CN101674472A - Multistage code rate control method of video code with a plurality of visual points - Google Patents

Abstract

The invention discloses a multistage code rate control method of a video code with a plurality of visual points. The method firstly distributes the code rate reasonably between the visual points according to a relativity function, then realizes the bit distribution of a frame layer according to a sublayer B image bit distribution strategy and finally realizes the code rate control of a macro blocklayer according to a Laplacian rate distortion model. The method comprises the following steps: (1) establishing a frame of coding a plurality of visual points simultaneously; (2) carrying out GGOP layer bit distribution and code rate control; (3) carrying out GOP layer bit distribution and code rate control; (4) carrying out Frame layer bit distribution and code rate control; and (5) carrying out Macroblock layer bit distribution and code rate control and the like; when judging that the current GGOP is the last GGOP, finishing the entire code rate control process. The method adopts the Laplacian rate distortion model, has lower complexity and smaller calculation quantity and also has the decoding image PSNR gain of about 0.1dB averagely, and the code rate control error is within 1%; reasonable bit distribution is carried out between the visual points, and an image coded by the visual points has better equality and actual operability.

Description

The multistage code rate control method of multiple view video coding

Technical field

The invention belongs to technical field of video coding, what relate to is a kind of bit rate control method of video coding, particularly a kind of multistage code rate control method towards multiple view video coding.

Background technology

In recent years, with the third dimension of uniqueness, feeling of immersion and roaming characteristic is the concern that the multi-view point video technology of feature just more and more is subjected to academia and industrial quarters, multi-view point video can be given a kind of strong sense of reality of people than the single view video world that more can truly reflect reality.But the multi-view point video data volume is huge, and storage and transmission are very difficult, must compress efficiently.Multi-view point video is delivered to after compression to transmit then on the channel at coding side and is decoded in decoding end, store or show according to user's request, in the whole flow process of coding → transmission → decoding, how existence solves the encoder bit rate control problem of the multi-view point video data adaptive channel after the compression, code stream if this problem can not get solving after encoding is directly delivered to channel, can cause channel or congested, or idle, and this will reduce the utilance of channel greatly.

At present, many mature technique have been arranged at the Rate Control of single view video coding, as: the TM5 among the Mpeg2, the VM8 among the Mpeg4, people such as TMN8, Z.He in H.263 propose based on the bit rate control method in ρ territory and JVT-G012 H.264.But these technology can not directly apply to multiple view video coding, and its reason is that multi-view point video not only has more viewpoint, and have more complicated coded prediction structure and more B image type.Therefore, how realize in multiple view video coding that Rate Control is one of difficult problem of facing of present multi-view point video technology.

The Rate Control of single view video coding, with regard to its essence is Rate Control to I frame and P frame, the B frame is not carried out Rate Control, B frame QP value is just simply definite according to the value of the QP of adjacent I frame and P frame or adjacent two P frames, when the B frame number between adjacent I frame and P frame or adjacent two the P frames was big, it is very poor that the precision of Rate Control will become.Multiple view video coding be provided with more B image, so the Rate Control of multiple view video coding will increase the Rate Control to the B frame for improving code efficiency.The bit rate control method of existing multiple view video coding is based on secondary model more, structural relation according to parallax prediction and motion prediction, all images is divided into polytype coded frame, then to the modeling respectively of various types of images, according to the target bit and the frame level quantized parameter of the calculation of parameter all kinds frame of model.This class methods control precision is not high, and complexity is higher, does not have actual operability.

Summary of the invention

The problem and shortage that exists of prior art in view of the above the object of the present invention is to provide a kind of multistage code rate control method of multiple view video coding.This method can not only accurately be controlled code check, can also improve the coding quality of image.

For achieving the above object, design of the present invention is:

At first,, realize the Rate Control of macroblock layer at last according to Laplce's rate-distortion model then according to the Bit Allocation in Discrete of layering B video bits allocation strategy achieve frame layer according to relevance function reasonable distribution code check between viewpoint.

According to above-mentioned design, technical scheme of the present invention is:

A kind of multistage code rate control method of multiple view video coding, its concrete steps are:

(1), set up the framework of a plurality of viewpoint continuous programming codes, realize the continuous programming code of a plurality of viewpoints;

(2), GGOP (the group of group of pictures) layer Bit Allocation in Discrete and Rate Control, obtain the target bit of current GGOP;

(3), GOP (group of pictures) layer Bit Allocation in Discrete and Rate Control, calculate the weight factor w of each viewpoint GOP _k, obtain the target bit of current GOP;

(4), Frame layer Bit Allocation in Discrete and Rate Control, obtain the bit number that current encoded frame is assigned to;

(5), Macroblock layer Bit Allocation in Discrete and Rate Control, determine the quantization parameter Q of current macro _Pc

(6), the quantization parameter Q that calculates according to step (5) _PcCurrent macro is encoded;

(7), judge that whether macro blocks all in the present frame all encodes, if all encode, then changes step (8); If all encode, then repeat to change step (5) to (6), up to all encoding, change step (8);

(8), judge that whether frames all in the current GOP all encodes, if all encode, then changes step (9); If all do not encode, then repeat to change step (4) to (7), up to all compiling all frames of current GOP.

(9), judge that whether GOP all in the current GGOP all encodes, if all encode, then changes step (10); If all do not encode, then repeat to change step (3) to (8), up to all compiling all GOP of current GGOP.

(10), judge whether current GGOP is last GGOP, if last GGOP, then whole Rate Control process finishes, otherwise, repeat to change step (2) to (9)

The multistage code rate control method of multiple view video coding of the present invention compared with the prior art, have following conspicuous substantive outstanding feature and remarkable advantage: this method has increased the Rate Control to the B frame, adopted simple relatively Laplce's rate-distortion model, complexity is lower, amount of calculation is less, the error of Rate Control can be controlled within 1%, and the decoded picture PSNR gain about average 0.1dB is arranged; This method is owing to carried out reasonable bit distribution between viewpoint, the quality balance of its viewpoint coded image is better, and can expand the viewpoint number according to user's request in the practical application, has actual operation, is convenient to use realize.

Description of drawings

Fig. 1 is the FB(flow block) of the multistage code rate control method of multiple view video coding of the present invention.

Fig. 2 is the schematic diagram of the framework of a plurality of viewpoint continuous programming codes among Fig. 1.

Fig. 3 is the GGOP layer Bit Allocation in Discrete among Fig. 1 and the flow chart of Rate Control.

Fig. 4 is the GOP layer Bit Allocation in Discrete among Fig. 1 and the flow chart of Rate Control.

Fig. 5 is the Frame layer Bit Allocation in Discrete among Fig. 1 and the flow chart of Rate Control.

Fig. 6 is the Macroblock layer Bit Allocation in Discrete among Fig. 1 and the flow chart of Rate Control.

Embodiment

Below in conjunction with accompanying drawing embodiments of the invention are described in further detail.Present embodiment is to implement under the prerequisite with technical scheme of the present invention, provided detailed execution mode, but protection scope of the present invention is not limited to following embodiment.

Referring to Fig. 1, the multistage code rate control method of a kind of multiple view video coding of the present invention, at first set up the framework of a plurality of viewpoint continuous programming codes, carry out the Bit Allocation in Discrete and the Rate Control of GGOP layer, GOP layer, Frame layer, Macroblock layer then successively, determine the quantization parameter Qpc of current macro, current macro is carried out the RDO coding, constantly repeat said process until the Rate Control of finishing whole all frames of many viewpoints cycle tests with Qpc, the Rate Control process finishes, and the steps include:

(1), sets up the framework of a plurality of viewpoint continuous programming codes, realize that the multiple view video coding reference model JMVM (Joint Multiview Video Model) that the continuous programming code of a plurality of viewpoints: JVT (Joint Video Team) provides is that viewpoint is encoded one by one when realizing, that is to say in the time of to encode next viewpoint after the current view point end-of-encode and need carry out the configuration of parameter again, coding when obviously this reference model can't carry out a plurality of viewpoint.The code book invention is based on JMVM when realizing a plurality of viewpoint, utilize the predict of multiple view video coding, for example GGOP prediction, sequential prediction, gridiron pattern decompose prediction, and projected relationship, the for example inter prediction in the viewpoint, the parallax prediction between viewpoint etc., set up the framework of a plurality of viewpoint continuous programming codes, realize the continuous programming code of a plurality of viewpoints.Referring to Fig. 2, wherein horizontal express time is vertically represented viewpoint, and for example T0 represents the moment 0, and S0 represents viewpoint 0; Dash box is represented I frame, P frame or B frame, the coded sequence of this frame of numeral in the dash box, and among the figure, coded sequence is 8,9,10,11,12,13,14,15 8 two field pictures constitute a viewpoint GOP, i.e. 1 of viewpoint 0 GOP, 1 GOP of viewpoint 1 is 16,17,18,19 for coded sequence then, 20,21,22,23, other viewpoints GOP and the like, coded sequence is GGOP of 64 two field pictures formation of 8 to 71.This step is the basis of whole Rate Control, and later step all is to implement on the framework that this step is set up.

(2), GGOP layer Bit Allocation in Discrete and Rate Control, obtain the target bit of current GGOP, referring to Fig. 3, its concrete steps are:

(2-1), the initialization of multiple view video coding Rate Control: target setting code check at first, size of initialization virtual cache then and full scale;

(2-2), calculate the bit number that current GGOP is assigned to:

$T_{\mathrm{GGOP}}\left(i\right)=\frac{u\left(i\right)}{F_r}\×N_{\mathrm{view}}\×N\left(i\right)$

N wherein _ViewThe number of expression viewpoint, the size of N (i) expression GOP, F _rThe expression frame per second, u (i) represents target bit rate.

(2-3), consider the virtual cache occupancy behind the previous GGOP coding, then current GGOP actual allocated to bit number be:

$T_{\mathrm{GGOP}}\left(i\right)=\frac{u\left(i\right)}{F_r}\×N_{\mathrm{view}}\×N\left(i\right)+B_c(i-1)$

N wherein _ViewThe number of expression viewpoint, the size of N (i) expression GOP, B _c(i-1) the virtual cache occupancy behind i-1 GGOP coding of expression, F _rThe expression frame per second, u (i) represents target bit rate.

(3), GOP layer Bit Allocation in Discrete and Rate Control, calculate the weight factor w of each viewpoint GOP _k, obtain the target bit of current GOP, referring to Fig. 4, its concrete steps are:

(3-1), calculate the weight factor w of each viewpoint _k:

$C(S_k,S_0)=\frac{\underset{i=0}{\overset{I-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{J-1}{\mathrm{\Σ}}}\left|\right[A(i,j)-\stackrel{\‾}{A}]\×[B(i,j)-\stackrel{\‾}{B}\left]\right|}{\sqrt{\underset{i=0}{\overset{I-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{J-1}{\mathrm{\Σ}}}{[A(i,j)-\stackrel{\‾}{A}]}^2}\×\sqrt{\underset{i=0}{\overset{I-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{J-1}{\mathrm{\Σ}}}{[B(i,j)-\stackrel{\‾}{B}]}^2}}$

γ _k＝λ _k×C(S _k，S ₀)

$w_k=\frac{{\mathrm{\γ}}_k}{\underset{m=0}{\overset{N_{\mathrm{view}}-1}{\mathrm{\Σ}}}{\mathrm{\γ}}_m}$

Wherein A (i, j), (i j) represents viewpoint S respectively to B _kAnd S ₀(i, pixel value j), I, J be the height, wide of representative image respectively, and A, B represent A, B image averaging pixel value, C (S respectively in the position at the image A of synchronization correspondence and B _k, S ₀) correlation factors of expression viewpoint k and viewpoint 0, γ _kThe composite factor of expression expression viewpoint k, λ _kThe factor, w are adjusted in expression _kThe weight factor of expression viewpoint k, i.e. significance level.

(3-2), calculate the bit number that current GOP is assigned to:

T′ _GOP(k)＝T _GGOP(i)×w _k

Wherein, T _GGOP(i) target bit of i GGOP of expression (current GGOP), w _kThe weight factor of expression viewpoint k, i.e. significance level.

(3-3), count T according to the remaining bits of a last viewpoint GOP ^* _GOP(k-1), total bit number of being assigned to of then current GOP is:

T _GOP(k)＝T′ _GOP(k)+T ^* _GOP(k-1)

Wherein, T _GOP(k) represent total bit number that current GOP actual allocated arrives, T ' _GOP(k) represent the bit number that current GOP is assigned to, T ^* _GOP(k-1) the remaining bits number of the last viewpoint GOP of expression.

(4), Frame layer Bit Allocation in Discrete and Rate Control, obtain the bit number that current encoded frame is assigned to, referring to Fig. 5, its concrete steps are:

(4-1), determine the zoom factor of current GOP:

$S_f\left[k\right]=S_f[k-1]\&times;\sqrt{3/2}\&times;N_{\mathrm{Bi}}(k,i)+\sqrt{2}\&times;({N^0}_{\mathrm{Uni}}(k,i)+{{\mathrm{<figure-callout\; id="1"\; label="N"\; filenames="A2009101965320014H1.png"\; state="\{\{state\}\}">N</figure-callout>}}^1}_{\mathrm{Uni}}(k,i))/2$

$X_{\mathrm{tot}}=X^I+N_P{X}^P+\underset{k}{\mathrm{\&Sigma;}}{N}_B\left(k\right)X^B\left(k\right)$

${\mathrm{\&omega;}}^{I/P}=\frac{X^{I/P}}{S_f[-1]\&times;X_{\mathrm{tot}}}$

${\mathrm{\&omega;}}_k=\frac{X^B\left(k\right)}{S_f\left[k\right]\&times;X_{\mathrm{tot}}}$

S wherein _fThe weight of k layering B image of [k] expression, S _fThe weight of [1] expression I image or P image, its initial value is 1, N _Bi(k, i) expression bi-directional association sample point number, N ⁰ _Uni(k, i) or N ¹ _Uni(k, i) the unidirectional related sample point number of expression, ω ^I/PThe zoom factor of expression I image or P image, ω _kThe zoom factor of representing k layering B image, N _B(k) the B frame number of k time horizon of expression, X presentation video complexity is obtained by the actual coding bit number and the quantization parameter of image.Above-mentioned various in the scope of k from 0 to d-1, d represents total layering number.

(4-2), determine that according to the residue B frame number of remaining bits number, each layering and the zoom factor of current GOP the bit number that present frame is assigned to is:

${T^B}_{l,i}=\frac{{\mathrm{\&omega;}}_l{B}_l\left(i\right)}{\underset{k=l}{\overset{d-1}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_k{N^i}_B\left(k\right)}$

$T^{I/P}=\frac{{\mathrm{\&omega;}}^{I/P}{B}_{\mathrm{tot}}}{{\mathrm{\&omega;}}^{I/P}+\underset{k=0}{\overset{d-1}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_k{N}_B\left(k\right)}$

Wherein, T ^B _{L, i}Represent the bit number that i B image of l layering of current GOP is assigned to, T ^I/PThe bit number that expression I image or P image are assigned to, B _l(i) the remaining bits number of expression when i B image of coding l layering of current GOP, N ⁱ _B(k) k still uncoded B picture number of layering of expression, B _TotRepresent total bit number that current GOP is assigned to, ω ^I/PAnd ω _kRepresent the zoom factor of I image or P image and the zoom factor of k layering B image respectively.

(5), Macroblock layer Bit Allocation in Discrete and Rate Control, finally determine the quantization parameter Q of current macro _Pc, referring to Fig. 6, its concrete steps are:

(5-1), according to the target bit of the current encoded frame that obtains previously, determine that the target bit of current coding macro block is:

$T_{\mathrm{mac}}=\frac{R_{\mathrm{FRM}}}{256\&times;N_R}$

Wherein, T _MacRepresent every pixel bit number, R _FRMRemaining bits number after the intact previous macro block of presentation code, N _RRepresent remaining not coded macroblocks number.

(5-2), determine that according to Laplce's rate-distortion model the quantization parameter Q of current coding macro block is:

$R\left(D\right)={\log }_2\left(\frac{1}{\mathrm{\&lambda;D}}\right)={\log }_2\left(\frac{\mathrm{\&sigma;}}{\sqrt{2}\mathrm{\&rho;Q}}\right)$

$Q=\frac{\mathrm{\&sigma;}\&times;2^{-{\mathrm{<figure-callout\; id="2"\; label="T\; mac"\; filenames="A2009101965320014H1.png"\; state="\{\{state\}\}">T</figure-callout>}}_{\mathrm{mac}}}}{\sqrt{2}\mathrm{\&rho;}}$

Wherein, R represents code check, and D represents distortion, and λ represents distributed constant, and σ represents the standard deviation of initial data, and ρ represents distortion parameter, and Q represents the quantization parameter of current macro, T _MacRepresent every pixel bit number.

(5-3), for keeping the harmony of picture quality, consider the quantization parameter Q of previous macro block _Pp, finally determine the quantization parameter Q of current macro _Pc:

${\tilde{Q}}_{\mathrm{pc}}=\min \{{\mathrm{<figure-callout\; id="2"\; label="Q\; pp\; +"\; filenames="A2009101965320014H1.png"\; state="\{\{state\}\}">Q</figure-callout>}}_{\mathrm{pp}}+2,\max \{Q_{\mathrm{pp}}-2,Q\left\}\right\}$

$Q_{\mathrm{pc}}=\min \{51,\max \{{\tilde{Q}}_{\mathrm{pc}},1\left\}\right\}$

Q _PpThe quantization parameter of representing previous macro block, Q _PcThe quantization parameter of the final current macro of determining of expression.

(6), use Q _PcCurrent macro is carried out the RDO coding.

(7), judge that whether macro blocks all in the present frame all encodes, if all encode, then changes step (8); If all encode, then repeat to change step (5) to (6), up to all encoding, change step (8);

(8), judge that whether frames all in the current GOP all encodes, if all encode, then changes step (9); If all do not encode, then repeat to change step (4) to (7), up to all compiling all frames of current GOP.

(9), judge that whether GOP all in the current GGOP all encodes, if all encode, then changes step (10); If all do not encode, then repeat to change step (3) to (8), up to all compiling all GOP of current GGOP.

(10), judge whether current GGOP is last GGOP, if last GGOP, then whole Rate Control process finishes, otherwise, repeat to change step (2) to (9).

Below test experiments is carried out in several multi-view point video sequential codings, on the multiple view video coding system platform that JVT provides, with JMVM adopt fixed quantisation value 28 respectively and the code check that obtained in 32 o'clock as the target bit rate of this bit rate control method, table 1 provides the test data parameter, table 2 provides test condition, and table 3 provides result of the present invention.

Cycle tests:

The Ballroom that MERL, KDDI provide, Exit, Race1, many viewpoints of Flamenco1 cycle tests test data parameter:

Table 1 test data parameter

Cycle tests	High * wide	The camera spacing	Viewpoint
				??Ballroom	??640×480	??20cm	??0～7
??Exit	??640×480	??20cm	??0～7
				??Race1	??640×480	??20cm	??0～7
??Flamenco1	??320×240	??20cm	??0～7

Test condition:

Table 2 test condition

Frame per second	??25	??ICMode	??0
				The GOP size	??8	Way of search	??FastSearch
The hunting zone	??64pixel	The coding frame number	??81
				Coding mode	??CABAC	Reference frame number	??2
??Base?Unit?Size	??1	Channel type	??CBR

Test result:

Table 3 test result

As shown in Table 3: bit rate control method of the present invention can also guarantee less code check deviation when having less code check deviation fluctuation range; Bit rate control method of the present invention has also obtained the decoded picture PSNR gain about average 0.1dB; The multistage code rate control method of multiple view video coding of the present invention can not only more accurately be controlled code check, and can also keep even slightly improve the coding quality of image.

Claims (6)

1, a kind of multistage code rate control method of multiple view video coding, it is characterized in that, at first according to relevance function reasonable distribution code check between viewpoint, then according to the Bit Allocation in Discrete of layering B video bits allocation strategy achieve frame layer, realize the Rate Control of macroblock layer at last according to Laplce's rate-distortion model, its step is as follows:

(1), set up the framework of a plurality of viewpoint continuous programming codes, realize the continuous programming code of a plurality of viewpoints;

(2), GGOP layer Bit Allocation in Discrete and Rate Control, obtain the target bit of current GGOP;

(3), GOP layer Bit Allocation in Discrete and Rate Control, calculate the weight factor w of each viewpoint GOP _k, obtain the target bit of current GOP;

(4), Frame layer Bit Allocation in Discrete and Rate Control, obtain the bit number that current encoded frame is assigned to;

(5), Macroblock layer Bit Allocation in Discrete and Rate Control, determine the quantization parameter Q of current macro _Pc

(6), the quantization parameter Q that calculates according to step (5) _PcCurrent macro is encoded.

(7), judge that whether macro blocks all in the present frame all encodes, if all encode, then changes step (8); If all encode, then repeat to change step (5) to (6), up to all encoding, change step (8);

(8), judge that whether frames all in the current GOP all encodes, if all encode, then changes step (9); If all do not encode, then repeat to change step (4) to (7), up to all compiling all frames of current GOP.

(9), judge that whether GOP all in the current GGOP all encodes, if all encode, then changes step (10); If all do not encode, then repeat to change step (3) to (8), up to all compiling all GOP of current GGOP.

(10), judge whether current GGOP is last GGOP, if last GGOP, then whole Rate Control process finishes, otherwise, repeat to change step (2) to (9)

2, the multistage code rate control method of multiple view video coding according to claim 1, it is characterized in that, the framework of setting up a plurality of viewpoint continuous programming codes described in the above-mentioned steps (1) is meant that with JVT reference software JMVM be basis coding when utilizing the predict of multiple view video coding and projected relationship to realize a plurality of viewpoint.

3, according to the multistage code rate control method of claim 2 described multiple view video codings, it is characterized in that, GGOP layer Bit Allocation in Discrete and the Rate Control described in the above-mentioned steps (2), its concrete steps are:

(2-1), the initialization of multiple view video coding Rate Control: target setting code check at first, size of initialization virtual cache then and full scale;

(2-2), calculate the bit number that current GGOP is assigned to:

$T_{\mathrm{GGOP}}\left(i\right)=\frac{u\left(i\right)}{F_r}\&times;N_{\mathrm{view}}\&times;N\left(i\right)$

N wherein _ViewThe number of expression viewpoint, the size of N (i) expression GOP, F _rThe expression frame per second, u (i) represents target bit rate.

(2-3), consider the virtual cache occupancy behind the previous GGOP coding, then current GGOP actual allocated to bit number be:

$T_{\mathrm{GGOP}}\left(i\right)=\frac{u\left(i\right)}{F_r}\&times;N_{\mathrm{view}}\&times;N\left(i\right)+B_c(i-1)$

N wherein _ViewThe number of expression viewpoint, the size of N (i) expression GOP, B _c(i-1) the virtual cache occupancy behind i-1 GGOP coding of expression, F _rThe expression frame per second, u (i) represents target bit rate.

4, the multistage code rate control method of multiple view video coding according to claim 1 is characterized in that, GOP layer Bit Allocation in Discrete and the Rate Control described in the above-mentioned steps (3), and its concrete steps are:

(3-1), calculate the weight factor w of each viewpoint _k:

$C(S_k,S_0)=\frac{\underset{i=0}{\overset{I-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{J-1}{\mathrm{\&Sigma;}}}\left|\right[A(i,j)-\stackrel{\&OverBar;}{A}]\&times;[B(i,j)-\stackrel{\&OverBar;}{B}]|}{\sqrt{\underset{i=0}{\overset{I-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{J-1}{\mathrm{\&Sigma;}}}{[A(i,j)-\stackrel{\&OverBar;}{A}]}^2}\&times;\sqrt{\underset{i=0}{\overset{I-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{J-1}{\mathrm{\&Sigma;}}}{[B(i,j)-\stackrel{\&OverBar;}{B}]}^2}}$

γ _k＝λ _k×C(S _k，S ₀)

$w_k=\frac{{\mathrm{\&gamma;}}_k}{\underset{m=0}{\overset{N_{\mathrm{view}}-1}{\mathrm{\&Sigma;}}}{\mathrm{\&gamma;}}_m}$

Wherein A (i, j), (i j) represents viewpoint S respectively to B _kAnd S ₀(i, pixel value j), I, J be the height, wide of representative image respectively, and A, B represent A, B image averaging pixel value, C (S respectively in the position at the image A of synchronization correspondence and B _k, S ₀) correlation factors of expression viewpoint k and viewpoint 0, γ _kThe composite factor of expression expression viewpoint k, λ _kThe factor, w are adjusted in expression _kThe weight factor of expression viewpoint k, i.e. significance level.

(3-2), calculate the bit number that current GOP is assigned to:

T′ _GOP(k)＝T _GGOP(i)×w _k

Wherein, T _GGOP(i) target bit of i GGOP of expression (current GGOP), w _kThe weight factor of expression viewpoint k, i.e. significance level.

(3-3), count T according to the remaining bits of a last viewpoint GOP ^* _GOP(k-1), then current GOP actual allocated to total bit number be:

T _GOP(k)＝T′ _GOP(k)+T ^* _GOP(k-1)

Wherein, T _GOP(k) represent total bit number that current GOP actual allocated arrives, T ' _GOP(k) represent the bit number that current GOP is assigned to, T ^* _GOP(k-1) the remaining bits number of the last viewpoint GOP of expression.

5, the multistage code rate control method of multiple view video coding according to claim 4 is characterized in that, Frame layer Bit Allocation in Discrete and the Rate Control described in the above-mentioned steps (4), and its concrete steps are:

(4-1), determine the zoom factor of current GOP:

$S_f\left[k\right]=S_f[k-1]\&times;\sqrt{3/2}\&times;N_{\mathrm{Bi}}(k,i)+\sqrt{2}\&times;({N^0}_{\mathrm{Uni}}(k,i)+{N^1}_{\mathrm{Uni}}(k,i))/2$

$X_{\mathrm{tot}}=X^I+N_P{X}^P+\underset{k}{\mathrm{\&Sigma;}}{N}_B\left(k\right)X^B\left(k\right)$

${\mathrm{\&omega;}}^{I/P}=\frac{X^{I/P}}{S_f[-1]\&times;X_{\mathrm{tot}}}$

${\mathrm{\&omega;}}_k=\frac{X^B\left(k\right)}{S_f\left[k\right]\&times;X_{\mathrm{tot}}}$

S wherein _fThe weight of k layering B image of [k] expression, S _fThe weight of [1] expression I image or P image, its initial value is 1, N _Bi(k, i) expression bi-directional association sample point number, N ⁰ _Uni(k, i) or N ¹ _Uni(k, i) the unidirectional related sample point number of expression, ω ^I/PThe zoom factor of expression I image or P image, ω _kThe zoom factor of representing k layering B image, N _B(k) the B frame number of k time horizon of expression, x presentation video complexity is obtained by the actual coding bit number and the quantization parameter of image.Above-mentioned various in the scope of k from 0 to d-1, d represents total layering number.

(4-2), determine that according to the residue B frame number of remaining bits number, each layering and the zoom factor of current GOP the bit number that present frame is assigned to is:

${T^B}_{l,i}=\frac{{\mathrm{\&omega;}}_l{B}_l\left(i\right)}{\underset{k=l}{\overset{d-1}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_k{N^i}_B\left(k\right)}$

$T^{I/P}=\frac{{\mathrm{\&omega;}}^{I/P}{B}_{\mathrm{tot}}}{{\mathrm{\&omega;}}^{I/P}+\underset{k=0}{\overset{d-1}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_k{N}_B\left(k\right)}$

Wherein, T ^B _{L, i}Represent the bit number that i B image of l layering of current GOP is assigned to, T ^I/PThe bit number that expression I image or P image are assigned to, B _i(i) the remaining bits number of expression when i B image of coding l layering of current GOP, N ⁱ _B(k) k still uncoded B picture number of layering of expression, B _TotRepresent total bit number that current GOP is assigned to, ω ^I/PAnd ω _kRepresent the zoom factor of I image or P image and the zoom factor of k layering B image respectively.

6, the multistage code rate control method of multiple view video coding according to claim 5 is characterized in that, Macroblock layer Bit Allocation in Discrete and the Rate Control described in the above-mentioned steps (5) finally determined the quantization parameter Q of current macro _Pc, its concrete steps are:

(5-1), according to the target bit of the current encoded frame that obtains previously, determine that the target bit of current coding macro block is:

$T_{\mathrm{mac}}=\frac{R_{\mathrm{FRM}}}{256\&times;N_R}$

Wherein, T _MacRepresent every pixel bit number, R _FRMRemaining bits number after the intact previous macro block of presentation code, N _RRepresent remaining not coded macroblocks number.

(5-2), determine that according to Laplce's rate-distortion model the quantization parameter Q of current coding macro block is:

$R\left(D\right)={\log }_2\left(\frac{1}{\mathrm{\&lambda;D}}\right)={\log }_2\left(\frac{\mathrm{\&sigma;}}{\sqrt{2}\mathrm{\&rho;Q}}\right)$

$Q=\frac{\mathrm{\&sigma;}\&times;2^{-T_{\mathrm{mac}}}}{\sqrt{2}\mathrm{\&rho;}}$

Wherein, R represents code check, and D represents distortion, and λ represents distributed constant, and σ represents the standard deviation of initial data, and ρ represents distortion parameter, and Q represents the quantization parameter of current macro, T _MacRepresent every pixel bit number.

(5-3), for keeping the harmony of picture quality, consider the quantization parameter Q of previous macro block _Pp, finally determine the quantization parameter Q of current macro _Pc:

${\tilde{Q}}_{\mathrm{pc}}=\min \{Q_{\mathrm{pp}}+2,\max \{Q_{\mathrm{pp}}-2,Q\left\}\right\}$

$Q_{\mathrm{pc}}=\min \{51,\max \{{\tilde{Q}}_{\mathrm{pc}},1\left\}\right\}$

Q _PpThe quantization parameter of representing previous macro block, Q _PcThe quantization parameter of the final current macro of determining of expression.

Images