Method and apparatus for estimating channel induced distortion
View Patent ↗There are provided methods and apparatus for estimating channel induced distortion. A method for estimating distortion for an inter-coded image includes the step of calculating an average channel distortion for received inter-coded macroblocks in the inter-coded image as a previous average channel distortion multiplied by a scaling factor.
1. A method in a distortion calculator computer for estimating distortion for an inter-coded image, comprising:
accessing with the distortion calculator computer an average channel distortion estimated for a previously coded image; and
calculating with the distortion calculator computer an average channel distortion for the inter-coded image as said average channel distortion estimated for the previously coded image scaled by a scaling factor (1-P)(a(1−β)+β)+hP, where P represents a packet loss rate, β represents an intra rate, a is a parameter associated with inter-prediction, c is a parameter associated with intra-prediction, and h is a parameter associated with concealment,
where:
the inter-coded image is encoded with inter-prediction and temporal concealment with non-integer motion vectors, and without intra-prediction and deblocking filtering,
the inter-coded image is encoded with inter-prediction and temporal concealment with non-integer motion vectors, with intra-prediction, and without deblocking filtering, or
the inter-coded image is encoded with inter-prediction and temporal concealment with non-integer motion vectors, and with intra-prediction and deblocking filtering.
2. The method according to claim 1 , wherein the parameter c is set to zero when the inter-coded image is encoded without intra-prediction or encoded with constrained intra-prediction.
3. The method according to claim 1 , wherein the parameter a is calculated as follows
a
=
w
E
{
∑
l
=
1
L
p
,
p
a
l
2
+
ρ
∑
l
,
r
,
l
≠
r
a
l
a
r
}
,
wherein a l and a r represent interpolation filter coefficients used for motion-compensated temporal prediction, L p,p represents a number of neighboring samples used for interpolation, ρ represents an average correlation coefficient between channel-induced errors in two adjacent pixels, and E{.} represents an averaging operation that averages values inside {.} for inter-coded pixels, and
wherein when the inter-coded image is encoded without deblocking filtering w is set to one, and when the inter-coded image is encoded with deblocking filtering w is determined as follows
w
=
E
{
∑
l
=
1
L
d
w
l
2
+
ρ
∑
l
,
r
,
l
≠
r
w
l
w
r
}
wherein w l and w r represent deblocking filter coefficients, and L d represents a number of neighboring samples used for the deblocking filtering.
4. The method according to claim 1 , wherein
the parameter c is calculated as follows
c
=
c
P
1
-
c
I
a
,
wherein C P and c I are respectively determined as follows
c
P
=
E
{
∑
l
:
P
-
neighbors
c
l
2
+
ρ
∑
l
,
r
:
P
-
neighbors
,
l
≠
r
c
l
c
r
}
,
c
I
=
E
{
∑
l
:
I
-
neighbors
c
l
2
+
ρ
∑
l
,
r
:
I
-
neighbors
,
l
≠
r
c
l
c
r
}
wherein c l and c r represent weighting coefficients used for intra-prediction, I-neighbors represents a set of all neighboring samples used for the intra-prediction that were previously coded using intra-mode, and P-neighbors represents a set of all neighboring samples used for the intra-prediction that were previously coded using inter-mode, and E{.} represents an averaging operation that averages values inside {.} for intra-coded pixels in P-frames.
5. The method according to claim 1 , wherein the parameter h is calculated as follows
h
=
w
E
{
∑
l
=
1
L
p
,
p
h
l
2
+
ρ
∑
l
,
r
,
l
≠
r
h
l
h
r
}
,
wherein h l and h r represent interpolation filter coefficients used for motion-compensated temporal error concealment L p,p represents a number of neighboring samples used for interpolation, ρ represents an average correlation coefficient between channel-induced errors in two adjacent pixels, and E{.} represents an averaging operation that averages values inside {.} for concealed pixels, and
wherein when deblocking filtering is not applied after motion-compensated temporal error concealment w is set to one, and when deblocking filtering is applied after motion-compensated temporal error concealment w is calculated as follows:
w
=
E
{
∑
l
=
1
L
d
w
l
2
+
ρ
∑
l
,
r
,
l
≠
r
w
l
w
r
}
;
wherein w l and w r are deblocking filter coefficients, and L d represents a number of neighboring samples used for the deblocking filtering.
6. The method according to claim 1 , further comprising determining an error concealment distortion for the inter-coded image, including:
defining a concealment distortion for a given macroblock in the inter-coded image as a channel distortion for the given macroblock when there is an absence of transmission losses in other frames and other macroblocks in the inter-coded image;
estimating the concealment distortion for a selected macroblock in the inter-coded image by setting the selected macroblock as lost and concealing the selected macroblock using a decoder error concealment;
repeating said estimating step for other selected macroblocks in the inter-coded image, and
calculating the error concealment distortion based upon the estimated concealment distortions for the selected macroblocks in the inter-coded image.
7. An apparatus for estimating distortion for an inter-coded image, comprising:
a distortion calculator computer for accessing an average channel distortion estimated for a previously coded image and for calculating an average channel distortion for the inter-coded image as said average channel distortion estimated for the previously coded image scaled by a scaling factor (1−P)(a(1−β)+cβ)+hP, where P represents a packet loss rate, β represents an intra rate, a is a parameter associated with inter-prediction, c is a parameter associated with intra-prediction, and h is a parameter associated with concealment,
where:
the inter-coded image is encoded with inter-prediction and temporal concealment with non-integer motion vectors, and without intra-prediction and deblocking filtering,
the inter-coded image is encoded with inter-prediction and temporal concealment with non-integer motion vectors, with intra-prediction, and without deblocking filtering, or
the inter-coded image is encoded with inter-prediction and temporal concealment with non-integer motion vectors, and with intra-prediction and deblocking filtering.
8. The apparatus according to claim 7 , wherein said distortion calculator computer sets the parameter c to zero when the inter-coded image is encoded without intra-prediction or encoded with constrained intra-prediction.
9. The apparatus according to claim 7 , wherein said distortion calculator computer calculates the parameter a as follows
a
=
w
E
{
∑
l
=
1
L
p
,
p
a
l
2
+
ρ
∑
l
,
r
,
l
≠
r
a
l
a
r
}
,
wherein a l and a r represent interpolation filter coefficients used for motion-compensated temporal prediction, L p,p represents a number of neighboring samples used for interpolation, ρ represents an average correlation coefficient between channel-induced errors in two adjacent pixels, and E{.} represents an averaging operation that averages values inside {.} for inter-coded pixels, and
wherein when the inter-coded image is encoded without deblocking filtering w is set to one, and when the inter-coded image is encoded with deblocking filtering w is determined as follows
w
=
E
{
∑
l
=
1
L
d
w
l
2
+
ρ
∑
l
,
r
,
l
≠
r
w
l
w
r
}
wherein w l and w r represent deblocking filter coefficients, and L d represents a number of neighboring samples used for the deblocking filtering.
10. The apparatus according to claim 7 , wherein said distortion calculator computer calculates the parameter c as follows
c
=
c
P
1
-
c
I
a
,
wherein c P and c I are respectively determined as follows
c
P
=
E
{
∑
l
:
P
-
neighbors
c
l
2
+
ρ
∑
l
,
r
:
P
-
neighbors
,
l
≠
r
c
l
c
r
}
,
c
I
=
E
{
∑
l
:
I
-
neighbors
c
l
2
+
ρ
∑
l
,
r
:
I
-
neighbors
,
l
≠
r
c
l
c
r
}
wherein c l and c r represent weighting coefficients used for intra-prediction, I-neighbors represents a set of all neighboring samples used for the intra-prediction that were previously coded using intra-mode, and P-neighbors represents a set of all neighboring samples used for the intra-prediction that were previously coded using inter-mode, and E{.} represents an averaging operation that averages values inside {.} for intra-coded pixels in P-frames.
11. The apparatus according to claim 7 , wherein said distortion calculator computer calculates the parameter h as follows
h
=
w
E
{
∑
l
=
1
L
p
,
p
h
l
2
+
ρ
∑
l
,
r
,
l
≠
r
h
l
h
r
}
,
wherein h l and h r represent interpolation filter coefficients used for motion-compensated temporal error concealment, L p,p represents a number of neighboring samples used for interpolation, ρ represents an average correlation coefficient between channel-induced errors in two adjacent pixels, and E{.} represents an averaging operation that averages values inside {.} for concealed pixels, and
wherein when deblocking filtering is not applied after motion-compensated temporal error concealment w is set to one, and when deblocking filtering is applied after motion-compensated temporal error concealment w is calculated as follows:
w
=
E
{
∑
l
=
1
L
d
w
l
2
+
ρ
∑
l
,
r
,
l
≠
r
w
l
w
r
}
;
wherein w l and w r are deblocking filter coefficients, and L d represents a number of neighboring samples used for the deblocking filtering.
12. The apparatus according to claim 7 further comprising a distortion estimator computer for
estimating a concealment distortion for a selected macroblock in the inter-coded image by setting the selected macroblock as lost and concealing the selected macroblock using a decoder error concealment,
repeating the estimating for other selected macroblocks in the inter-coded image, and
calculating an error concealment distortion based upon the estimated concealment distortions for the selected macroblocks in the inter-coded image.