![](bg5.png)
the same simulation setup. For each sequence, all
frames are encoded with intra encoding under H.264
baseline profile, and QP=28. For the MB loss, we
use FMO (flexible macroblock ordering) tool with 2
slices per frame in encoding (Stephan Wenger, 2003),
while 1 slice is lost for channel simulation. In Fig.
14(b), green MBs compose a lost slice while all 4
NMBs are available for each LMB. To evaluate im-
age recovery, both subjective and objective image
qualities are observed.
Fig. 11 shows the comparison of objective image
quality measurement, luminance PSNR for con-
cealed frames using different methods. A larger
PSNR value leads to a better image recovery. By
observing Fig. 13, some conclusions can be ob-
tained:
I.
For different sequence, within DI, BI, and
BNM, different method wins the best recov-
ery, e.g., DI wins for foreman, BI wins for
bus, while BNM wins for flower. In other
words, in order to find a best image recovery,
an adaptive scheme is necessary.
II.
For adaptive method in (Dimitris, 2006),
compared with BI and DI, it dos not always
win the best recovery, such as bus, although
the difference with the winner is very small.
Same case was found in (Z Rongfu, 2004).
Compared with BI, DI, and BNM, method in
(Z Rongfu, 2004) does not always win the
best, such as foreman. However, for proposed
one, it always wins the best recovery.
Figure 11: Objective image recovery comparison.
Fig. 12 shows the subjective image quality com-
parison for flower sequence. Proposed method
achieves the best image recovery. It is easy to see
that, the optimal MB type for the flowers area should
be texture. However in algorithm of (Z Rongfu, 2004)
whose result is shown in Fig. 12(g), some LMBs in
flower area are regarded as SMB and recovered by
BI, which is a false MB type decision. In contrast,
our proposal avoided such false decision wel.
4 CONCLUSIONS
Considering there are 3 types of MBs in natural vid-
eo frame, i.e., TMB, EMB, and SMB, this paper
proposed an adaptive spatial error concealment,
which can choose 3 different methods for these 3
different MBs. For criterion of choosing, both edge
statistics measurement and intra prediction mode for
H.264 are taken into consideration. In terms of sub-
jective and objective image quality evaluation, expe-
riments show that the proposed method achieves the
best image recovery compared with previous work.
ACKNOWLEDGEMENTS
This work was supported by CREST, JST and Glob-
al-COE program of Japan.
REFERENCES
Jao-Won Suh, et al., 2002: Error concealment techniques
for digital TV. IEEE Trans. Broadcasting, Vol. 48
Yao.Wang, et al., 1998: Error control and concealment for
video communication: a review. Proceedings of IEEE,
pp947~997
Y.K.Wang, et al, 2002: The error concealment feature in
the H.26L test model. Proc. ICIP, vol.2, pp729~732
Jae-Won Suh, et al., 1997: Error concealment based on
directional interpolation. IEEE Trans. on consumer
electronics
Yan Zhao, et al, 2005: Spatial error concealment based on
directional decision and intra prediction. ISCAS
Dimitris, et al.: Enhanced error concealment with mode
selection. IEEE Trans. on circuits & sys for video tech
(2006)
Zhou Wang, et al., 1998: Best Neighborhood Matching:
An information loss based image coding systems.
IEEE Trans. on image processing
Z Rongfu, et al. 2004: Content-adaptive spatial error con-
cealment for video communication. IEEE Trans. on
consumer electronics, vol. 50, No.1
Thomas Wiegand, et al., 2003: Overview of the
H.264/AVC video coding standard. IEEE Trans. on
circuits and system for video technology
Stephan Wenger, 2003: H.264/AVC over IP. IEEE Trans.
on circuits and system for video technology
APPENDIX
In order to show the fact that 16x16 mode is more
suited for coding very smooth area, i.e., SMB, while
4x4 is well suited for area with significant detail, i.e.,
TMB, we did experiments to observe this.
AN ADAPTIVE SPATIAL ERROR CONCEALMENT FOR H.264/AVC VIDEO STREAM
27