Statistical adaptive video rate control
View Patent ↗A method and apparatus for video rate control for a video encoding system for encoding a video signal includes determining 12 a variance of pixel values over a time t of the video signal and calculating 13 a quantization parameter for the time t using the pixel variance and a number of output bits for the time t and a quantization parameter, pixel variance and a number of output bits used for an immediately previous time t−1.
1. A method of video rate control for a video encoding system for encoding a video signal comprising:
a) determining, using a processor, a variance of pixel values over a time t of the video signal; and
b) calculating, using a processor, a quantization parameter for the time t using the pixel variance and a number of output bits over the time t and a quantization parameter used for an immediately preceding time t−1, a pixel variance over the immediately preceding time t−1, and a number of output bits over the immediately preceding time t−1;
wherein the variance is a modified variance being a function of the calculated variance over the time t and a similarly modified variance over the immediately preceding time t−1, and
wherein the modified variance V(t) is defined as: V(t)=(1−α)V a (t)+αV(t−1) where V a (t) is the calculated variance over a time t, V(t−1) is the modified variance over the immediately preceding time t−1 and α is a constant such that 0<α<1.
2. A method as claimed in claim 1 , wherein the quantization parameter over the time t is given by:
QP ( t )=(1−β)Δ qp+βQP ( t− 1)
where:
Δ qp =(ε( V ( t )− V ( t− 1))+ R ( t− 1) V ( t− 1)− R ( t ) V ( t ))/η
where QP(t−1) is the quantization parameter over an immediately preceding time t−1, β is a constant such that 0>β>1, R(t) is the number of output bits over the time t, R(t−1) is the number of output bits over the immediately preceding time t−1 and ε and η are constants.
3. A method as claimed in claim 1 , wherein the quantization parameter over the time t is given by:
QP ( t )=(1−β)Δ qp+βQP ( t− 1)
where:
Δ qp =(ε( V ( t )− V ( t− 1))+ R ( t− 1) V ( t− 1)− R ( t ) V ( t ))/η
where QP(t−1) is the quantization parameter over an immediately preceding time t−1, β is a constant such that 0>β>1, R(t) is the number of output bits over the time t, R(t−1) is the number of output bits over the immediately preceding time t−1 and:
η
=
R
(
t
-
1
)
-
R
(
t
)
QP
(
t
)
V
(
t
)
-
QP
(
t
-
1
)
V
(
t
-
1
)
,
and
ɛ
=
R
(
t
)
+
R
(
t
-
1
)
-
R
(
t
)
1
-
V
(
t
)
·
QP
(
t
-
1
)
V
(
t
-
1
)
·
QP
(
t
)
.
4. A method as claimed in claim 1 , wherein the quantization parameter is further dependent on a current fill rate of an output buffer.
5. A method as claimed in claim 4 , wherein the quantization parameter is increased if the rate exceeds an upper threshold corresponding to overfilling of the output buffer and decreased if the rate is below a lower threshold corresponding to underfilling of the output buffer.
6. A method as claimed in claim 1 , wherein the time t corresponds to one of a macroblock, a slice, a picture and a group of pictures.
7. A method as claimed in claim 1 , further comprising calculating the quantization parameter dependant on a current average value of a motion vector for the video signal.
8. A method as claimed in claim 7 , wherein the quantization parameter is adapted by the equation: QP m =QP s +ΔQP where ΔQP=φ(V m −V s )/V s +τ|MV| 2 where φ and τ are parameters selected to provide a chosen degree of adaptation and QP m and QP s are values of the quantization parameter for a current macroblock and current slice respectively and V m and V s are the variances of the current macroblock and current slice respectively and MV is an average of the N most recent values of the motion vector such that
MV
2
=
1
N
∑
i
=
0
N
-
1
(
x
i
2
+
y
i
2
)
,
where x i and y i are horizontal and vertical components respectively of the motion vector.
9. A video rate controller for a video encoding system for encoding a video signal, the video rate controller comprising:
at least one processor;
a non-transitory, computer-readable storage medium, coupled to the at least one processor, wherein the non-transitory, computer-readable storage medium further comprises computer-readable instructions, when executed by the at least one processor, configured to:
determine a variance of pixel values over a time t of the video signal; and
calculate a quantization parameter for the time t using the pixel variance and a number of output bits over the time t and a quantization parameter used for an immediately preceding time t−1, a pixel variance over the immediately preceding time t−1, and a number of output bits over the immediately preceding time t−1, wherein
the variance is a modified variance being a function of the calculated variance over the time t and a similarly modified variance over the immediately preceding time t−1, and
wherein the modified variance V(t) is defined as: V(t)=(1−α)V a (t)+αV(t−1) where V a (t) is the calculated variance over a time t, V(t−1) is the modified variance over the immediately preceding time t−1 and α is a constant such that 0<α<1.
10. A video rate controller as claimed in claim 9 , wherein the quantization parameter over the time t is given by:
QP ( t )=(1−β)Δ qp+βQP ( t− 1)
where:
Δ qp =(ε( V ( t )− V ( t− 1))+ R ( t− 1) V ( t− 1)− R ( t ) V ( t ))/η
where QP(t−1) is the quantization parameter over an immediately preceding time t−1, β is a constant such that 0>β>1, R(t) is a number of output bits over the time t, R(t−1) is a number of output bits over the immediately preceding time t−1 and ε and η are constants.
11. A video rate controller as claimed in claim 9 , wherein the quantization parameter over the time t is given by:
QP ( t )=(1−β)Δ qp+βQP ( t− 1)
where:
Δ qp =(ε( V ( t )− V ( t− 1))+ R ( t− 1) V ( t− 1)− R ( t ) V ( t ))/η
where QP(t−1) is the quantization parameter over an immediately preceding time t−1, β is a constant such that 0>β>1, R(t) is a number of output bits over the time t, R(t−1) is a number of output bits over the immediately preceding time t−1, where:
η
=
R
(
t
-
1
)
-
R
(
t
)
QP
(
t
)
V
(
t
)
-
QP
(
t
-
1
)
V
(
t
-
1
)
,
and
ɛ
=
R
(
t
)
+
R
(
t
-
1
)
-
R
(
t
)
1
-
V
(
t
)
·
QP
(
t
-
1
)
V
(
t
-
1
)
·
QP
(
t
)
.
12. A video rate controller as claimed in claim 9 , wherein the quantization parameter is further dependent on a current fill rate of an output buffer.
13. A video rate controller as claimed in claim 12 , wherein the quantization parameter is increased if the rate exceeds an upper threshold corresponding to overfilling of the output buffer and decreased if the rate is below a lower threshold corresponding to underfilling of the output buffer.
14. A video rate controller as claimed in claim 9 , wherein the time t corresponds to one of a macroblock, a slice, a picture and a group of pictures.
15. A video rate controller as claimed in claim 9 , further arranged to calculate the quantization parameter dependant on a current average value of a motion vector for the video signal.
16. A video rate controller as claimed in claim 15 , wherein the quantization parameter is adapted by the equation: QP m =QP s +ΔQP where ΔQP=φ(V m −V s )/V s +τ|MV| 2 where φ and τ are parameters selected for a chosen degree of adaptation and QP m and QP s are values of the uantization parameter for a current macroblock and current slice respectively and V m and V s are the variances of the current macroblock and current slice respectively and MV is an average of the N most recent values of the motion vector such that
MV
2
=
1
N
∑
i
=
0
N
-
1
(
x
i
2
+
y
i
2
)
,
where x i and y i are horizontal and vertical components respectively of the motion vector.
17. A non-transitory computer readable medium embodying stored computer executable software code, when executed by a processor, the computer executable software code is configured to:
determine a variance of pixel values over a time t of the video signal; and
calculate a quantization parameter for the time t using the pixel variance and a number of output bits over the time t and a quantization parameter used for an immediately preceding time t−1, a pixel variance over the immediately preceding time t−1, and a number of output bits over the immediately preceding time t−1, wherein
the variance is a modified variance being a function of the calculated variance over the time t and a similarly modified variance over the immediately preceding time t−1, and
wherein the modified variance V(t) is defined as: V(t)=(1−α)V a (t)+αV(t−1) where V a (t) is the calculated variance over a time t, V(t−1) is the modified variance over the immediately preceding time t−1 and a is a constant such that 0<α<1.
18. A computer readable medium as claimed in claim 17 , wherein the quantization parameter over the time t is given by:
QP ( t )=(1−β)Δ qp+βQP ( t− 1)
where:
Δ qp =(ε( V ( t )− V ( t− 1))+ R ( t− 1) V ( t− 1)− R ( t ) V ( t ))/η
where QP(t−1) is the quantization parameter used for an immediately preceding time t−1, β is a constant such that 0<β<1, R(t) is the number of output bits over the time t, R(t−1) is the number of output bits over the immediately preceding time t−1 and ε and η are constants.
19. A computer readable medium as claimed in claim 17 , wherein the quantization parameter over the time t is given by:
QP ( t )=(1−β)Δ qp+βQP ( t− 1)
where:
Δ qp =(ε( V ( t )− V ( t− 1))+ R ( t− 1) V ( t− 1)− R ( t ) V ( t ))/η
where QP(t−1) is the quantization parameter used for an immediately preceding time t−1, β is a constant such that 0<β<1, R(t) is the number of output bits over the time t, R(t−1) is the number of output bits over the immediately preceding time t−1, where:
η
=
R
(
t
-
1
)
-
R
(
t
)
QP
(
t
)
V
(
t
)
-
QP
(
t
-
1
)
V
(
t
-
1
)
,
and
ɛ
=
R
(
t
)
+
R
(
t
-
1
)
-
R
(
t
)
1
-
V
(
t
)
·
QP
(
t
-
1
)
V
(
t
-
1
)
·
QP
(
t
)
.
20. A computer readable medium as claimed in claim 17 , wherein the quantization parameter is further dependent on a current fill rate of an output buffer.
21. A computer readable medium as claimed in claim 20 , wherein the quantization parameter is increased if the rate exceeds an upper threshold corresponding to overfilling of the output buffer and decreased if the rate is below a lower threshold corresponding to underfilling of the output buffer.
22. A computer readable medium as claimed in claim 17 , wherein the time t corresponds to one of a macroblock, a slice, a picture and a group of pictures.
23. A computer readable medium as claimed in claim 17 , for further calculating the quantization parameter dependant on a current average value of a motion vector for the video signal.
24. A computer readable medium as claimed in claim 23 , wherein the quantization parameter is adapted by the equation: QP m =QP s +ΔQP where ΔQP=φ(V m −V s )/V s +τ|MV| 2 where φ and τ are parameters selected for a chosen degree of adaptation and QP m and QP s are values of the quantization parameter for a current macroblock and current slice respectively and V m and V s are the variances of the current macroblock and current slice respectively and MV is an average of the N most recent values of the motion vector such that
MV
2
=
1
N
∑
i
=
0
N
-
1
(
x
i
2
+
y
i
2
)
,
where x i and y i are horizontal and vertical components respectively of the motion vector.