CN102780885A - Rate distortion optimization method - Google Patents

Abstract

The invention relates to a rate distortion optimization method, which comprises the following steps: obtaining a block layer Lagrange parameter of each basic coding unit in a current frame and a frame layer Lagrange parameter of the current frame; acquiring a final Lagrange parameter of each basic coding unit in the current frame according to the Lagrange parameter of the block layer and the Lagrange parameter of the frame layer; according to the final Lagrange parameter, rate distortion optimization coding is carried out on each basic coding unit in the current frame to obtain a code rate value and a distortion value of each basic coding unit in the current frame, and then the code rate value and the distortion value of the current frame are obtained; and establishing a frame layer rate-distortion model according to the code rate values and the distortion values of at least two basic coding units in the current frame, and estimating the frame layer Lagrange parameters of the next frame according to the frame layer rate-distortion model and the code rate values and the distortion values of the current frame. The rate distortion optimization method realizes the optimal rate distortion optimization of video contents by integrating the rate distortion optimization of a frame layer and the rate distortion optimization of a basic coding unit layer.

Description

A kind of rate-distortion optimization method

Technical field

The present invention relates to image/video encoding and decoding field, more particularly, relate to a kind of rate-distortion optimization method.

Background technology

With international standard H264 is reference, when encoded in the basic coding unit, can select the different coding pattern.The selection of coding mode comprises the selection to infra-frame prediction mode or inter prediction mode; Also can comprise the selection of partitioning scheme to the basic coding unit (routine INTRA-4x4, INTRA-8x8, INTRA-16x16, SKIP, DIRECT, INTER-16x16, INTER-16x8, INTER-8x16, INTER-8x8, can further be divided into INTER-8x8, INTER-8x4, INTER-4x8, INTER-4x4), can also comprise selection predict blocks position (routine Intra_4x4_Vertical, Intra_4x4_Horizontal, Intra_4x4_Diagonal_Down_Left, Intra_4x4_Diagonal_Down_Right, Intra4x4_Vertical_Right, Intra_4x4_Hori_zontal_Down, Intra_4x4_Vertical_Left, Intra_4x4_Horizontal_Up, Intra_4x4_DC) to INTER-8x8.Confirming of coding mode is that the percent of pass aberration optimizing realizes that wherein rate-distortion optimization is the process that following cost function J is minimized,

J(s,c,mode|QP)＝D(s,c,mode|QP)+λ _modeR(s,c,mode|QP)，

Wherein D is a distortion value; R is the code check value; S and c represent former figure and the corresponding basic coding unit of handling through encoding and decoding of building image again respectively; Mode representes the coding mode selected of basic coding unit, and QP is a quantization parameter, and λ mode is used for the Lagrangian parameter of compromise distortion value and code check value.

Rate-distortion optimization is under the fixed condition of quantization parameter QP, confirms the mode that can make above-mentioned cost function J minimum.In the H264 standard, λ mode is promptly had by quantization parameter QP decision

λ _mode＝0.85×2 ^(QP-12)/3，

Above-mentioned rate-distortion optimization mode is a statistical approximation result, lacks the adaptive ability to video content, does not have the continuity of content between considered frame, the rate distortion between successive frame is concerned the restriction that has no.So might cause the rate distortion between successive frame uneven, cause distortion or code check between successive frame to rise and fall excessive.

So, be necessary to provide a kind of rate-distortion optimization method, to solve the existing in prior technology problem.

Summary of the invention

The technical problem that the present invention will solve is; Can not carry out the defective of adaptive optimization to video content to rate-distortion optimization method of the prior art; Provide a kind of, realize rate-distortion optimization method the optimal ratio aberration optimizing of video content through the rate-distortion optimization of comprehensive frame layer and the rate-distortion optimization of basic coding elementary layer.

The technical solution adopted for the present invention to solve the technical problems is: the present invention relates to a kind of rate-distortion optimization method, it comprises step:

The frame layer Lagrange parameter of A, the piece layer Lagrange parameter of obtaining each basic coding unit in the present frame and present frame;

B, according to the frame layer Lagrange parameter of the piece layer of said basic coding unit Lagrange parameter and said present frame, obtain the final Lagrangian parameter of each basic coding unit in the said present frame;

C, based on said final Lagrangian parameter; Each basic coding unit in the said present frame is carried out the rate-distortion optimization coding; Obtain each basic coding unit corresponding code rate value and distortion value in the said present frame, and then obtain present frame corresponding code rate value and distortion value;

D, based at least two basic coding unit corresponding code rate values in the said present frame and distortion value; Set up frame layer rate-distortion model; Based on said frame layer rate-distortion model and said present frame corresponding code rate value and distortion value, the frame layer Lagrange parameter of estimation next frame;

Wherein said step B is specially: obtain said final Lagrangian parameter through following formula,

λ _mode,n＝w ₁λ _B,n+w ₂λ _F,0+w ₃，

λ wherein _{Mode, n}Be the final Lagrangian parameter of n basic coding unit, λ _{F, 0}Be the frame layer Lagrange parameter of present frame, λ _{B, n}Be the piece layer Lagrange parameter of n basic coding unit, w ₁, w ₂, w ₃Be corresponding weight coefficient, said weight coefficient w ₁, w ₂, w ₃Obtain according to the coding result of said present frame and the coding result of nearest M frame, M is a positive integer.

In rate-distortion optimization method of the present invention, obtain said weight coefficient w through following steps ₁, w ₂, w ₃:

Store the frame layer Lagrange parameter lambda of said present frame _{F, 0}, the nearest frame layer Lagrange parameter lambda of M frame _{F, m}And intermediate parameters w _{23, m}, w wherein _{23, m}Obtain through following formula:

$w_{23,m}=\frac{\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

X wherein _nBe the piece layer Lagrange parameter of n basic coding unit of m frame, y _nBe the final Lagrangian parameter of n basic coding unit of m frame:

x _n＝λ _B,n，

y _n=λ _mode,n，

Piece layer Lagrange parameter x according to n basic coding unit of said m frame _nFinal Lagrangian parameter y with n basic coding unit of said m frame _n, calculate w ₁:

$w_1=\frac{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

Frame layer Lagrange parameter lambda according to present frame _{F, 0}, the nearest frame layer Lagrange parameter lambda of M frame _{F, m}With said intermediate parameters w _{23, m}, calculate w ₂And w ₃,

u _m＝λ _F,m，

v _m＝w _23,m，

$w_2=\frac{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}\left(u_m{v}_m\right)-\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{v}_m}{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2-{\left(\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\right)}^2},$

$w_3=\frac{\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{v}_m-\underset{n=0}{\overset{N}{\mathrm{\Σ}}}\left(u_m{v}_m\right)\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m}{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2-{\left(\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\right)}^2},$

Wherein, m is more than or equal to 0, smaller or equal to the integer of M, and N is the number of the said basic coding unit selected.

In rate-distortion optimization method of the present invention, the piece layer of said each basic coding unit Lagrange parameter is:

λ _B＝0.85×2 ^(QP-12)/3，

λ wherein _BBe the piece layer Lagrange parameter of said each basic coding unit, QP is the quantization parameter of said basic coding unit.

In rate-distortion optimization method of the present invention, said step C comprises: obtain the minimum code cost of rate-distortion optimization of the said basic coding unit of said present frame through following formula,

J(s,c,mode|QP)＝min[D(s,c,mode|QP)+λ _modeR(s,c,mode|QP)]，

Wherein J is the minimum code cost; D is a distortion value, and R is the code check value, and s representes the basic coding unit of former figure; C representes the corresponding basic coding unit of building image again through the encoding and decoding processing; Mode representes the coding mode selected of said basic coding unit, and QP is the quantization parameter of said basic coding unit, λ _ModeBe corresponding final Lagrangian parameter;

According to said minimum code cost, obtain corresponding forced coding pattern, the said basic coding unit of said present frame is carried out the rate-distortion optimization coding.

In rate-distortion optimization method of the present invention, said step D comprises: according to the code check value and the distortion value of at least two said basic coding unit of said present frame, adopt following formula to set up said frame layer rate-distortion model,

R＝aln(b/D)，

Wherein R is the code check value, and D is corresponding distortion value, and a, b are model parameter.

In rate-distortion optimization method of the present invention, when the code check value and the distortion value that use two said basic coding unit, when setting up said frame layer rate-distortion model, said model parameter obtains through following formula,

$a=\frac{R_1-R_2}{\ln D_2-\ln D_1},$

$b=e^{\frac{R_1\ln D_2-R_2\ln D_1}{R_1-R_2}},$

R wherein ₁And R ₂Be the code check value of said two basic coding unit, D ₁And D ₂Be corresponding distortion value.

In rate-distortion optimization method of the present invention, when the code check value and the distortion value that use N said basic coding unit, when setting up said frame layer rate-distortion model, said model parameter obtains through following formula,

a＝-p，

b＝e ^-q/p，

$p=\frac{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

$q=\frac{\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

x _n＝lnD _n，y _n=R _n，

N＞2，

R wherein _nBe the code check value of said n basic coding unit, D _nBe corresponding distortion value.

In rate-distortion optimization method of the present invention, said step D also comprises: through the frame layer Lagrange parameter of following formula estimation next frame,

λ _F＝-(D(A ₁)-D(A ₂))/(R(A ₁)-R(A ₂))，

R (A wherein ₁) be the code check value of first intersection point, R (A ₂) be the code check value of second intersection point, D (A ₁) be the distortion value of first intersection point, D (A ₂) be the distortion value of second intersection point, said first intersection point is R=R in R～D plane _AWith said rate-distortion model the intersection point of definite curve, said second intersection point is D=D _AWith said rate-distortion model the intersection point of definite curve, R _ABe frame layer bit rate value, D _ABe frame layer distortion value.

In rate-distortion optimization method of the present invention, said frame layer bit rate value R _ABe the code check value sum of all basic coding unit in the said frame, said frame layer distortion value D _AThrough asking mean square error to obtain to the former figure of said frame and the image of building again of said frame.

The rate-distortion optimization method of embodiment of the present invention has following beneficial effect: through the rate-distortion optimization of comprehensive frame layer and the rate-distortion optimization of basic coding elementary layer, realize the optimal ratio aberration optimizing to video content.Avoided the rate-distortion optimization method of prior art can not carry out the technical problem of adaptive optimization to video content.

Description of drawings

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

Fig. 1 is the flow chart of first preferred embodiment of rate-distortion optimization method of the present invention;

Fig. 2 is the idiographic flow block diagram of first preferred embodiment of rate-distortion optimization method of the present invention;

Fig. 3 is the sketch map that obtains the frame layer Lagrange parameter of next frame in the rate-distortion optimization method of the present invention.

Embodiment

In order to make the object of the invention, technical scheme and advantage clearer,, the present invention is further elaborated below in conjunction with accompanying drawing and embodiment.Should be appreciated that specific embodiment described herein only in order to explanation the present invention, and be not used in qualification the present invention.

Please with reference to Fig. 1, Fig. 1 is the flow chart of first preferred embodiment of rate-distortion optimization method of the present invention.This rate-distortion optimization method starts from:

Step 101 is obtained the piece layer Lagrange parameter of each basic coding unit in the present frame and the frame layer Lagrange parameter of present frame;

Step 102 according to the piece layer Lagrange parameter of said basic coding unit and the frame layer Lagrange parameter of said present frame, is obtained the final Lagrangian parameter of each basic coding unit in the said present frame;

Step 103; According to said final Lagrangian parameter; Each said basic coding unit in the said present frame is carried out the rate-distortion optimization coding, obtain each basic coding unit corresponding code rate value and distortion value, and then obtain present frame corresponding code rate value and distortion value;

Step 104; According at least two basic coding unit corresponding code rate values in the present frame and distortion value; Set up frame layer rate-distortion model, according to said present frame corresponding code rate value and distortion value, and the frame layer rate-distortion model of present frame; The frame layer Lagrange parameter of estimation next frame, and used weight coefficient when calculating the final Lagrangian parameter of next frame.

Please see figures.1.and.2, Fig. 2 is the idiographic flow block diagram of first preferred embodiment of rate-distortion optimization method of the present invention.Pass through the practical implementation process of the detailed explanation rate-distortion optimization method of the present invention of Fig. 2 below.

At first; Before step 101, import an absolute coding part of video to be encoded, such as GOP (picture group; Group of Pictures) frame in is to code device; Wherein each frame can have the different coding mode in the GOP, and according to existing video encoding standard, coded system can be divided into I frame mode, P frame mode and B frame mode.Rate-distortion optimization method of the present invention can be applicable to I frame, P frame and B frame respectively, for the simplicity of explaining is no longer distinguished I frame, P frame and B frame here, and is to use uniform way to describe.

After image frame is input to code device, this image frame is divided into the basic coding unit, can be divided into frame the macro block of 16x16 pixel size simply by basic coding unit acquiring unit, with each macro block as a basic coding unit.Here do not get rid of the dividing method that adopts other, as adopting band as the basic coding unit.Then all coding modes selected of basic coding unit are selected one by one; Can select coding mode to each; Accomplish respective coding and decoding processing; Obtain the basic coding unit corresponding build image again, the processing method of Code And Decode can be carried out according to various video encoding standards.Such as, encoding process can be carried out following each step successively: confirm predict blocks, residual computations, dct transform (discrete cosine transform, Discrete Cosine Transform), quantification, entropy coding.Decoding processing then is the reverse operating to above-mentioned each step.The concrete grammar that rate-distortion optimization method of the present invention is handled this Code And Decode does not limit.

Come step 101 subsequently, in step 101, obtain the piece layer Lagrange parameter lambda of each basic coding unit in the present frame _BAnd the frame layer of present frame Lagrange parameter lambda _FWherein the piece layer of basic coding unit Lagrange parameter is:

λ _B＝0.85×2 ^(QP-12)/3，

λ wherein _BBe the piece layer Lagrange parameter of said basic coding unit, QP is the quantization parameter of said basic coding unit.It is pointed out that can there be different λ each basic coding unit _BWith the QP value, but, do not use λ here for interest of clarity with making any distinction between _BRepresent the piece layer Lagrange parameter and the quantization parameter of a basic coding unit with QP.The frame layer Lagrange parameter lambda of present frame _FCoding result by to previous frame confirms that the concrete grammar hereinafter is set forth in detail.If present frame is the initial frame of a coded system (being I frame, P frame or B frame mode), then desirable λ _FBe a designation number, for example 0.

Come step 102 subsequently, in step 102, according to the piece layer Lagrange parameter lambda of each basic coding unit in the present frame _{B, n}Frame layer Lagrange parameter lambda with present frame _F, obtain the final Lagrangian parameter lambda of each said basic coding unit _{Mode, n}Be specially the final Lagrangian parameter of obtaining each basic coding unit through following formula,

λ _mode,n＝w ₁λ _B,n+w ₂λ _F,0+w ₃，

λ wherein _{Mode, n}Be the final Lagrangian parameter of n basic coding unit, λ _FThe frame layer Lagrange parameter of present frame, λ _{B, n}Be the piece layer Lagrange parameter of n basic coding unit, w ₁, w ₂, w ₃Be corresponding weight coefficient, said weight coefficient w ₁, w ₂, w ₃Obtain according to the coding result of present frame and the coding result of nearest M frame, M is a positive integer, and concrete grammar is set forth below in detail.If present frame is an initial frame, then can be taken as designation number, for example w ₁Be 1, w ₂Be 0, w ₃Be 0 etc.

Come step 103 subsequently; In step 103; According to each basic coding unit final Lagrangian parameter separately, each basic coding unit in the present frame is carried out the rate-distortion optimization coding, obtain corresponding code rate value and distortion value; And further obtain present frame corresponding code rate value and distortion value, set up the frame layer rate-distortion model of present frame simultaneously.Specifically comprise:

Obtain the minimum code cost of the rate-distortion optimization of a basic coding unit in the said present frame through following formula,

J(s,c,mode|QP)＝min[D(s,c,mode|QP)+λ _modeR(s,c,mode|QP)]，

Wherein J is the minimum code cost; D is a distortion value, and R is the code check value, and s representes the basic coding unit of former figure; C representes the corresponding basic coding unit of building image again through the encoding and decoding processing; Mode representes the coding mode selected of said basic coding unit, and QP is the quantization parameter of said basic coding unit, λ _ModeBe corresponding final Lagrangian parameter;

Find the minimum code cost of rate-distortion optimization through following formula, can from optional coding mode, find the forced coding pattern of each basic coding unit like this, adopt this forced coding pattern to be encoded in this basic coding unit then.Above-mentioned coding has all been accomplished in all basic coding unit in a frame, and then the coding of this frame is accomplished.

Simultaneously, adopt above-mentioned forced coding pattern, can obtain this basic coding unit corresponding code rate value and distortion value the basic coding cell encoding.

The distortion value of above-mentioned basic coding unit can obtain through the MSE (mean square error, Mean Squared Error) that calculates former figure and build again between the corresponding basic coding unit of image.

Subsequently, according to above-mentioned coding, can obtain present frame corresponding code rate value and distortion value to each basic coding unit in the present frame.Wherein, the code check value is the code check value sum of all basic coding unit in the present frame, and distortion value is through to the former figure of entire frame with build image again and ask MSE to obtain.

Come step 104 subsequently, in step 104,, adopt following formula to set up the frame layer rate-distortion model of descriptor frame layer rate distortion relation according to the code check value R and the distortion value D of at least two basic coding unit in the present frame,

R＝aln(b/D)，

Wherein R representes the code check value, and D representes distortion value.A, b are model parameter.

Be specially, when the code check value of using two basic coding unit and distortion value were set up frame layer rate-distortion model, available following formula obtained the model parameter of frame layer rate-distortion model

$a=\frac{R_1-R_2}{\ln D_2-\ln D_1},$

$b=e^{\frac{R_1\ln D_2-R_2\ln D_1}{R_1-R_2}},$

R wherein ₁And R ₂Be said two basic coding unit corresponding code rate values, D ₁And D ₂Be the corresponding distortion value in said two basic coding unit.

When the code check value of using N basic coding unit and distortion value were set up frame layer rate-distortion model, available following formula obtained the model parameter of frame layer rate-distortion model

a＝-p，

b＝e ^-q/p，

$p=\frac{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

$q=\frac{\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

x _n＝lnD _n，y _n=R _n，

N＞2，

R wherein _nBe n (the individual basic coding of n=1～N) unit corresponding code rate value, D _nBe the corresponding distortion value in this basic coding unit.

According to the frame layer rate-distortion model of present frame, and present frame corresponding code rate value and distortion value, the frame layer Lagrange parameter of estimation next frame and the corresponding weight coefficient w of next frame ₁, w ₂, w ₃

As shown in Figure 3, Fig. 3 is the sketch map that obtains the frame layer Lagrange parameter of next frame in the rate-distortion optimization method of the present invention.Curve among Fig. 3 is the determined curve of said frame layer rate-distortion model, and wherein the A coordinate of ordering is present frame corresponding code rate value R _AWith distortion value D _ACoordinate in this coordinate system, as previously mentioned, code check value R _ABe the code check value sum of all basic coding unit in the present frame, distortion value D _AThrough to the former figure of entire frame with build image again and ask mean square error to obtain.What supposed that A orders is A1 point and A2 point with the intersection point of curve respectively with horizontal straight line vertically, and wherein the A1 coordinate of ordering is (R (A ₁), D (A ₁)), the coordinate that A2 is ordered is (R (A ₂), D (A ₂)), the straight slope that A1 point and A2 point constitute just can be used as the frame layer Lagrange parameter lambda of next frame _FBut also arrive the nearest point that A is ordered on the calculated curve here, curve promptly can be used as the frame layer Lagrange parameter lambda of next frame at the slope of this point _F

Particularly, can estimate the frame layer Lagrange parameter of next frame through following formula,

λ _F＝-(D(A ₁)-D(A ₂))/(R(A ₁)-R(A ₂))，

R (A wherein ₁) be the code check value of first intersection point, R (A ₂) be the code check value of second intersection point, D (A ₁) be the distortion value of first intersection point, D (A ₂) be the distortion value of second intersection point, first intersection point is R=R in R～D plane _AWith the frame layer rate-distortion model of above-mentioned present frame the intersection point of definite curve, second intersection point is D=D _AWith above-mentioned rate-distortion model the intersection point of definite curve, R _ABe above-mentioned present frame corresponding code rate value, D _ABe the corresponding distortion value of above-mentioned present frame.

Weight coefficient w ₁, w ₂, w ₃Computational methods following,

At first pass through the λ of N basic coding unit in the present frame _{B, n}And λ _{Mode, n}Obtain w ₁With intermediate variable w _{23, m},

$w_1=\frac{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

$w_{23,m}=\frac{\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

x _n＝λ _B,n，

y _n=λ _mode,n，

Subsequently, again through the Lagrangian parameter lambda of the frame layer of present frame _{F, 0}, the nearest frame layer Lagrange parameter lambda of M frame _{F, m}The frame layer Lagrange parameter w corresponding with every frame _{23, m}(m=0～M), calculate w ₂And w ₃, specific as follows:

u _m＝λ _F，m，

v _m＝w _23,m，

$w_2=\frac{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}\left(u_m{v}_m\right)-\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{v}_m}{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2-{\left(\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\right)}^2},$

$w_3=\frac{\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{v}_m-\underset{n=0}{\overset{N}{\mathrm{\Σ}}}\left(u_m{v}_m\right)\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m}{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2-{\left(\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\right)}^2},$

In the above calculating formula, λ _{Mode, n}Be the final Lagrangian parameter of n basic coding unit of present frame, λ _{F, m}Be the frame layer Lagrange parameter of m frame, λ _{F, 0}Be the frame layer Lagrange parameter of present frame, λ _{B, n}Be the piece layer Lagrange parameter of n basic coding unit in the frame, w ₁, w ₂, w ₃Be corresponding weight coefficient, N is the basic coding unit number from a frame, selected, and M is the quantity of the nearest frame selected.

The weight coefficient that will use in the time of can obtaining to calculate the final Lagrangian parameter of each basic coding unit of next frame through following formula like this.

In sum, rate-distortion optimization method of the present invention has guaranteed the continuity of the rate distortion characteristic between successive frame through the rate-distortion optimization of comprehensive frame layer and the rate-distortion optimization of basic coding elementary layer, realizes the optimal ratio aberration optimizing to video content.The simultaneity factor aberration optimizing can be realized according to the type (being I frame, P frame or B frame) of frame coding is independent separately; Rate-distortion optimization method of the present invention has well avoided the rate-distortion optimization method of prior art can not carry out the technical problem of adaptive optimization to video content.

The above is merely embodiments of the invention; Be not so limit claim of the present invention; Every equivalent structure transformation that utilizes specification of the present invention and accompanying drawing content to be done, or directly or indirectly be used in other relevant technical fields, all in like manner be included in the scope of patent protection of the present invention.

Claims (9)

1. a rate-distortion optimization method is characterized in that, comprises step:

The frame layer Lagrange parameter of A, the piece layer Lagrange parameter of obtaining each basic coding unit in the present frame and present frame;

B, according to the frame layer Lagrange parameter of the piece layer of said basic coding unit Lagrange parameter and said present frame, obtain the final Lagrangian parameter of each basic coding unit in the said present frame;

C, based on said final Lagrangian parameter; Each basic coding unit in the said present frame is carried out the rate-distortion optimization coding; Obtain each basic coding unit corresponding code rate value and distortion value in the said present frame, and then obtain present frame corresponding code rate value and distortion value;

D, based at least two basic coding unit corresponding code rate values in the said present frame and distortion value; Set up frame layer rate-distortion model; Based on said frame layer rate-distortion model and said present frame corresponding code rate value and distortion value, the frame layer Lagrange parameter of estimation next frame;

Wherein said step B is specially: obtain said final Lagrangian parameter through following formula,

λ _mode,n＝w ₁λ _B,n+w ₂λ _F,0+w ₃，

λ wherein _{Mode, n}Be the final Lagrangian parameter of n basic coding unit, λ _{F, 0}Be the frame layer Lagrange parameter of present frame, λ _{B, n}Be the piece layer Lagrange parameter of n basic coding unit, w ₁, w ₂, w ₃Be corresponding weight coefficient, said weight coefficient w ₁, w ₂, w ₃Obtain according to the coding result of said present frame and the coding result of nearest M frame, M is a positive integer.

2. rate-distortion optimization method according to claim 1 is characterized in that, obtains said weight coefficient w through following steps ₁, w ₂, w ₃:

Store the frame layer Lagrange parameter lambda of said present frame _{F, 0}, the nearest frame layer Lagrange parameter lambda of M frame _{F, m}And intermediate parameters w _{23, m}, w wherein _{23, m}Obtain through following formula:

$w_{23,m}=\frac{\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

X wherein _nBe the piece layer Lagrange parameter of n basic coding unit of m frame, y _nBe the final Lagrangian parameter of n basic coding unit of m frame:

x _n＝λ _B,n，

y _n=λ _mode,n，

Piece layer Lagrange parameter x according to n basic coding unit of said m frame _nFinal Lagrangian parameter y with n basic coding unit of said m frame _n, calculate w ₁:

$w_1=\frac{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

Frame layer Lagrange parameter lambda according to present frame _{F, 0}, the nearest frame layer Lagrange parameter lambda of M frame _{F, m}With said intermediate parameters w _{23, m}, calculate w ₂And w ₃,

u _m＝λ _F,m，

v _m＝w _23,m，

$w_2=\frac{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}\left(u_m{v}_m\right)-\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{v}_m}{M\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m^2-{\left(\underset{m=0}{\overset{M}{\mathrm{\Σ}}}{u}_m\right)}^2},$

Wherein, m is more than or equal to 0, smaller or equal to the integer of M, and N is the number of the said basic coding unit selected.

3. rate-distortion optimization method according to claim 1 is characterized in that, the piece layer Lagrange parameter of said each basic coding unit is:

λ _B＝0.85×2 ^(QP-12)/3，

λ wherein _BBe the piece layer Lagrange parameter of said each basic coding unit, QP is the quantization parameter of said basic coding unit.

4. rate-distortion optimization method according to claim 1 is characterized in that, said step C comprises: obtain the minimum code cost of rate-distortion optimization of the said basic coding unit of said present frame through following formula,

J(s,c,mode|QP)＝min[D(s,c,mode|QP)+λ _modeR(s,c,mode|QP)]，

Wherein J is the minimum code cost; D is a distortion value, and R is the code check value, and s representes the basic coding unit of former figure; C representes the corresponding basic coding unit of building image again through the encoding and decoding processing; Mode representes the coding mode selected of said basic coding unit, and QP is the quantization parameter of said basic coding unit, λ _ModeBe corresponding final Lagrangian parameter;

According to said minimum code cost, obtain corresponding forced coding pattern, the said basic coding unit of said present frame is carried out the rate-distortion optimization coding.

5. rate-distortion optimization method according to claim 1 is characterized in that, said step D comprises: according to the code check value and the distortion value of at least two said basic coding unit of said present frame, adopt following formula to set up said frame layer rate-distortion model,

R＝aln(b/D)，

Wherein R is the code check value, and D is corresponding distortion value, and a, b are model parameter.

6. rate-distortion optimization method according to claim 5 is characterized in that, when the code check value and the distortion value that use two said basic coding unit, when setting up said frame layer rate-distortion model, said model parameter obtains through following formula,

$a=\frac{R_1-R_2}{\ln D_2-\ln D_1},$

$b=e^{\frac{R_1\ln D_2-R_2\ln D_1}{R_1-R_2}},$

R wherein ₁And R ₂Be the code check value of said two basic coding unit, D ₁And D ₂Be corresponding distortion value.

7. rate-distortion optimization method according to claim 5 is characterized in that, when the code check value and the distortion value that use N said basic coding unit, when setting up said frame layer rate-distortion model, said model parameter obtains through following formula,

a＝-p，

b＝e ^-q/p，

$p=\frac{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

$q=\frac{\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{y}_n-\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\left(x_n{y}_n\right)\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n}{N\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n^2-{\left(\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_n\right)}^2},$

_xn＝lnD _n，y _n=R _n，

N＞2，

R wherein _nBe the code check value of said n basic coding unit, D _nBe corresponding distortion value.

8. rate-distortion optimization method according to claim 5 is characterized in that, said step D also comprises: through the frame layer Lagrange parameter of following formula estimation next frame,

λ _F＝-(D(A ₁)-D(A ₂))/(R(A ₁)-R(A ₂))，

R (A wherein ₁) be the code check value of first intersection point, R (A ₂) be the code check value of second intersection point, D (A ₁) be the distortion value of first intersection point, D (A ₂) be the distortion value of second intersection point, said first intersection point is R=R in R～D plane _AWith said rate-distortion model the intersection point of definite curve, said second intersection point is D=D _AWith said rate-distortion model the intersection point of definite curve, R _ABe frame layer bit rate value, D _ABe frame layer distortion value.

9. rate-distortion optimization method according to claim 8 is characterized in that, said frame layer bit rate value R _ABe the code check value sum of all basic coding unit in the said frame, said frame layer distortion value D _AThrough asking mean square error to obtain to the former figure of said frame and the image of building again of said frame.

Images