CN101534432A - Method for controlling code rate based on human eye sensing model - Google Patents

Abstract

The invention provides a method for controlling code rate based on a human eye sensing model. The method comprises the following steps: through extracting static characters and dynamic characters of a video image, establishing a static character model and a dynamic character model respectively; further establishing a static and dynamic comprehensive character model; controlling code rate of frame level by the dynamic character model; and controlling code rate of macro block level by the static and dynamic comprehensive character model. The method for controlling the code rate not only can ensure that the decoded video has better visual effect, but also can realize distribution of target bit more effectively under a condition of limited number of bit.

Description

Bit rate control method based on human eye sensing model

Technical field

The present invention relates to the digital video coding technology, particularly relate to a kind of bit rate control method based on human eye sensing model.

Background technology

Rate Control is one of key technology of video coding, is that encoder is based on the process that will send to the video bits flow rate on the channel to the estimation decision of network availability bandwidth.The key of control is to find a balance between code check size and video compression quality.Therefore the quality of rate control algorithm has directly influenced the performance and the efficient of encoder.H.264/AVC and AVS in existing several compression standards, as, its bit rate control method only carries out Data Rate Distribution according to the complexity of video content, that is, distribute the more bits number for the part of texture complexity.But the final recipient of video is human, and the people can be decided to the apperceive characteristic of video with to the understanding of video content by human brain for the concern situation of video.Under many scientific workers' effort, we have realized that human eye can drop into the more concern degree for the edge contour of moving object in the video and object.Therefore, if think to be different frame, different macroblock allocation target bit more effectively, just must consider visual characteristics of human eyes in the Rate Control and go.

Summary of the invention

The object of the present invention is to provide a kind of bit rate control method based on human eye sensing model, described method is according to the behavioral characteristics and the static nature of video image, set up the static attention model and the dynamic attention model of human vision, and extract the zone that the human eye in the video image is paid close attention to by these two kinds of models, to distinguish the non-zone of paying close attention to, realization is carried out the classification quantitative coding at the different zone of significance level in the image, and then realizes the bit rate control method based on the human eye perception.

The object of the present invention is achieved like this: a kind of bit rate control method based on human eye sensing model, described bit rate control method carries out frame level and macro-block level Rate Control respectively, to determine the distribution of bit number, wherein, the size of macro block is h*h, h is a natural number, and described bit rate control method is realized by following steps:

The static nature of step 1, extraction video image, and set up the static nature model;

The behavioral characteristics of step 2, extraction video image, and set up the behavioral characteristics model;

Step 3, get the common factor of static nature model and behavioral characteristics model, obtain the moving image portion and the profile thereof of this video, i.e. sound attitude comprehensive characteristics model;

Step 4, in the frame level bit-rate control algolithm, utilize the behavioral characteristics model that extracts, calculate the ratio of the shared whole macroblock number of motion object piece in each frame, and with former linear prediction model MAD (the n)=k of mean absolute difference MAD of corresponding macro block pixels point in each macro block in this ratio correction current encoded frame and the corresponding reference frame of coding ₁* MAD (n-1)+k ₂, when the preassignment bit number,, reflect the motion complexity of present frame according to what of motion object piece, thus decision allocation bit number how much realize Rate Control, wherein k ₁Be the linear scale factor of this model, k ₂Be the constant coefficient of this model, n is the picture frame sequence number;

Step 5, in the macro-block level Rate Control, at first according to sound attitude comprehensive characteristics model, calculate the ratio of total pixel in the shared macro block of pixel in the sound attitude comprehensive characteristics model, and by an adjusting of this ratio acquisition parameter, in Rate Control, use and regulate the original target bit mean allocation of parameter modification formula Bit (n, I, J)=f _Rb(n, I, J)/N _Ub, realize the best Rate Control effect under the human eye perception, wherein, n is the picture frame sequence number, ((n, I J) are target bit, f to Bit for I, the J) coordinate of expression macro block _Rb(n, I J) are total bit number to be allocated, N _UbIt is number of macroblocks to be allocated.

The foundation of the extraction of the static nature in the described step 1 and static nature model further realizes by following steps:

Step 1.1, convert each pixel of current video to rgb format from yuv format; Extract red R, green G, blue B and yellow Y four primary colours, and the chromaticity figure that obtains four primary colours be respectively R (r, g, b), G (r, g, b), B (r, g, b) and Y (r, g, b); Extract red/green, indigo plant/yellow two groups of antagonistic pairs, and obtain red/green characteristic pattern RG (i, j)=| (R (and i, j)-G (i, j)) | and indigo plant/yellow characteristic pattern BY (i, j)=| (B (and i, j)-Y (i, j)) |, wherein (i, j) coordinate of remarked pixel point;

Step 1.2, calculate each pixel and the mean value of red/green, the indigo plant/yellow characteristic value of eight points on every side thereof ${\mathrm{Diff}}_{\mathrm{RG}}(i,j)=\frac{1}{9}\underset{n=i-1}{\overset{i+1}{\mathrm{\Σ}}}\underset{m=j-1}{\overset{j+1}{\mathrm{\Σ}}}|\mathrm{RG}(i,j)-\mathrm{RG}(n,m)|$ With ${\mathrm{Diff}}_{\mathrm{BY}}(i,j)=\frac{1}{9}\underset{n=i-1}{\overset{i+1}{\mathrm{\Σ}}}\underset{m=j-1}{\overset{j+1}{\mathrm{\Σ}}}|\mathrm{BY}(i,j)-\mathrm{BY}(n,m)|;$

Step 1.3, calculate the static nature figure of each pixel $\mathrm{StaticMap}(i,j)=\sqrt{{\mathrm{Diff}}_{\mathrm{BY}}{(i,j)}^2+{\mathrm{Diff}}_{\mathrm{RG}}{(i,j)}^2},$ The maximum of static nature figure is designated as StaticMap _Max, minimum value is designated as StaticMap _Min, establish and judge that the static nature figure threshold value whether pixel meets the static nature model is T, then T=(StaticMap _Max+ StaticMap _Min)/2, (i, during j) greater than T, this pixel promptly is included in the static nature model, otherwise is not included in the static nature model as the static nature figure of pixel StaticMap.

The foundation of the extraction of the behavioral characteristics in the described step 2 and behavioral characteristics model further realizes by following steps:

Step 2.1, establish n frame (I, J) motion vector of macro block be designated as PV (I, J)=(x _{N, I, J}, y _{N, I, J}), the motion vector of this macro block is defaulted as the motion vector of this each pixel of macro block, the direction of motion of this vector is expressed as θ _{N, i, i}=arctan (y _{N, i, j}/ x _{N, i, j});

Step 2.2, calculate the current pixel point and the probability histogram distribution function of the motion vector value of eight points on every side thereof $P_s\left(n\right)=\frac{{\mathrm{SH}}_{i,j}^w\left(n\right)}{\underset{l=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{SH}}_{i,j}^w\left(l\right)};$ Wherein SH () is by the current pixel point and the direction value θ of the motion vector of eight points on every side thereof _{N, i, i}The histogram of being formed, m are histogram space size, and w represents the search window size of N*N; Calculate the spatial coherence entropy of the motion vector value of each pixel according to the probability distribution situation of gained $\mathrm{Cs}(i,j)=-\underset{n=1}{\overset{m}{\mathrm{\Σ}}}{P}_s\left(n\right)\mathrm{Log}\left(P_s\left(n\right)\right);$ Wherein, the spatial information entropy of Cs () expression motion vector, P _sIt is the corresponding probability-distribution function of histogram SH ();

The probability histogram distribution function of the motion vector value of the pixel on the same position of step 2.3, the motion vector value that calculates current pixel point and front and back three frames thereof $P_t\left(n\right)=\frac{{\mathrm{TH}}_{i,j}^L\left(n\right)}{\underset{l=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{TH}}_{i,j}^L\left(l\right)},$ Wherein TH () is the motion vector direction indication value θ by current pixel point and front and back three frame relevant position pixels thereof _{N, i, i}The histogram of being formed, P _tBe the corresponding probability-distribution function of histogram TH (), m is a histogram space size, the relevant frame number on the L express time axle; Calculate the temporal correlation entropy of the motion vector value of each pixel thus $\mathrm{Ct}(i,j)=-\underset{n=1}{\overset{m}{\mathrm{\Σ}}}{P}_t\left(n\right)\mathrm{Log}\left(P_t\left(n\right)\right);$ The temporal information entropy of Ct () expression motion vector;

Step 2.4, generalized time and spatial information, obtain final space time information entropy C (i, j)=a ₁ ^*Ct (i, j)+a ₂ ^*Cs (i, j), a wherein ₁+ a ₂=1;

Step 2.5, in a two field picture, the minimum space time information entropy of order be Min[C (i, j)], represent with message level 0, make maximum space time information entropy be Max[C (i, j)], l-1 represents with message level, R={0,1..., l-1} represent the set of message level; Definition N _p(p ∈ R) pixel quantity when being p for message level promptly has the pixel number of identical information entropy; For threshold value t ∈ R, need find out wherein the pairing space time information entropy of certain one-level in the l-1 grade as threshold value t in 0 grade, and carry out self adaptation according to threshold value t and divide, the comentropy that promptly is lower than threshold value is $E_A=-\underset{j=i+1}{\overset{l-1}{\mathrm{\Σ}}}\frac{N_j}{\underset{n=t+1}{\overset{l-1}{\mathrm{\Σ}}}{N}_n}\mathrm{Log}\left(\frac{N_j}{\underset{n=t+1}{\overset{l-1}{\mathrm{\Σ}}}{N}_n}\right);$ The comentropy that is higher than threshold value is $E_B=-\underset{i=0}{\overset{t}{\mathrm{\Σ}}}\frac{N_i}{\underset{m=0}{\overset{t}{\mathrm{\Σ}}}{N}_m}\mathrm{Log}\left(\frac{N_i}{\underset{m=0}{\overset{t}{\mathrm{\Σ}}}{N}_m}\right);$ Threshold value wherein $t=\underset{R}{\arg \max }(E_A+E_B);$ Promptly when the space time information entropy of pixel during greater than threshold value t, promptly this pixel is in the moving region, otherwise is in non-moving region.

The ratio of the shared whole macroblock number of motion object piece is R in each frame in the described step 4 _Mb(n)=N _Motion(n)/N _All, the linear prediction model of revised mean absolute difference MAD is MAD (n)=k ₁* R _Mb(n) * MAD (n-1)+k ₂, N wherein _Motion(n) be the number of the shared macro block of motion object in the current n frame, N _AllBe the macro block sum that each frame comprised.

The ratio of total pixel is R in the shared macro block of pixel in the sound attitude comprehensive characteristics model in the described step 5 _Pixel(n, I, J)=N (n, I, J)/N _All, wherein (n, I J) are (I, the number of the pixel of the static nature that J) is extracted on the piece in the current encoded frame n frame to N; N _AllBe the total pixel number of each macro block.

Adjusting parameter in the described step 5 be α (n, I, J)=R _Pixel(n, I, J)+and b, the target bit mean allocation formula after regulating is $\mathrm{Bit}(n,I,J)\′=\mathrm{\α}(n,I,J)f_{\mathrm{rb}}(n,I,J)/\underset{l=I,k=J}{\overset{l=c,k=d}{\mathrm{\Σ}}}\mathrm{\α}(n,l,k),$ Wherein b prevents that bit number is assigned as 0 radix that is provided with;

For in the present frame from the (I, J) macro block begin all not coded macroblocks the adjusting parameter and, c represents the macro block number that every row comprised in each two field picture, d represents the macro block number that every row comprised in each two field picture.

In the step 1.1, think that every adjacent h pixel has identical YUV value in the described macro block.Described pixel is as follows from the transformational relation that yuv format converts rgb format to: R=Y+1.402 (V-128); G=Y-0.34414 (U-128)-0.71414 (V-128); B=Y+1.772 (U-128).The relation of described R, G, B and Y and three color component r, g and b is as follows: R=r-(g+b)/2; G=g-(r+b)/2; B=b-(r+g)/2; Y=r+g-2 (| r-g|+b).

Comprise also in the step 2.1 that motion vector is made further average value filtering to be handled.

The present invention compared with prior art, not only makes decoded video have better visual effect owing to adopted above-mentioned technical scheme; And under the situation of limit bit number, can realize the distribution of target bits more efficiently.

Description of drawings

Bit rate control method based on human eye sensing model of the present invention is provided in detail by following embodiment and accompanying drawing.

Fig. 1 is the implementation procedure schematic diagram of the embodiment of the invention based on the bit rate control method of human eye sensing model;

Fig. 2 is an a certain two field picture in the video of the embodiment of the invention;

The video static nature hum pattern of Fig. 3 for extracting according to the color contrast principle in the embodiment of the invention;

Fig. 4 is the video behavioral characteristics figure before adaptive threshold is divided in the embodiment of the invention;

Fig. 5 is the video behavioral characteristics figure after adaptive threshold is divided in the embodiment of the invention;

Fig. 6 is video sound attitude characteristic synthetic figure in the embodiment of the invention.

Embodiment

Below will be described in further detail the bit rate control method based on human eye sensing model of the present invention.

Accompanying drawing 1 has simply presented the implementation procedure of present embodiment based on the bit rate control method of human eye sensing model, when video image is as shown in Figure 2 encoded, described bit rate control method carries out frame level and macro-block level Rate Control respectively, the size of macro block is 4*4, and described bit rate control method is realized by following steps:

The static nature of step 1, extraction video image, and set up the static nature model, as shown in Figure 3, specifically comprise the following steps:

Step 1.1, convert each pixel of current video image to rgb format from yuv format, i.e. R=Y+1.402 (V-128); G=Y-0.34414 (U-128)-0.71414 (V-128); B=Y+1.772 (U-128); In order to handle conveniently, in the present embodiment, every adjacent 4 pixels can think to have identical YUV value in the macro block, promptly the YUV value of first pixel are composed the YUV value of giving other 3 pixels; Then extract red R, green G, blue B and yellow Y four primary colours, and the chromaticity figure that obtains four primary colours is respectively R (r, g, b), G (r, g, b), B (r, g, b) and Y (r, g, b), be expressed as follows with color three-component r, g and b: R=r-(g+b)/2, G=g-(r+b)/2, B=b-(r+g)/2, Y=r+g-2 (| r-g|+b); Extract red/green, indigo plant/yellow two groups of antagonistic pairs, and obtain red/green characteristic pattern RG (i, j)=| (R (and i, j)-G (i, j)) | and indigo plant/yellow characteristic pattern BY (i, j)=| (B (and i, j)-Y (i, j)) |, wherein (i, j) coordinate of remarked pixel point;

Step 1.2, calculate each pixel and the mean value of red/green, the indigo plant/yellow characteristic value of eight points on every side thereof ${\mathrm{Diff}}_{\mathrm{RG}}(i,j)=\frac{1}{9}\underset{n=i-1}{\overset{i+1}{\mathrm{\Σ}}}\underset{m=j-1}{\overset{j+1}{\mathrm{\Σ}}}|\mathrm{RG}(i,j)-\mathrm{RG}(n,m)|$ With ${\mathrm{Diff}}_{\mathrm{BY}}(i,j)=\frac{1}{9}\underset{n=i-1}{\overset{i+1}{\mathrm{\Σ}}}\underset{m=j-1}{\overset{j+1}{\mathrm{\Σ}}}|\mathrm{BY}(i,j)-\mathrm{BY}(n,m)|;$

Step 1.3, calculate each pixel static nature figure StaticMap (i, j), wherein $\mathrm{StaticMap}(i,j)=\sqrt{{\mathrm{Diff}}_{\mathrm{BY}}{(i,j)}^2+{\mathrm{Diff}}_{\mathrm{RG}}{(i,j)}^2},$ The maximum of static nature figure is designated as StaticMap _Max, minimum value is designated as StaticMap _Min, establish and judge that the static nature figure threshold value whether pixel meets the static nature model is T, then T=(StaticMap _Max+ StaticMap _Min)/2; That is to say that (i, during j) greater than T, this pixel promptly is included in the static nature model, otherwise is not included in the static nature model, the video information of paying close attention to for non-human eye as the static nature figure of pixel StaticMap.

The behavioral characteristics of step 2, extraction video image, and set up the behavioral characteristics model, as shown in Figure 4, specifically comprise the following steps:

Step 2.1, establish n frame (I, J) motion vector of macro block be designated as PV (I, J)=(x _{N, I, J}, y _{N, I, J}), the motion vector of this macro block is defaulted as the motion vector of this each pixel of macro block, the direction of motion of this vector can be expressed as θ _{N, i, i}=arctan (y _{N, i, j}/ x _{N, i, j}), in order to obtain better effects, further motion vector is made average value filtering and handle;

Step 2.2, calculate the current pixel point and the probability histogram distribution function of the motion vector value of eight points on every side thereof $P_s\left(n\right)=\frac{{\mathrm{SH}}_{i,j}^w\left(n\right)}{\underset{l=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{SH}}_{i,j}^w\left(l\right)};$ Wherein SH () is by the current pixel point and the direction value θ of the motion vector of eight points on every side thereof _{N, i, i}The histogram of being formed; Calculate the spatial coherence entropy of the motion vector value of each pixel according to the probability distribution situation of gained $\mathrm{Cs}(i,j)=-\underset{n=1}{\overset{m}{\mathrm{\Σ}}}{P}_s\left(n\right)\mathrm{Log}\left(P_s\left(n\right)\right);$ Wherein, the spatial information entropy of Cs () expression motion vector, P _sBe the corresponding probability-distribution function of histogram SH (), m is a histogram space size, is the amount of calculation in the probability histogram function, is obtained by the statistical computation value; W represents the search window size of N*N, and present embodiment adopts the 3*3 window, and promptly w is 9.

The probability histogram distribution function of the motion vector value of the pixel on the same position of the motion vector value of step 2.3, current pixel point and front and back three frames thereof is $P_t\left(n\right)=\frac{{\mathrm{TH}}_{i,j}^L\left(n\right)}{\underset{l=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{TH}}_{i,j}^L\left(l\right)};$ Calculate the temporal correlation entropy of the motion vector value of each pixel thus $\mathrm{Ct}(i,j)=-\underset{n=1}{\overset{m}{\mathrm{\Σ}}}{P}_t\left(n\right)\mathrm{Log}\left(P_t\left(n\right)\right);$ Wherein TH () is the motion vector direction indication value θ by current pixel point and front and back three frame relevant position pixels thereof _{N, i, i}The histogram of being formed; The temporal information entropy of Ct () expression motion vector, P _tBe the corresponding probability-distribution function of histogram TH (), m is a histogram space size, and relevant frame number gets 7 in the present embodiment on the L express time axle;

Step 2.4, generalized time and spatial information, obtain final space time information entropy C (i, j)=a ₁* Ct (i, j)+a ₂* Cs (i, j), a wherein ₁+ a ₂=1, a in the present embodiment ₁=0.7; a ₂=0.3;

Step 2.5, in a two field picture, the minimum space time information entropy of order be Min[C (i, j)], represent with message level 0, make maximum space time information entropy be Max[C (i, j)], l-1 represents with message level.R={0,1..., l-1} represent the set of message level; Definition N _p(p ∈ R) pixel quantity when being p for message level promptly has the pixel number of identical information entropy; For threshold value t ∈ R, need find out wherein the pairing space time information entropy of certain one-level in the 1-1 grade as threshold value t in 0 grade, and carry out self adaptation according to threshold value t and divide, the comentropy that promptly is lower than threshold value is $E_A=-\underset{j=t+1}{\overset{l-1}{\mathrm{\Σ}}}\frac{N_j}{\underset{n=t+1}{\overset{l-1}{\mathrm{\Σ}}}{N}_n}\mathrm{Log}\left(\frac{N_j}{\underset{n=t+1}{\overset{l-1}{\mathrm{\Σ}}}{N}_n}\right);$ The comentropy that is higher than threshold value is $E_B=-\underset{i=0}{\overset{t}{\mathrm{\Σ}}}\frac{N_i}{\underset{m=0}{\overset{t}{\mathrm{\Σ}}}{N}_m}\mathrm{Log}\left(\frac{N_i}{\underset{m=0}{\overset{t}{\mathrm{\Σ}}}{N}_m}\right);$ Threshold value wherein $t=\underset{R}{\arg \max }(E_A+E_B);$ Promptly when the space time information entropy of pixel during greater than threshold value t, promptly this pixel is in the moving region, otherwise is in non-moving region, the video behavioral characteristics figure before dividing as shown in Figure 4, the video behavioral characteristics figure after the division is as shown in Figure 5;

Step 3, get the common factor of static nature model and behavioral characteristics model, obtain the moving image portion and the profile thereof of this video, i.e. sound attitude comprehensive characteristics model, as shown in Figure 6;

Step 4, in the frame level bit-rate control algolithm, the behavioral characteristics model that utilize to extract calculates motion object piece in each frame, i.e. the macroblock number that comprises in the motion model, the ratio R of shared whole macroblock number _Mb(n)=N _Motion(n)/N _All, and with former linear prediction model MAD (n)=k of this ratio correction mean absolute difference MAD ₁* MAD (n-1)+k ₂, revised MAD model be MAD ' (n)=k ₁* R _Mb(n) * MAD ' (n-1)+k ₂, when the preassignment bit number, according to the shared macroblock number of motion object what, reflect the motion complexity of present frame, thus decision allocation bit number how much realize Rate Control, wherein k ₁Be the linear scale factor of this model, k ₂Be the constant coefficient of this model, N _Motion(n) be the number of the shared macro block of motion object in the current n frame, N _AllBe the macro block sum that each frame comprised;

Step 5, in the macro-block level Rate Control, at first according to sound attitude comprehensive characteristics model, calculate the pixel in the sound attitude comprehensive characteristics model, the pixel paid close attention to of human eye just, the ratio R of total pixel in the shared macro block _Pixel(n, I, J)=N (n, I, J)/N _All, and by this ratio obtain one regulate parameter alpha (n, I, J)=R _Pixel(n, I, J)+b, the original target bit mean allocation of usefulness adjusting parameter modification formula Bit in Rate Control (n, I, J)=f _Rb(n, I, J)/N _Ub, the target bit mean allocation formula after regulating is $\mathrm{Bit}(n,I,J)\′=\mathrm{\α}(n,I,J)f_{\mathrm{rb}}(n,I,J)/\underset{l=I,k=J}{\overset{l=c,k=d}{\mathrm{\Σ}}}\mathrm{\α}(n,l,k),$ Wherein b prevents that bit number is assigned as 0 radix that is provided with, and gets 1 herein;

For in the present frame from (I, J) macro block begin all not coded macroblocks the adjusting parameter and, c represents the macro block number that every row comprised in each two field picture, and d represents the macro block number that every row comprised in each two field picture, thereby realizes the best Rate Control effect under the human eye perception.

By the bit rate control method based on human eye sensing model of the present invention, not only make decoded video have better visual effect; And under the situation of limit bit number, can realize the distribution of target bits more efficiently.

Claims (10)

1, a kind of bit rate control method based on human eye sensing model, described bit rate control method carries out frame level and macro-block level Rate Control respectively, to determine the distribution of target bit, wherein, the size of macro block is h*h, h is a natural number, it is characterized in that, described bit rate control method is realized by following steps:

The static nature of step 1, extraction video image, and set up the static nature model;

The behavioral characteristics of step 2, extraction video image, and set up the behavioral characteristics model;

Step 3, get the common factor of static nature model and behavioral characteristics model, obtain the moving image portion and the profile thereof of this video, i.e. sound attitude comprehensive characteristics model;

Step 4, in the frame level bit-rate control algolithm, utilize the behavioral characteristics model that extracts, calculate the ratio of the shared whole macroblock number of motion object piece in each frame, and with former linear prediction model MAD (the n)=k of mean absolute difference MAD of corresponding macro block pixels point in each macro block in this ratio correction current encoded frame and the corresponding reference frame of coding ₁* MAD (n-1)+k ₂, when the preassignment bit number,, reflect the motion complexity of present frame according to what of motion object piece, thus decision allocation bit number how much realize Rate Control, wherein k ₁Be the linear scale factor of this model, k ₂Be the constant coefficient of this model, n is the picture frame sequence number;

Step 5, in the macro-block level Rate Control, at first according to sound attitude comprehensive characteristics model, calculate the ratio of total pixel in the shared macro block of pixel in the sound attitude comprehensive characteristics model, and by an adjusting of this ratio acquisition parameter, in Rate Control, use and regulate the original target bit mean allocation of parameter modification formula Bit (n, I, J)=f _Rb(n, I, J)/N _Ub, realize the best Rate Control effect under the human eye perception, wherein, n is the picture frame sequence number, ((n, I J) are target bit, f to Bit for I, the J) coordinate of expression macro block _Rb(n, I J) are total bit number to be allocated, N _UbIt is number of macroblocks to be allocated.

2, the bit rate control method based on human eye sensing model as claimed in claim 1 is characterized in that, the foundation of the extraction of the static nature in the described step 1 and static nature model further realizes by following steps:

Step 1.1, convert each pixel of current video to rgb format from yuv format; Extract red R, green G, blue B and yellow Y four primary colours, and the chromaticity figure that obtains four primary colours be respectively R (r, g, b), G (r, g, b), B (r, g, b) and Y (r, g, b); Extract red/green, indigo plant/yellow two groups of antagonistic pairs, and obtain red/green characteristic pattern RG (i, j)=| (R (and i, j)-G (i, j)) | and indigo plant/yellow characteristic pattern BY (i, j)=| (B (and i, j)-Y (i, j)) |, wherein (i, j) coordinate of remarked pixel point;

Step 1.2, calculate each pixel and the mean value of red/green, the indigo plant/yellow characteristic value of eight points on every side thereof ${\mathrm{Diff}}_{\mathrm{RG}}(i,j)=\frac{1}{9}\underset{n=i-1}{\overset{i+1}{\mathrm{\Σ}}}\underset{m=j-1}{\overset{j+1}{\mathrm{\Σ}}}|\mathrm{RG}(i,j)-\mathrm{RG}(n,m)|$ With ${\mathrm{Diff}}_{\mathrm{BY}}(i,j)=\frac{1}{9}\underset{n=i-1}{\overset{i+1}{\mathrm{\Σ}}}\underset{m=j-1}{\overset{j+1}{\mathrm{\Σ}}}|\mathrm{BY}(i,j)-\mathrm{BY}(n,m)|;$

Step 1.3, calculate the static nature figure of each pixel $\mathrm{Stati\; cMap}(i,j)=\sqrt{{\mathrm{Diff}}_{\mathrm{BY}}{(i,j)}^2+{\mathrm{Diff}}_{\mathrm{RG}}{(i,j)}^2},$ The maximum of static nature figure is designated as StaticMap _Max, minimum value is designated as StaticMap _Min, establish and judge that the static nature figure threshold value whether pixel meets the static nature model is T, then T=(StaticMap _Max+ StaticMap _Min)/2, (i, during j) greater than T, this pixel promptly is included in the static nature model, otherwise is not included in the static nature model as the static nature figure of pixel StaticMap.

3, the bit rate control method based on human eye sensing model as claimed in claim 1 is characterized in that, the foundation of the extraction of the behavioral characteristics in the described step 2 and behavioral characteristics model further realizes by following steps:

Step 2.1, establish n frame (I, J) motion vector of macro block be designated as PV (I, J)=(x _{N, I, J}, y _{N, I, J}), the motion vector of this macro block is defaulted as the motion vector of this each pixel of macro block, the direction of motion of this vector is expressed as θ _{N, i, j}=arctan (y _{N, i, j}/ x _{N, i, j});

Step 2.2, calculate the current pixel point and the probability histogram distribution function of the motion vector value of eight points on every side thereof $P_s\left(n\right)=\frac{{\mathrm{SH}}_{i,j}^w\left(n\right)}{\underset{l=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{SH}}_{i,j}^w\left(l\right)};$ Wherein SH () is by the current pixel point and the direction value θ of the motion vector of eight points on every side thereof _{N, i, i}The histogram of being formed, m are histogram space size, and w represents the search window size of N*N; Calculate the spatial coherence entropy of the motion vector value of each pixel according to the probability distribution situation of gained $\mathrm{Cs}(i,j)=-\underset{n=1}{\overset{m}{\mathrm{\Σ}}}{P}_s\left(n\right)\mathrm{Log}\left(P_S\left(n\right)\right);$ Wherein, the spatial information entropy of Cs () expression motion vector, P _sIt is the corresponding probability-distribution function of histogram SH ();

The probability histogram distribution function of the motion vector value of the pixel on the same position of step 2.3, the motion vector value that calculates current pixel point and front and back three frames thereof $P_t\left(n\right)=\frac{{\mathrm{TH}}_{i,j}^L\left(n\right)}{\underset{l=1}{\overset{m}{\mathrm{\Σ}}}{\mathrm{TH}}_{i,j}^L\left(l\right)},$ Wherein TH () is the motion vector direction indication value θ by current pixel point and front and back three frame relevant position pixels thereof _{N, i, i}The histogram of being formed, P _tBe the corresponding probability-distribution function of histogram TH (), m is a histogram space size, the relevant frame number on the L express time axle; Calculate the temporal correlation entropy of the motion vector value of each pixel thus $\mathrm{Ct}(i,j)=-\underset{n=1}{\overset{m}{\mathrm{\Σ}}}{P}_t\left(n\right)\mathrm{Log}\left(P_t\left(n\right)\right);$ The temporal information entropy of Ct () expression motion vector;

Step 2.4, generalized time and spatial information, obtain final space time information entropy C (i, j)=a ₁* Ct (i, j)+a ₂* Cs (i, j), a wherein ₁+ a ₂=1;

Step 2.5, in a two field picture, the minimum space time information entropy of order be Min[C (i, j)], represent with message level 0, make maximum space time information entropy be Max[C (i, j)], l-1 represents with message level, R={0,1..., l-1} represent the set of message level; Definition N _p(p ∈ R) pixel quantity when being p for message level promptly has the pixel number of identical information entropy; For threshold value t ∈ R, need find out wherein the pairing space time information entropy of certain one-level in the l-1 grade as threshold value t in 0 grade, and carry out self adaptation according to threshold value t and divide, the comentropy that promptly is lower than threshold value is $E_A=-\underset{j=t+1}{\overset{l-1}{\mathrm{\Σ}}}\frac{N_j}{\underset{n=t+1}{\overset{l-1}{\mathrm{\Σ}}}{N}_n}\mathrm{Log}\left(\frac{N_j}{\underset{n=t+1}{\overset{l-1}{\mathrm{\Σ}}}{N}_n}\right);$ The comentropy that is higher than threshold value is $E_B=-\underset{i=0}{\mathrm{\Σ}}\frac{N_i}{\underset{m=0}{\overset{t}{\mathrm{\Σ}}}{N}_m}\mathrm{Log}\left(\frac{N_i}{\underset{m=0}{\overset{t}{\mathrm{\Σ}}}{N}_m}\right);$ Threshold value t=arg max (E wherein _A+ E _B); Promptly when the space time information entropy of pixel during greater than threshold value t, promptly this pixel is in the moving region, otherwise is in non-moving region.

4, the bit rate control method based on human eye sensing model as claimed in claim 1 is characterized in that, the ratio of the shared whole macroblock number of motion object piece is R in each frame in the described step 4 _Mb(n)=N _Motion(n)/N _All, the linear prediction model of revised mean absolute difference MAD be MAD ' (n)=k ₁* R _Mb(n) * MAD ' (n-1)+k ₂, N wherein _Motion(n) be the number of the shared macro block of motion object in the current n frame, N _AllBe the macro block sum that each frame comprised.

5, the bit rate control method based on human eye sensing model as claimed in claim 1 is characterized in that, the ratio of total pixel is R in the shared macro block of pixel in the sound attitude comprehensive characteristics model in the described step 5 _Pixel(n, I, J)=N (n, I, J)/N _All, wherein (n, I J) are (I, the number of the pixel of the static nature that J) is extracted on the piece in the current encoded frame n frame to N; N _AllBe the total pixel number of each macro block.

6, the bit rate control method based on human eye sensing model as claimed in claim 5 is characterized in that, the adjusting parameter in the described step 5 be α (n, I, J)=R _Pixel(n, I, J)+and b, the target bit mean allocation formula after regulating is $\mathrm{Bit}(n,I,J)\′=\mathrm{\α}(n,I,J)f_{\mathrm{rb}}(n,I,J)/\underset{l=I,k=J}{\overset{l=c,k=d}{\mathrm{\Σ}}}\mathrm{\α}(n,l,k),$ Wherein b prevents that bit number is assigned as 0 radix that is provided with;

For in the present frame from the (I, J) macro block begin all not coded macroblocks the adjusting parameter and, c represents the macro block number that every row comprised in each two field picture, d represents the macro block number that every row comprised in each two field picture.

7, the bit rate control method based on human eye sensing model as claimed in claim 2 is characterized in that, in the step 1.1, every adjacent h pixel is considered to have identical YUV value in the described macro block.

8, the bit rate control method based on human eye sensing model as claimed in claim 2 is characterized in that, in the step 1.1, described pixel is as follows from the transformational relation that yuv format converts rgb format to: R=Y+1.402 (V-128); G=Y-0.34414 (U-128)-0.71414 (V-128); B=Y+1.772 (U-128).

9, the bit rate control method based on human eye sensing model as claimed in claim 2 is characterized in that, in the step 1.1, the relation of described R, G, B and Y and three color component r, g and b is as follows: R=r-(g+b)/2; G=g-(r+b)/2; B=b-(r+g)/2; Y=r+g-2 (| r-g|+b).

10, the bit rate control method based on human eye sensing model as claimed in claim 3 is characterized in that: comprise also in the step 2.1 that motion vector is made further average value filtering to be handled.

Images