Systems and methods for allocating bits to macroblocks within a picture depending on the motion activity of macroblocks as calculated by an L1 norm of the residual signals of the macroblocks
View Patent ↗The invention is related to methods and apparatus that can advantageously be used in a video encoder to improve picture quality, to improve the speed of encoding, and the like. One embodiment of the invention advantageously computes activity measures using an efficient L1-norm, which can advantageously be relatively quickly computed by selected microprocessors. Another embodiment of the invention advantageously allocates bits to macroblocks of a picture based at least in part on the motion activities of the macroblocks.
1. A method of allocating bits to macroblocks of a picture in a video encoding process, where the bits are allocated at least in part according to a motion activity of the macroblocks, the method comprising the steps of:
a. receiving a first group of pictures, each picture of said first group of pictures being selected from a group consisting of an I-picture, a P-picture, and a B-picture, wherein the macroblocks of an I-picture are intra coded, the macroblocks of a P-picture are either intra coded or forward predictively coded, and the macroblocks of a B-picture are either intra coded, or forward predictively coded, or backward predictively coded, or interpolated;
b. retrieving a picture mode of said first group of pictures, said picture mode defining a sequence of said I-pictures, P-pictures, and B-pictures in said first group of pictures;
c. determining a number N p , N b of each of said P-pictures and B-pictures, respectively, in said first group of pictures to be encoded;
d. calculating values of complexity estimators X i , X p , X b for each of said I-pictures, P-pictures, and B-pictures, respectively, in said first group of pictures;
e. calculating a number R of bits allocated to said first group of pictures to be encoded remaining after encoding a respective I-picture, P-picture or B-picture, wherein R=R prev −S i,p,b ,
the R prev being a number of bits allocated to said first group of pictures prior to encoding of said respective I-picture, P-picture, or B-picture, and the S i,p,b being a number of bits used to encode said respective I-picture, P-picture, or B-picture, respectively; and
f. calculating an initial target number of bits T i , T p , T b for each I-picture, P-picture, and B-picture, respectively, to be encoded in said first group of pictures, wherein
T
i
=
max
{
(
R
(
1
_
+
N
p
X
p
X
i
K
p
+
N
b
X
b
X
i
K
b
)
)
,
(
bit_rate
8
·
picture_rate
)
}
T
p
=
max
{
(
R
(
N
p
+
N
b
K
p
X
b
K
b
X
p
)
)
,
(
bit_rate
8
·
picture_rate
)
}
T
b
=
max
{
(
R
(
N
b
+
N
p
K
b
X
p
K
p
X
b
)
,
bit_rate
8
·
picture_rate
}
wherein bit_rate corresponds to a bit rate of a data transmission channel,
wherein picture_rate corresponds to a number of pictures per second transmitted via said data transmission channel, and
wherein K p and K b are universal constants depending on quantization matrices for the P-pictures and B-pictures, respectively, to be encoded;
allocating a respective virtual buffer for each of said I-pictures, P-pictures and B-pictures in said first group of pictures;
calculating virtual buffer fullness d j i , d j p , and d j b of each of said respective virtual buffer for said I-pictures, P-pictures, and B-pictures, wherein j is a number of a macroblock being encoded; and
updating a respective virtual buffer fullness d j i , d j p , and d j b in accordance with
d
j
i
=
d
o
i
+
B
j
-
1
-
(
T
i
·
Mact_sum
j
-
1
MACT
)
;
d
j
p
=
d
0
p
+
B
j
-
1
-
(
T
p
·
Mact_sum
j
-
1
MACT
)
;
and
d
j
b
=
d
o
b
+
B
j
-
1
-
(
T
b
·
Mact_sum
j
-
1
MACT
)
,
wherein d 0 i , d 0 p , and d j b correspond to said respective virtual buffer fullness prior to encoding of the j th macroblock,
wherein B j−1 corresponds to a number of bits used to encode macroblocks prior to encoding said j th macroblock,
wherein T i corresponds to the target bit allocation for a next picture to be encoded when the picture is the I-picture that starts a group of pictures, T p corresponds to the target bit allocation for a next picture to be encoded when the next picture is a P-picture, and T b corresponds to the target bit allocation for a next picture to be encoded when the next picture is a B-picture,
wherein the variable MACT represents the sum of the motion activity of all of the macroblocks in the pictures of the first group of pictures, and wherein the variable Mact_sum j−1 corresponds to the sum of the motion activity of the macroblocks in the picture that have been encoded.
2. A method of allocating bits to macroblocks of a picture in a video encoding process, where the bits are allocated at least in part according to a motion activity of the macroblocks, the method comprising the steps of:
a. receiving a first group of pictures, each picture of said first group of pictures being selected from a group consisting of an I-picture, a P-picture, and a B-picture, wherein the macroblocks of an I-picture are intra coded, the macroblocks of a P-picture are either intra coded or forward predictively coded, and the macroblocks of a B-picture are either intra coded, or forward predictively coded, or backward predictively coded, or interpolated;
b. retrieving a picture mode of said first group of pictures, said picture mode defining a sequence of said I-pictures, P-pictures, and B-pictures in said first group of pictures;
c. determining a number N p , N b of each of said P-pictures and B-pictures, respectively, in said first group of pictures to be encoded;
d. calculating values of complexity estimators X i , X p , X b for each of said I-pictures, P-pictures, and B-pictures, respectively, in said first group of pictures;
e. calculating a number R of bits allocated to said first group of pictures to be encoded remaining after encoding a respective I-picture, P-picture or B-picture, wherein R=R prev −S i,p,b ,
the R prev being a number of bits allocated to said first group of pictures prior to encoding of said respective I-picture, P-picture, or B-picture, and the S i,p,b being a number of bits used to encode said respective I-picture, P-picture, or B-picture, respectively; and
f. calculating an initial target number of bits T i , TP, T b for each I-picture, P-picture, and B-picture, respectively, to be encoded in said first group of pictures, wherein
T
i
=
max
{
(
R
(
1
_
+
N
p
X
p
X
i
K
p
+
N
b
X
b
X
i
K
b
)
)
,
(
bit_rate
8
·
picture_rate
)
}
T
p
=
max
{
(
R
(
N
p
+
N
p
K
p
X
b
K
b
X
p
)
)
,
(
bit_rate
8
·
picture_rate
)
}
T
b
=
max
{
R
(
N
b
+
N
p
K
b
X
p
K
p
X
b
)
,
bit_rate
8
·
picture_rate
}
wherein bit_rate corresponds to a bit rate of a data transmission channel,
wherein picture_rate corresponds to a number of pictures per second transmitted via said data transmission channel, and
wherein K p and K b are universal constants depending on quantization matrices for the P-pictures and B-pictures, respectively, to be encoded;
allocating a respective virtual buffer for each of said I-pictures, P-pictures and B-pictures in said first group of pictures;
calculating virtual buffer fullness d j i , d j p , and d j b of each of said respective virtual buffer for said I-pictures, P-pictures, and B-pictures, wherein j is a number of a macroblock being encoded; and
updating a respective virtual buffer fullness d j i , d j p ,and d j b in accordance with
d
j
i
=
d
o
i
+
B
j
-
1
-
(
α
i
T
i
·
(
j
-
1
)
MB_cnt
+
(
1
-
α
i
)
T
i
·
Mact_sum
j
-
1
MACT
)
;
d
j
p
=
d
o
p
+
B
j
-
1
-
(
α
p
T
p
·
(
j
-
1
)
MB_cnt
+
(
1
-
α
p
)
T
p
·
Mact_sum
j
-
1
MACT
)
;
and
d
j
b
=
d
o
b
+
B
j
-
1
-
(
α
b
T
b
·
(
j
-
1
)
MB_cnt
+
(
1
-
α
b
)
T
b
·
Mact_sum
j
-
1
MACT
)
;
wherein the variables d j i , d j p , and d j b represent a respective virtual buffer fullness for I-pictures, for P-pictures, and for B-pictures, respectively,
wherein the variable j represents the number of the encoded macroblock,
wherein B j−1 corresponds to the number of bits used to encode the macroblocks up to but not including the j-th macroblock,
wherein T i corresponds to the target bit allocation for the next picture to be encoded when the picture is the I-picture that starts a group of pictures, T p corresponds to the target bit allocation for a next picture to be encoded when the next picture is a P-picture, and T b corresponds to the target bit allocation for a next picture to be encoded when the next picture is a B-picture,
wherein the variable MACT represents the sum of the motion activity of all of the macroblocks in said first group of pictures,
wherein the variable Mact_sum j−1 corresponds to the sum of the motion activity of the macroblocks in the picture that have been encoded, and MB_cnt corresponds to the number of macroblocks in the picture, and
wherein a i , a p , and a b correspond to weighting factors for allocation of bits to macroblocks within I-pictures, P-pictures, and B-pictures, respectively.
3. The method as defined in claim 2 , wherein α i , α p , and α b are selected in a range from about 0 to about 1.
4. The method as defined in claim 2 , wherein α i , α p , and α b are selected in a range from about 0 to about 0.5.
5. The method as defined in claim 2 , wherein α i , α p , and α b selected from the group consisting of 0.2, 0.3, 0.4 and 0.5.