CN104702961A - Code rate control method for distributed video coding - Google Patents

Abstract

The invention discloses a method for building related noise model and estimation model parameters in distributed video coding. A hybrid model uses K-Mediods to divide residual coefficients into small residual and large residual, uses improved Laplacian distribution to describe the distribution of small residual coefficients and uses cauchy distribution to describe large residual coefficients. The hybrid model (Hybrid Distribution Correlation Noise Model, HDCNM) is capable of precisely describing the residual coefficient distribution between WZ frame and side information, and accordingly the rate-distortion performance of the transform domain distributed type video coding is effectively improved, and the computation complexity of the decoding end of a system is reduced.

Description

Bit rate control method in a kind of distributed video coding

Technical field

The present invention relates to one bit rate control method in distributed video coding, belong to field of video compression.

Background technology

The information theory that distributed video coding (Distributed Video Coding, DVC) proposes based on 20 century 70 Slepian.Wolf and Wyner.Ziv is set up, and encoder computational complexity is transferred to decoder.Be characterized in that coding is simple, decoding is more complicated, compression performance is strong close to traditional coded system, resist miscode ability, be applicable to video encoder etc. resource-constrained in wireless network.

In distributed video coding, Turbo code or the such chnnel coding of LDPC code of using is encoded to WZ frame more.Although Turbo code and LDPC code are all can close to the channel coding schemes of theoretical circles, the overall performance of LDPC code is better than Turbo code to have result of study to show, for the video of motion intense, LDPC scheme resist miscode rate performance is better.In distributed video coding, between original WZ frame and corresponding edge information, the accuracy of correlated noise model has a significant impact code efficiency, and correlated noise model is more accurate, LDPC code successfully decode need check digit fewer, reduce code check on the one hand, improve compression efficiency; On the other hand, the amount of calculation of LDPC code decoding is decreased.So the accurate modeling of decoding end to correlated noise is a key technology of distributed video coding.

Laplacian distribution model due to sub-band levels has good compromise and obtains extensive accreditation on computation complexity and accuracy, DVC system now adopts laplacian distribution to describe the noise relationship between original WZ frame and side information mostly, is mainly how to obtain more accurate Laplacian parameter in recent years to the research of correlated noise model.But the DCT residual error coefficient Distribution Statistics in DVC system and not in full conformity with laplacian distribution this hypothesis, but there is sharper peak feature and longer afterbody, for adapting to these two characteristics of residual error coefficient, the modeling method mixing relevant noise model based on the Laplce-Cauchy of K-Medoids cluster is proposed herein.This mixed model utilizes K-Medoids that subband residual error is divided into large residual error coefficient and little residual error coefficient, laplace model is used to describe to the distribution of little residual error coefficient, use Cauchy to distribute to the distribution of large residual error coefficient to describe, add the precision of model, thus improve the distortion performance of system.

Summary of the invention

Technical problem: there is the problem that large residual sum little residual error coefficient Distribution Statistics and traditional laplacian distribution exist certain deviation for the residual error coefficient of original Wyner-Ziv (WZ) frame and corresponding edge information in Transform Domain Distribution formula Video coding.In order to reduce this species diversity, a kind of mixed distribution correlated noise model based on K-Mediods and parameter estimation algorithm thereof are proposed.This mixed model utilizes the laplacian distribution improved to describe the distribution of little residual error coefficient, adopts Cauchy to distribute and describes large residual error coefficient.Mixed model modeling method in this paper more accurately can describe the residual error coefficient distribution between WZ frame and side information, thus effectively improves the distortion performance of Transform Domain Distribution formula Video coding, and reduces system decodes end computation complexity.

Technical scheme:

1. a correlated noise model modelling approach in distributed video coding, it is characterized in that, the method includes the steps of:

1) for current decoded sub-band b _k, the absolute value of the distance of each coefficient and it and this subband residual mean square (RMS) forms 2 dimensional feature vectors, i.e. subband b _kin the n-th residual error characteristic vector be expressed as wherein $D_{b_k}(\mathrm{\μ},v)=C_{R_{\mathrm{XY}}}^{b_x}(\mathrm{\μ},v)-E\left(C_{R_{\mathrm{XY}}}^{b_k}\right);$

2) K-Medoids clustering algorithm is utilized residual error characteristic vector to be divided into large residual error class and little residual error class;

3) use the laplacian distribution improved to describe little residual error class, use Cauchy to distribute and describe large residual error class.Calculate corresponding distributed constant respectively, finally obtain mixed distribution formula correlated noise model;

In step 2) in, carry out K-Medoids cluster in accordance with the following steps:

1) initialization cluster centre: select three the residual error characteristic vectors started as initial cluster center corresponding class is S ₁ ^(k), S ₂ ^(k), S ₃ ^(k), make k=0;

2) sample clustering: by set of eigenvectors to be sorted the a certain class in three classes is allocated to one by one, if that is: by minimal distance principle $d_{\mathrm{ij}}^{\left(k\right)}=\underset{j}{\min }\left[d_{\mathrm{ij}}^{\left(k\right)}\right],(i=\mathrm{1,2},...,N),j=\left(\mathrm{1,2,3}\right),$ Then in formula represent and class center distance, superscript k represents iterations, and distance here selects Euclidean distance, so produce new cluster

3) cluster centre is recalculated: different from K-Means, K-Means selects the mean value of all data points in current cluster to be new central point, and in K-Medoids, the minimum point of the distance sum of (in current cluster) point to other are all will be chosen as central point from current cluster;

4) end condition is judged: if (j=1,2,3), then terminate, and obtains 3 clusters, and DCT residual error coefficient is divided into three S set ₁, S ₂, S ₃, otherwise k=k+1, goes to 2).

In described step 3) in, the idiographic flow calculating the parameter of laplacian distribution and the parameter of Cauchy's distribution is:

After cluster, calculate three residual error coefficient S set _jthe variance that (j=1,2,3) are respective, then by order from small to large, is designated as S respectively by the residual error coefficient set of its correspondence ₁, S ₂, S ₃, wherein S ₁be designated as little coefficient set, and S ₂, S ₃then be designated as large coefficient set, S ₁substantially symmetrical about 0, in order to calculate the convenience of the parameter of Cauchy's distribution, to S ₁revise, note S ₁the upper bound and the minimum value of absolute value of lower bound be TL, large residual error coefficient collection will be described with Cauchy's distribution afterwards, and describe little residual error coefficient collection by laplacian distribution;

Cauchy's distribution can be expressed as:

$p\left(x\right)=\frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+{(x-\mathrm{\μ})}^2}$

Wherein λ is form parameter, and μ is location parameter, because in video compression coding, residual error coefficient is substantially symmetrical about 0, so make μ=0.Can also meet to keep the probability density finally represented by laplacian distribution and Cauchy's distributed combination so for Cauchy's distribution, a λ must be found to make the integrated value of its probability density in [-TL, TL] interval equal the integrated value of probability density in [-TL, TL] interval of Laplce, and still can maintain its heavy-tailed property.So just λ can be derived.

The integrated value of Laplacian probability density in [-TL, TL] is made to be P _l(TL), it can be calculated as follows:

$P_L\left(\mathrm{TL}\right)=P(\mathrm{TL}\≤x\≤\mathrm{TL})={\∫}_{-\mathrm{TL}}^{\mathrm{TL}}\frac{{\mathrm{\α}}_{b_k}}{2}\exp (-{\mathrm{\α}}_{b_k}|x\left|\right)\mathrm{dx}=C$

Wherein press ${\mathrm{\α}}_{b_k}=\sqrt{2/{\mathrm{\σ}}_{b_k}^2},{\mathrm{\σ}}_{b_k}^2=E\left({\left|C_{R_{\mathrm{XY}}}^{b_k}\right|}^2\right)-E{\left(\right|C_{R_{\mathrm{XY}}}^{b_k}\left|\right)}^2$ Calculate.

Make P _c(LH) represent the integrated value of Cauchy probability density in [-TL, TL], be calculated as follows:

$\begin{array}{c}{P}_C\left(\mathrm{TL}\right)=P(-\mathrm{TL}\≤x\≤\mathrm{TL})={\∫}_{-\mathrm{TL}}^{\mathrm{TL}}\frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+x^2}\mathrm{dx}\\ =\frac{2}{\mathrm{\π}}{\tan }^{-1}\left(\frac{\mathrm{TL}}{\mathrm{\λ}}\right)=P_L\left(\mathrm{TL}\right)=C\end{array}$

So just can obtain: $\mathrm{\λ}=\frac{\mathrm{TL}}{\tan (\mathrm{\πC}/2)}.$

Utilize following formula to calculate and belong to S ₁the Laplacian parameter of residual error coefficient set:

${\mathrm{\α}}_1=\sqrt{\frac{2}{{\mathrm{\σ}}_1^2}}$

s ₁the variance of residual error coefficient set.

Finally can represent correlated noise model like this:

$p\left(C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\right)=\left\{\begin{array}{c}\frac{{\mathrm{\α}}_0}{2}\exp (-{\mathrm{\α}}_0|C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\left|\right),C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\∈S_1\\ \frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+{\left[C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\right]}^2},C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\∈S_2,S_3\end{array}\right.$

Beneficial effect: the present invention compared with prior art, has the following advantages:

A) the present invention program proposes the relevant noise model modeling method of a kind of mixing, in the method, adopts K-Medoids clustering algorithm to improve accuracy and the robustness of classification.Use laplacian distribution to describe the distribution of little residual error, use Cauchy to distribute and describe the distribution of large residual error.Improve accuracy and the adaptive performance of correlated noise model.

B) context of methods is by the improvement in decoding end, effectively improves the distortion performance of DVC system.

Accompanying drawing explanation

Fig. 1 is the existing distributed video coding frame diagram based on transform domain.

Fig. 2 is the video coding framework figure that the present invention improves.

Fig. 3 is that mixing of the present invention is correlated with noise model modeling method flow chart.

Embodiment

Below in conjunction with accompanying drawing and embodiment, the present invention is described in further detail.

Fig. 1 is the existing distributed video coding frame diagram based on transform domain, the present invention proposes a kind of method setting up correlated noise model and estimation model parameter in distributed video coding, first this mixed model utilizes K-Mediods residual error coefficient to be divided into the large residual error of little residual sum, utilize the laplacian distribution improved to describe the distribution of little residual error coefficient, adopt Cauchy to distribute and describe large residual error coefficient.Mixed model (the Hybrid Distribution CorrelationNoise Model that this article proposes, HDCNM) residual error coefficient distribution between WZ frame and side information can more accurately be described, thus effectively improve the distortion performance of Transform Domain Distribution formula Video coding, and reduce system decodes end computation complexity.Video coding framework after improvement as shown in Figure 2.

Provide the specific embodiment of the inventive method below:

1) for current decoded sub-band b _k, the absolute value of the distance of each coefficient and it and this subband residual mean square (RMS) forms 2 dimensional feature vectors, i.e. subband b _kin the n-th residual error characteristic vector be expressed as wherein $D_{b_k}(\mathrm{\μ},v)=C_{R_{\mathrm{XY}}}^{b_x}(\mathrm{\μ},v)-E\left(C_{R_{\mathrm{XY}}}^{b_k}\right);$

2) K-Medoids clustering algorithm is utilized residual error characteristic vector to be divided into large residual error class and little residual error class

3) use the laplacian distribution improved to describe little residual error class, use Cauchy to distribute and describe large residual error class.Calculate corresponding distributed constant respectively, finally obtain mixed distribution formula correlated noise model.

In step 2) in, carry out K-Medoids cluster in accordance with the following steps:

1) initialization cluster centre: select three the residual error characteristic vectors started as initial cluster center corresponding class is S ₁ ^(k), S ₂ ^(k), S ₃ ^(k), make k=0;

2) sample clustering: by set of eigenvectors to be sorted the a certain class in three classes is allocated to one by one, if that is: by minimal distance principle $d_{\mathrm{ij}}^{\left(k\right)}=\underset{j}{\min }\left[d_{\mathrm{ij}}^{\left(k\right)}\right],(i=\mathrm{1,2},...,N),j=\left(\mathrm{1,2,3}\right),$ Then in formula represent and class center distance, superscript k represents iterations, and distance here selects Euclidean distance, so produce new cluster

3) cluster centre is recalculated: different from K-Means, K-Means selects the mean value of all data points in current cluster to be new central point, and in K-Medoids, the minimum point of the distance sum of (in current cluster) point to other are all will be chosen as central point from current cluster;

4) end condition is judged: if (j=1,2,3), then terminate, and obtains 3 clusters, and DCT residual error coefficient is divided into three S set ₁, S ₂, S ₃, otherwise k=k+1, goes to 2).

In described step 3) in, the idiographic flow calculating the parameter of laplacian distribution and the parameter of Cauchy's distribution is:

After cluster, calculate three residual error coefficient S set _jthe variance that (j=1,2,3) are respective, then by order from small to large, is designated as S respectively by the residual error coefficient set of its correspondence ₁, S ₂, S ₃, wherein S ₁be designated as little coefficient set, and S ₂, S ₃then be designated as large coefficient set, S ₁substantially symmetrical about 0, in order to calculate the convenience of the parameter of Cauchy's distribution, to S ₁revise, note S ₁the upper bound and the minimum value of absolute value of lower bound be TL, large residual error coefficient collection will be described with Cauchy's distribution afterwards, and describe little residual error coefficient collection by laplacian distribution;

Cauchy's distribution can be expressed as:

$p\left(x\right)=\frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+{(x-\mathrm{\μ})}^2}$

Wherein λ is form parameter, and μ is location parameter, because in video compression coding, residual error coefficient is substantially symmetrical about 0, so make μ=0.Can also meet to keep the probability density finally represented by laplacian distribution and Cauchy's distributed combination so for Cauchy's distribution, a λ must be found to make the integrated value of its probability density in [-TL, TL] interval equal the integrated value of probability density in [-TL, TL] interval of Laplce, and still can maintain its heavy-tailed property.So just λ can be derived.

The integrated value of Laplacian probability density in [-TL, TL] is made to be P _l(TL), it can be calculated as follows:

$P_L\left(\mathrm{TL}\right)=P(\mathrm{TL}\≤x\≤\mathrm{TL})={\∫}_{-\mathrm{TL}}^{\mathrm{TL}}\frac{{\mathrm{\α}}_{b_k}}{2}\exp (-{\mathrm{\α}}_{b_k}|x\left|\right)\mathrm{dx}=C$

Wherein press ${\mathrm{\α}}_{b_k}=\sqrt{2/{\mathrm{\σ}}_{b_k}^2},{\mathrm{\σ}}_{b_k}^2=E\left({\left|C_{R_{\mathrm{XY}}}^{b_k}\right|}^2\right)-E{\left(\right|C_{R_{\mathrm{XY}}}^{b_k}\left|\right)}^2$ Calculate.

Make P _c(LH) represent the integrated value of Cauchy probability density in [-TL, TL], be calculated as follows:

$\begin{array}{c}{P}_C\left(\mathrm{TL}\right)=P(-\mathrm{TL}\≤x\≤\mathrm{TL})={\∫}_{-\mathrm{TL}}^{\mathrm{TL}}\frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+x^2}\mathrm{dx}\\ =\frac{2}{\mathrm{\π}}{\tan }^{-1}\left(\frac{\mathrm{TL}}{\mathrm{\λ}}\right)=P_L\left(\mathrm{TL}\right)=C\end{array}$

So just can obtain: $\mathrm{\λ}=\frac{\mathrm{TL}}{\tan (\mathrm{\πC}/2)}.$

Utilize following formula to calculate and belong to S ₁the Laplacian parameter of residual error coefficient set:

${\mathrm{\α}}_1=\sqrt{\frac{2}{{\mathrm{\σ}}_1^2}}$

s ₁the variance of residual error coefficient set.

Finally can represent correlated noise model like this:

$p\left(C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\right)=\left\{\begin{array}{c}\frac{{\mathrm{\α}}_0}{2}\exp (-{\mathrm{\α}}_0|C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\left|\right),C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\∈S_1\\ \frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+{\left[C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\right]}^2},C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\∈S_2,S_3\end{array}\right.$

Claims (3)

1. a correlated noise model modelling approach in distributed video coding, it is characterized in that, the method includes the steps of:

1) for current decoded sub-band b _k, the absolute value of the distance of each coefficient and it and this subband residual mean square (RMS) forms 2 dimensional feature vectors, i.e. subband b _kin the n-th residual error characteristic vector be expressed as wherein $D_{b_k}(\mathrm{\μ},v)=C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},v)-E\left(C_{R_{\mathrm{XY}}}^{b_k}\right);$

2) K-Medoids clustering algorithm is utilized residual error characteristic vector to be divided into large residual error class and little residual error class;

3) use the laplacian distribution improved to describe little residual error class, use Cauchy to distribute and describe large residual error class; Calculate corresponding distributed constant respectively, finally obtain mixed distribution formula correlated noise model.

2. bit rate control method in a kind of distributed video coding according to claim 1, is characterized in that, described step 2) in, carry out K-Medoids cluster in accordance with the following steps:

1) initialization cluster centre: select three the residual error characteristic vectors started as initial cluster center corresponding class is make k=0;

2) sample clustering: by set of eigenvectors to be sorted the a certain class in three classes is allocated to one by one, if that is: by minimal distance principle (i=1,2 ..., N), j=(1,2,3), then in formula represent and class center distance, superscript k represents iterations, and distance here selects Euclidean distance, so produce new cluster

3) cluster centre is recalculated: different from K-Means, K-Means selects the mean value of all data points in current cluster to be new central point, and in K-Medoids, the minimum point of the distance sum of (in current cluster) point to other are all will be chosen as central point from current cluster;

4) end condition is judged: if (j=1,2,3), then terminate, and obtains 3 clusters, and DCT residual error coefficient is divided into three S set ₁, S ₂, S ₃, otherwise k=k+1, goes to 2).

3. in a kind of distributed video coding according to claim 1 and 2, side information is improved one's methods, and it is characterized in that, step 3) in, the idiographic flow calculating the parameter of laplacian distribution and the parameter of Cauchy's distribution is:

After cluster, calculate three residual error coefficient S set _jthe variance that (j=1,2,3) are respective, then by order from small to large, is designated as S respectively by the residual error coefficient set of its correspondence ₁, S ₂, S ₃, wherein S ₁be designated as little coefficient set, and S ₂, S ₃then be designated as large coefficient set, S ₁substantially symmetrical about 0, in order to calculate the convenience of the parameter of Cauchy's distribution, to S ₁revise, note S ₁the upper bound and the minimum value of absolute value of lower bound be TL, large residual error coefficient collection will be described with Cauchy's distribution afterwards, and describe little residual error coefficient collection by laplacian distribution;

Cauchy's distribution can be expressed as:

$p\left(x\right)=\frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+{(x-\mathrm{\μ})}^2}$

Wherein λ is form parameter, and μ is location parameter, because in video compression coding, residual error coefficient is substantially symmetrical about 0, so make μ=0; Can also meet to keep the probability density finally represented by laplacian distribution and Cauchy's distributed combination so for Cauchy's distribution, a λ must be found to make the integrated value of its probability density in [-TL, TL] interval equal the integrated value of probability density in [-TL, TL] interval of Laplce, and still can maintain its heavy-tailed property; So just λ can be derived;

The integrated value of Laplacian probability density in [-TL, TL] is made to be P _l(TL), it can be calculated as follows:

$P_L\left(\mathrm{TL}\right)=P(-\mathrm{TL}\≤x\≤\mathrm{TL})={\∫}_{-\mathrm{TL}}^{\mathrm{TL}}\frac{{\mathrm{\α}}_{b_k}}{2}\exp \left({-\mathrm{\α}}_{b_k}\right|x\left|\right)\mathrm{dx}=C$

Wherein press ${\mathrm{\α}}_{b_k}=\sqrt{2/{\mathrm{\σ}}_{b_k}^2},{\mathrm{\σ}}_{b_k}^2=E\left({\left|C_{R_{\mathrm{XY}}}^{b_k}\right|}^2\right)-E{\left(\right|C_{R_{\mathrm{XY}}}^{b_k}\left|\right)}^2$ Calculate;

Make P _c(LH) represent the integrated value of Cauchy probability density in [-TL, TL], be calculated as follows:

$\begin{array}{c}{P}_C\left(\mathrm{TL}\right)=P(-\mathrm{TL}\≤x\≤\mathrm{TL})={\∫}_{-\mathrm{TL}}^{\mathrm{TL}}\frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+x^2}\mathrm{dx}\\ =\frac{2}{\mathrm{\π}}{\tan }^{-1}\left(\frac{\mathrm{TL}}{\mathrm{\λ}}\right)=P_L\left(\mathrm{TL}\right)=C\end{array}$

So just can obtain: $\mathrm{\λ}=\frac{\mathrm{TL}}{\tan (\mathrm{\πC}/2)};$

Utilize following formula to calculate and belong to S ₁the Laplacian parameter of residual error coefficient set:

${\mathrm{\α}}_1=\sqrt{\frac{2}{{\mathrm{\σ}}_1^2}}$

s ₁the variance of residual error coefficient set.

Final expression correlated noise model:

$p\left(C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\right)=\left\{\begin{array}{c}\frac{{\mathrm{\α}}_0}{2}\exp \left({-\mathrm{\α}}_0\right|C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\left|\right),C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\∈S_1\\ \frac{1}{\mathrm{\π}}\frac{\mathrm{\λ}}{{\mathrm{\λ}}^2+{\left[C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\right]}^2},C_{R_{\mathrm{XY}}}^{b_k}(\mathrm{\μ},\mathrm{\υ})\∈S_2,S_3\end{array}\right.$