Information Technology Reference
In-Depth Information
Taking the video packet as the base unit, the expected frame quality can be
written as:
⎛
N
∑
−
=
1
⎞
(
)
E
(
Q
)
=
10
⋅
log
⎜
⎝
g
/
E
D
⎟
⎠
(5)
l
,
m
10
l
,
m
,
n
n
0
where
E
(
Q
l,m
) is the expected quality of
m
th
frame of the
l
th
object,
E
(
D
l,m,n
) is the
expected distortion of
n
th
video packet and
N
is the total number of video packets.
g
is a constant defined by the color depth and by the size of given composite video
frame,
g =
c
2
h
, where
c
is the color depth factor,
w
and
h
are the width and
height of the video frame, respectively.
E
(
D
l,m,n
) can be written as:
(
⋅
w
⋅
) (
)
(
)
(
)
ED
=
ED
Q
+
ρ
e
lmn ED
,,
⋅
T
+
f lmn
tp
(,,)
(6)
,,
,,
,,
lmn
lmn
lmn
.
Calculation of each term shown in (6) depends on the implemented concealment
techniques, and the applied video coding scheme and its parameters. Quantization
distortion is computed from the squared difference of original and reconstructed
luminance values for every macroblock in the particular video packet.
The probability
e
(
l, m
,
n
) of receiving a video packet with errors depends on
the channel's Bit Error Rate (BER) and size of the video packet. If
ρ
b
is the BER
ρ
e
(
l, m
,
n
) is given as:
of the transmission channel then
ρ
(
)
(
)
p
l
,
m
,
n
ρ
e
l
,
m
,
n
=
1
−
(
−
ρ
b
)
(7)
b
where
p
b
represents the size of the packet in bits.
The extent of temporal concealment distortion depends on the algorithm used
for the temporal concealment. Here a simple concealment approach is followed,
where macroblocks in the corrupted video packet are replaced by data of the cor-
responding macroblocks of the previous frame. For the estimation it is assumed
that the neighbouring video packets and reference frames are received correctly.
Therefore, the concealment distortion is given as:
E
(
D
T
)
=
Y
−
Y
′
(8)
l
,
m
,
n
l
,
m
,
n
l
,
m
−
1
n
which is the squared difference of the luminance components of macroblocks of
reconstructed part of current frame
Y
m
,
n
and the same spatial area in the previous
frame
Y
'
m
- 1,
n
.
The temporal error propagation due to MB mismatch between adjacent video
frames is quantified by the term
f
tp
(
l, m
,
n
) in (3), which is computed as:
(
(
)
)
(
)
tp
e
tp
m
f
l
,
m
,
n
=
1
−
ρ
l
,
m
,
n
⋅
p
⋅
l
,
−
1
(9)
K
∑
−
=
1
[
]
(
)
(
)
(
)
(
)
ρ
e
l
,
m
−
1
n
⋅
E
D
T
+
1
−
ρ
e
l
,
m
−
1
n
⋅
p
tp
m
k
,
l
,
m
−
1
n
l
,
−
2
k
0
