Page 266 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 10.4. Simulation Results 243

th
(a) Original 58 frame (b) Damaged blocks, 30%

(c) Concealed using BM (d) Concealed using BM-MFI

Figure 10.8: Subjective quality of concealed 58 th frame of QSIF TABLE TENNIS at 30 frames=s with
a blockloss rate of 30%

10.4.2 Results Within an H.263 Decoder
This set of simulations tests the performance of the techniques when incor-
porated within an H.263 decoder. In this case, the assumptions made in the
previous set of simulations will be relaxed. In other words, previously re-
constructed, possibly damaged and concealed, frames will be used for both
prediction and concealment. This will result in temporal error propagation. In
addition, spatial error propagation will also occur, since H.263 uses VLC and
predictive coding.
The Telenor implementation [144] of H.263 was used in this simulation.
The decoder was modi$ed to perform error detection by detecting syntax and
semantic violations, as was described in Section 9.5. When an error is detected,
the decoding process is stopped, the decoder searches for the next synchro-
nization codeword, and decoding is resumed. All macroblocks between the
point where the error was detected and the synchronization point are marked
as damaged macroblocks. In this simulation, the H.263 encoder option to in-
sert synchronization codewords at the start of each GOB was switched on. All