Page 260 - Video Coding for Mobile Communications Efficiency, Complexity, and Resilience

Section 10.4. Simulation Results 237

Thus, the concealed pel is a weighted sum of two predictions. The function
w (·; ·) is used to adjust the weights given to MFI and BM according to the
spatial location of the pel within the damaged block. Knowledge of the way
both BM and MFI workcan provide some insights into designing a suitable
w (·; ·). For example, the SMD measure of the BM technique involves the
border pels of the damaged block. It is expected, therefore, that BM will
perform well at those pels. Therefore, BM must be given high weights at
the borders of the blockand low weights at the center. To achieve this, the
following function was used:
g (x n )g (y n )+1
w (x n ;y n )= ; (10.9)
2
where

 1 − k ( (4a−1))−k( ) ; 0 ≤ a ≤ 1
g (a)= k(− )−k( ) 2 (10.10)
 g (1 − a); 1 2 ¡a ≤ 1

and k (·) is de$ned by Equation (10.4). The parameter is used to control the
smoothness of w (·; ·), as illustrated in Figure 10.2.
Before proceeding to present simulation results, it is valuable at this point
to highlight the main di,erences between the two novel algorithms, MFI and
BM-MFI, and conventional temporal error concealment techniques. These are
summarized in Table 10.2.

10.4 Simulation Results

10.4.1 Results Within an Isolated Error Environment

It is very important to evaluate the performance of the techniques in isolation
from any external e,ects, like temporal and spatial error propagation and the
choice of the error detection algorithm. This is particularly important for a fair
comparison, since such error mechanisms and algorithm choices may randomly
a,ect one technique more than another. Thus, in this set of simulations, the
following assumptions were made
1. There is no temporal error propagation. This was achieved by using
original reference frames for the concealment process.
2. There is no spatial error propagation. This is equivalent to using $xed-
length codes and no predictive coding.

3. The concealment process is supported by an ideal error detection algo-
rithm that can identify all damaged blocks.