
25 24
ZR-ZR ZR-ZR
AV-ZR AV-ZR
24 BM-ZR 23.5 BM-ZR
MFI-ZR MFI-ZR
23
23
22
22.5
PSNR Y (dB) 21 PSNR Y (dB) 22
20
21.5
19
21
18
17 20.5
16 20
10 20 30 40 50 1 2 3
4
Macroblock error rate (%) Frame skip
(a) Performance over a range of error rates (b) Performance over a range of frame skips
Figure 10.15: Spatial-components recovery for QSIF FOREMAN with M =10 and QP =10
Spatial-components recovery, Table Tennis, M=10, QP=10, Skip=3 Spatial-components recovery, Table Tennis, M=10, QP=10, Macroblock error rate=20%
25 23.5
ZR-ZR ZR-ZR
AV-ZR AV-ZR
24 BM-ZR BM-ZR
MFI-ZR 23 MFI-ZR
23
22.5
22
PSNR Y (dB) 21 PSNR Y (dB) 22
20
21.5
19
21
18
17 20.5
10 20 30 40 50 1 2 3 4
Macroblock error rate (%) Frame skip
(a) Performance over a range of error rates (b) Performance over a range of frame skips
Figure 10.16: Spatial-components recovery for QSIF TABLE TENNIS with M =10 and QP =10

recovery process. A very interesting point to note is that the performance of
MFI starts to deteriorate at high frame skips. This may be due to the fact that
at high frame skips, the spatial components within the motion $eld become
less correlated (see Property 6:3:1:3 and Figures 6.4(a) and 6.4(b)). Since MFI
assumes a high correlation between the spatial components, its performance
will deteriorate with decreased correlation.

10.5.3 Spatial-Temporal-Components Recovery

Comparing the results of Section 10.5.1 to those of Section 10.5.2 it can be
concluded that spatial-components recovery is, in general, more important than