Systems and methods for resetting rate control state variables upon the detection of a scene change within a group of pictures
View Patent ↗The invention is related to methods and apparatus that advantageously improve picture quality in a video encoder, such as an MPEG video encoder. Scene changes typically occur relatively frequently in picture sequences, such as movies. One embodiment of the invention detects a scene change within a group of pictures and allocates bits within the group of pictures in response to the detected scene change without changing a predetermined structure for the group of pictures.
1. A method for allocating bits in a video encoding process, the method comprising:
receiving an indication for a group of pictures for encoding, wherein the indication includes a structure indicating picture types for the pictures in the group of pictures;
comparing a picture in the group of pictures to a previous picture in the group of pictures, the picture having bits allocated thereto for encoding;
determining from the comparison that a scene change has occurred in the group of pictures at the picture;
determining whether the picture is a P-picture; and
reallocating bits for encoding remaining pictures in the group of pictures in response to the determination that there has been a scene change at the picture and that the picture is a P-picture, wherein such reallocation occurs without changing the structure of picture types, and the reallocated bits are allocated to the P-picture with the scene change according to:
T
p
′
=
max
{
(
R
(
1
+
(
N
p
-
1
)
X
p
X
i
K
p
+
N
b
X
b
X
i
K
b
)
)
,
(
bit_rate
8
·
picture_rate
)
}
where T p′ corresponds to the reallocation bits for the P-picture, where R corresponds to a remaining number of bits reallocated to the group of pictures, where N p corresponds to a number of P-pictures remaining to be encoded in the group, where N b corresponds to a number of B-pictures remaining to be encoded in the group, where bit_rate corresponds to a constant bit rate at which data is transmitted in a data channel, where picture_rate corresponds to a rate at which pictures are presented, where X i , X p , and X b correspond to constant bit rate complexity estimators respectively for I-pictures, for P-pictures, and for B-pictures, and where K p and K b are values determined by quantization matrices used to encode the pictures.
2. The method as defined in claim 1 , wherein the previous picture is determined by encoding order and not by presentation order.
3. The method as defined in claim 1 , wherein the video encoding process is performed in real time.
4. The method as defined in claim 1 , wherein the video encoding process is adapted to encode video for transmission over a constant-bit-rate data channel.
5. The method as defined in claim 1 , wherein comparing is performed by computing a sum of absolute differences (SAD) measurement.
6. The method as defined in claim 1 , wherein comparing further comprises reusing the results of a comparison between pictures that has already been performed in connection with a motion estimation task.
7. The method as defined in claim 1 , wherein reallocating bits among the remaining pictures further comprises allocating more bits to the P-picture with the scene change than were originally allocated and allocating fewer bits to the other pictures in the remaining pictures than were originally allocated.
8. The method as defined in claim 1 , wherein reallocating bits further comprises allocating bits to the P-picture with the scene change as if the P-picture were an I-picture.
9. The method as defined in claim 1 , further comprising resetting complexity estimators X i , X p , and X b upon the detection of the scene change in the P-picture.
10. A method for allocating bits in a real-time video encoding process, the method comprising:
detecting a scene change in a P-picture of a group of pictures, the P-picture having bits allocated thereto; and
reallocating bits in response to the detection of the scene change, where the reallocation excludes changing a type of picture in the group of pictures, and the bits are reallocated to the P-picture with the scene change according to:
T
p
′
=
max
{
(
R
(
1
+
(
N
p
-
1
)
X
p
X
i
K
p
+
N
b
X
b
X
i
K
b
)
)
,
(
bit_rate
8
·
picture_rate
)
}
where T p′ corresponds to the reallocation bits for the P-picture, where R corresponds to a remaining number of bits reallocated to the group of pictures, where N p corresponds to a number of P-pictures remaining to be encoded in the group, where N b corresponds to a number of B-pictures remaining to be encoded in the group, where bit_rate corresponds to a constant bit rate at which data is transmitted in a data channel, where picture_rate corresponds to a rate at which pictures are presented, where X i , X p , and X b correspond to constant bit rate complexity estimators respectively for I-pictures, for P-pictures, and for B-pictures, and where K p and K b are values determined by quantization matrices that are used to encode the pictures.
11. The method as defined in claim 10 , wherein the P-picture is reallocated with an amount of bits that would have been allocated had the P-picture been an I-picture.
12. A computer readable medium encoded with computer executable instructions in the form of a computer program for allocating bits in a video encoding process, that:
receives an indication for a group of pictures for encoding, wherein the indication includes a structure indicating picture types for the pictures in the group of pictures;
allocates a target number of bits for a picture in the group of pictures;
compares the picture in the group of pictures to a previous picture in the group of pictures;
determines from the comparison to determine that a scene change has occurred in the group of pictures at the picture;
determines that the picture is a P-picture; and
reallocates bits for encoding remaining pictures in the group of pictures in response to the determination that there has been a scene change at the picture and that the picture is a P-picture, wherein such reallocation occurs without changing the structure of picture types, and the bits are allocated to the P-picture with the scene change according to:
T
p
′
=
max
{
(
R
(
1
+
(
N
p
-
1
)
X
p
X
i
K
p
+
N
b
X
b
X
i
K
b
)
)
,
(
bit_rate
8
·
picture_rate
)
}
where T p′ corresponds to the reallocation bits for the P-picture, where R corresponds to a remaining number of bits reallocated to the group of pictures, where N p corresponds to a number of P-pictures remaining to be encoded in the group, where N b corresponds to a number of B-pictures remaining to be encoded in the group, where bit_rate corresponds to a constant bit rate at which data is transmitted in a data channel, where picture_rate corresponds to a rate at which pictures are presented, where X i , X p , and X b correspond to constant bit rate complexity estimators respectively for I-pictures, for P-pictures, and for B-pictures, and where K p and K b are values determined from quantization matrices used to encode the pictures.
13. A circuit adapted to allocate bits in a video encoding process comprising:
a module adapted to receiving an indication for a group of pictures for encoding, wherein the indication includes a structure indicating picture types for the pictures in the group of pictures;
a module adapted to compare a picture in the group of pictures to a previous picture in the group of pictures, the picture having bits for encoding allocated thereto;
a module adapted to determine whether a scene change has occurred in the group of pictures at the picture;
a module adapted to determine whether the picture is a P-picture; and
a module adapted to reallocate bits for encoding remaining pictures of the group of pictures in response to the determination that there has been a scene change at the picture and that the picture is a P-picture, wherein such reallocation occurs without changing the structure of picture types, and the bits are reallocated to the P-picture with the scene change according to:
T
p
′
=
max
{
(
R
(
1
+
(
N
p
-
1
)
X
p
X
i
K
p
+
N
b
X
b
X
i
K
b
)
)
,
(
bit_rate
8
·
picture_rate
)
}
where T p′ corresponds to the reallocation bits for the P-picture, where R corresponds to a remaining number of bits reallocated to the group of pictures, where N p corresponds to a number of P-pictures remaining to be encoded in the group, where N b corresponds to a number of B-pictures remaining to be encoded in the group, where bit_rate corresponds to a constant bit rate at which data is transmitted in a data channel, where picture_rate corresponds to a rate at which pictures are presented, where X i , X p , and X b correspond to constant bit rate complexity estimators respectively for I-pictures, for P-pictures, and for B-pictures, and where K p and K b are values determined from quantization matrices used to encode the pictures.
14. A video encoder that embodies the circuit of claim 13 .