Rate control method of perceptual-based rate-distortion optimized bit allocation
A rate control method of perceptual-based rate-distortion (R-D) optimized bit allocation is disclosed. An input frame is firstly determined as a key frame or non-key frame. A key frame is additionally encoded to generate rate-distortion (R-D) points. The R-D model of each basic unit (BU) is updated, followed by perceptual-based bit allocation, thereby generating a target bit rate. A quantization parameter (QP) is computed according to the target bit rate, and the current BU is encoded according to the QP. A rate-quantization (R-Q) model is updated if not all BUs have been encoded.
1. A rate control method of perceptual-based rate-distortion (R-D) optimized bit allocation, comprising:
using a coder to perform the steps of:
determining an input frame as a key frame;
if the input frame is the key frame, additionally encoding the key frame at least one time, thereby generating a corresponding R-D point;
updating an R-D model of each basic unit (BU);
performing perceptual-based bit allocation, thereby generating a target bit rate for each said BU;
computing a quantization parameter (QP) according to the target bit rate;
encoding the current BU according to the QP; and
updating a rate-quantization (R-Q) model if not all the BUs have been encoded;
wherein the R-D model is expressed as
D ( R )= αe −βR
where R is a bit rate, D is distortion, α and β are model parameters, which are positive;
wherein the updated model parameters α* and β* are obtained by minimizing
∑
k
=
1
D
d
~
k
(
ln
d
~
k
-
α
-
β
r
~
k
)
2
where {tilde over (r)} is encoded bits of the encoded BU, {tilde over (d)} is encoded distortion of the encoded BU, D is a set of data points for updating the R-D model of the BU, and |D| is a number of the data points in D.
2. The method of claim 1 , wherein a first frame of a whole video sequence or a first frame of a sub-sequence is the key frame.
3. The method of claim 1 , further comprising a step of storing the at least one additionally generated R-D point.
4. The method of claim 1 , further comprising a step of storing the R-D point generated from the encoded current BU.
5. The method of claim 1 , wherein the key frame is additionally encoded using the different QP at each time.
6. The method of claim 5 , wherein the key frame is additionally encoded twice to obtain two different said QPs, which are (QP avg +Δ)and (QP avg −Δ) respectively, where the QP avg is the average QP of all the BUs in a previous frame and Δ is a constant.
7. The method of claim 1 , wherein the R-D model of each said BU is updated according to a previous frame or frames.
8. The method of claim 1 , wherein the R-D model of each said BU is updated according to the previous BU or BUs in the current frame.
9. The method of claim 1 , wherein the updated model parameters α* and β* are obtained according to least mean square error (LMSE) by regression.
10. The method of claim 1 , wherein a structural similarity (SSIM) index is based to construct the R-D model for the bit allocation.
11. The method of claim 10 , wherein the SSIM-based bit allocation is expressed as
min
r
i
∑
i
=
1
N
b
α
i
ⅇ
β
i
r
i
where α i and β i are model parameters of the i-th BU, r i is bit budget allocated to the i-th BU, and N b is a number of the BUs in the frame;
wherein a sum of the bit budget of all the BUs is less than or equal to the target bit rate for the current frame.
12. The method of claim 11 , wherein the bit budget for each said BU has an upper bound, which is a maximum number of bits for encoding one of the BUs in the previous frame; and the bit budget for each said BU has a lower bound, which is proportional to a channel rate and is inversely proportional to a frame rate.
13. The method of claim 1 , wherein the step of updating the R-Q model further comprises:
updating a mean absolute difference (MAD) model.
14. The method of claim 13 , wherein the QP is computed using the R-Q model.
15. The method of claim 14 , wherein the step of computing the QP comprises:
solving the R-Q model of multi-degree to obtain a quantization step, wherein the R-Q model includes the MAD model; and
obtaining the QP by mapping the quantization step to the QP by a lookup table.
16. The method of claim 15 , wherein an absolute difference between the QP and the average QP is less than or equal to a predetermined value.
17. The method of claim 13 , wherein the R-Q model and the MAD model are updated according the previous BU or BUs in the current frame.
18. The method of claim 13 , wherein the R-Q model and the MAD model are updated according the previous frame or frames.
19. The method of claim 13 , wherein the step of updating the R-Q model and the MAD model further comprises:
generating the target bit rate for the remaining BUs in the current frame.
20. The method of claim 19 , in the step of generating the target bit rate for the remaining BUs in the current frame, the allocated bit r i* in the beginning of the current encoded frame is reused, and the target bit rate for the i-th BU is expressed as follows:
t
i
=
T
i
-
1
r
i
*
∑
k
=
1
N
b
r
k
*
where T i is the target bit rate for the remaining BUs in the current frame after the i-th BU is encoded.