CN100563338C - A kind of control method of constant code rate - Google Patents

Abstract

The invention provides a kind of control method of constant code rate, described method is divided into several image sets to present image, adopts frame level bit-rate control and macro-block level Rate Control to come image is encoded to each image sets.Constant code rate control method provided by the invention, the control precision height, the code stream fluctuation of output is little, and the coding subjective picture quality is good, and computation complexity is very low, is fit to the application of various software and hardware codings.

Description

A kind of control method of constant code rate

Technical field

The invention belongs to the transmission control correlative technology field of digital audio/video, particularly relate to a kind of control method of constant code rate.

Background technology

In the real time communication of video, code stream need transmit on the network of different bandwidth.For when guaranteeing higher relatively video image quality, improve utilization rate of network resource as much as possible, just the control to the transmission code rate of video flowing has proposed demand.Present Rate Control technology is exactly to control the occupancy of bit in the buffering area by ceaselessly adjusting coding parameter, makes it reach stable as far as possible, the situation of overflow and underflow do not occur; Guaranteeing to guarantee the stable of picture quality under the stable prerequisite of buffering bit occupancy simultaneously as far as possible.In various video compression standards, Rate Control is being played the part of very important role always, for example: the TM5 of MPEG2, TMN8 H.263 etc.Yet at up-to-date video standard, as H.264/MPEG-4AVC with AVS in owing to adopted a lot of brand-new coding techniquess, former a lot of rate control algorithms can not be applied in the new coding standard.

Summary of the invention

The object of the present invention is to provide a kind of control method of constant code rate, described bit rate control method is by carrying out constant code rate control to each image sets (GOP) for unit.Described method is the control precision height not only, and the code stream fluctuation of output is little, and the coding subjective picture quality is good, and computation complexity is low simultaneously, is fit to the application of various software and hardware codings.

The object of the present invention is achieved like this: a kind of control method of constant code rate, and described method realizes Rate Control based on the quantization parameter value of regulating the video image code stream, and described code stream comprises I frame, P frame and B frame, and concrete steps are as follows:

Step 1, video image is divided into several image sets GOP, each GOP is one group of image continuous in time, and its target bit is designated as Bit _Gop, ${\mathrm{Bit}}_{\mathrm{gop}}=(1+N_P+N_B)\×\left(\frac{R}{F}\right)+{\mathrm{Bit}}_{\mathrm{less}},$ N in the following formula _PWith

N _BBe respectively the B frame among each GOP and the frame number of P two field picture, R is a channel object transmission code check, and F is an output frame rate, Bit _LessBe the remaining bit number in last GOP coding back, Bit _LessInitial value be 0;

Step 2, the control of achieve frame level bit-rate, it further comprises:

The bit number that step 2.1, the current B frame of calculating or P frame can access is Bit _{Cur_pic}=λ _Cur* (Bit _Gop-Bit _{Cur_gop_used}), Bit in the following formula _{Cur_gop_used}Be the already used bit number of current GOP, λ _CurBe the weight coefficient of present image, calculate according to the complexity of present image B frame or P frame;

Step 2.2, calculate in every frame the quantization parameter QP of i P frame or B frame respectively _{P_i}Or QP _{B_i}, i=1,2,3......;

Step 2.2.1, calculating QP _{P_i}:

${\mathrm{QP}}_{P\_i}=\mathrm{\α}\×{\mathrm{QP}}_{\mathrm{ref}}+\mathrm{\η}\×\frac{L_0-2n}{L_0-2},$ n＝1，2，...，L ₀-1

In the following formula, α is the coded-bit Bit of present frame _{Cur_pic}Coded-bit Bit with reference frame _{Ref_pic}Ratio, promptly $\mathrm{\α}=\frac{{\mathrm{Bit}}_{\mathrm{cur}\_\mathrm{pic}}}{{\mathrm{Bit}}_{\mathrm{ref}\_\mathrm{pic}}};$ QP _RefBe the quantization parameter of the reference frame of P frame, L ₀Be the length of GOP, η is an empirical value, generally gets 1.8;

Step 2.2.2, calculating QP _{B_i}:

Represent between I frame and the P frame or the B number of frames that comprises between P frame and the P frame with L, the quantization parameter of establishing two two field pictures of B frame reference is respectively QP ₁And QP ₂

Then when L greater than 1 the time, ${\mathrm{QP}}_{B\_i}={\mathrm{QP}}_1+\mathrm{\Δ}+\max \left\{\min \right\{\frac{({\mathrm{QP}}_2-{\mathrm{QP}}_1)(i-1)}{L-1},2(i-1)\},-2(i-1)\}$

Set the QP of the 1st I frame among the GOP and the QP value of the 1st P frame and be QP1, then

QP _{B_1}＝QP ₁+Δ；

Δ is QP _{B_1}With QP ₁Poor, the Δ value is as follows:

When L equals 1, QP _{B_i}=QP ₁+ 2;

Step 2.3, according to the quantization parameter of each frame, again each frame is encoded, until frame end; If frame does not finish, then jump to step 1;

Step 3, realization macro-block level Rate Control, it further comprises:

The average absolute quantizing distortion MAQD of step 3.1, computing macro block is designated as MAQD _MB,

${\mathrm{MAQD}}_{\mathrm{MB}}=\left\{\begin{array}{cc}\frac{1}{N}\underset{i=0}{\overset{N-1}{\mathrm{\Σ}}}\left|C_i\right|& \left|C_i\right|\≤\mathrm{\β}\×\mathrm{step}\\ 0& \left|C_i\right|>\mathrm{\β}\×\mathrm{step}\end{array}\right.$

In the following formula, C _iBe i DCT coefficient behind the residual error dct transform of macro block, N is a number of pixels that macro block comprises, and β is an experiment value, generally gets 0.98, and step is the quantization step of macro block;

Step 3.2, the average absolute quantizing distortion of calculating current macro are designated as MAQD _{MB_cur}, obtaining by the MAQD weighting of the macro block of the relevant position of the MAQD of current macro left side adjacent macroblocks and reference frame, the MAQD of the macro block of the relevant position of left side adjacent macroblocks and reference frame is designated as MAQD respectively _{MB_left}And MAQD _{MB_ref}, then:

The target bits of step 3.3, calculating current macro is designated as Bit _{MB_cur},

${\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}=\mathrm{\θ}\×\frac{{\mathrm{MAQD}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(\mathrm{MAQD}\right)}\×\frac{{\mathrm{Bit}}_{\mathrm{cur}\_\mathrm{pic}}}{N}$

In the following formula, θ is an experiment value, is determined by macro block (mb) type and frame type; N is the macroblock number that present frame comprises, general desirable 0.93; Bit _{Cur_pic}Bit for the present frame distribution; AVG (MAQD) is the arithmetic average of the MAQD of current encoded macro block;

The quantization parameter of step 3.4, calculating current macro is designated as QP _{MB_cur},

QP _{MB_cur}＝λ×QP _{MB_prev}+ΔQP

In the following formula, λ represents the coded-bit Bit of current macro _{MB_cur}Mean value AVG (Bit with the coded-bit of adjacent coded macroblocks _MB) ratio, computing formula is $\mathrm{\λ}=\frac{{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(B{\mathrm{it}}_{\mathrm{MB}}\right)},$ Wherein

$\mathrm{AVG}\left({\mathrm{Bit}}_{\mathrm{MB}}\right)=\frac{1}{i_{\mathrm{MB}\_\mathrm{left}}+i_{\mathrm{MB}\_\mathrm{upleft}}+i_{\mathrm{MB}\_\mathrm{up}}+i_{\mathrm{MB}\_\mathrm{upright}}}({\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{left}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{upleft}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{up}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{upright}})$

Bit _{MB_left}, Bit _{MB_upleft}, Bit _{MB_up}And Bit _{MB_upright}Be respectively the left side of macroblock to be encoded, the last left side, top and the bit of going up the adjacent coded macroblocks in the right; i _{MB_left}, i _{MB_upleft}, i _{MB_up}, i _{MB_upright}Value followingly determine:

QP _{MB_prev}Quantization parameter for the previous encoded macro block of current coding macro block; Δ QP is the adjustment amount of quantization parameter, and Δ QP value is definite by the scope of ω, $\mathrm{\ω}=\frac{{\mathrm{MAQD}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(\mathrm{MAQD}\right)},$

Step 3.5, adjust quantization parameter, finish the coding of current macro according to the Δ QP of macro-block level;

Step 3.6, judge whether all macroblock codings of present image are finished,, then jump to step 2 if macroblock coding is finished; If macroblock coding is not finished, then jump to step 3.1;

Step 4, upgrade the coded quantization parameter of present image, finish the present frame coding according to step 2; Judge whether the frame coding of two field pictures all among the present image group GOP is finished, if fruit is finished, then continues step 5, otherwise, Bit upgraded _Less, jump to step 2.1;

If the coding of all images group of step 5 present image is all finished, then finish, otherwise jump to step 1.

The present invention has been owing to adopted above-mentioned technical scheme, compared with prior art, control precision height not only, the code stream fluctuation of output is little, and the coding subjective picture quality is good, and computation complexity is low, is fit to the application of various software and hardwares.

Embodiment

Below will the control method of constant code rate of the present invention be described in further detail.

Present embodiment requires comprising an I frame, and 4 P frames, the image of 11 B frames are encoded and transmission code rate control, and the image size is 352 * 288, is divided into 16 image sets (GOP), and target frame rate is 400Kbps.The invention provides a kind of control method of constant code rate, described method realizes that based on regulating quantization parameter QP value the specific implementation step is as follows:

Step 1, video image by temporal several GOP that are divided into continuously, the target bit of each GOP is designated as Bit _Gop, ${\mathrm{Bit}}_{\mathrm{gop}}=(1+N_P+N_B)\×\left(\frac{R}{F}\right)+{\mathrm{Bit}}_{\mathrm{less}},$ N in the following formula _PAnd N _BBe respectively the B frame among each image sets GOP and the frame number of P two field picture, R is a channel object transmission code check, and F is an output frame rate, Bit _LessBe the remaining bit number in last GOP coding back, Bit _LessInitial value be 0;

Step 2, the control of achieve frame level bit-rate;

The bit number that step 2.1, the current B frame of calculating or P frame can access is Bit _{Cur_pic}=λ _Cur* (Bit _Gop-Bit _{Cur_gop_used}), Bit in the following formula _{Cur_gop_used}Be the already used bit number of current GOP, λ _CurBe the weight coefficient of present image, calculate according to the complexity of present image B frame or P frame; If the code check that current GOP obtains is Bit _Gop=400kbps, the bit number Bit that the I frame uses _I=95824bit, the bit number Bit that the P frame uses _P=23832bit, the bit number that connects 2 B frames uses of back is respectively 8488bit, 12928bit, Bit like this _{Cur_gop_used}=258928bit, Bit _Gop-Bit _{Cur_gop_used}=141072bit calculates λ to the P frame _CurGet 0.13, then the Bit of P frame _{Cur_pic}=18339bit; The B frame is calculated λ _CurGet 0.05, then the Bit of B frame _{Cur_pic}=7053bit.

Step 2.2, calculate in every frame the quantization parameter QP of i P frame or B frame respectively _{P_i}Or QP _{B_i}, i=1,2,3......;

Step 2.2.1, calculating QP _{P_i}:

${\mathrm{QP}}_{P\_i}=\mathrm{\α}\×{\mathrm{QP}}_{\mathrm{ref}}+\mathrm{\η}\×\frac{L_0-2n}{L_0-2},$ n＝1，2，...，L ₀-1

In the following formula, α is the coded-bit Bit of present frame _{Cur_pic}With the coded-bit of reference frame be Bit _{Ref_pic}Ratio, if known Bit _{Ref_pic}=23832bit, then $\mathrm{\α}=\frac{{\mathrm{Bit}}_{\mathrm{cur}\_\mathrm{pic}}}{{\mathrm{Bit}}_{\mathrm{ref}\_\mathrm{pic}}}=18339/23832=0.77;$ QP _RefBe the quantization parameter of the reference frame of P frame, be the average quantisation parameter of a reference frame, as can be known QP _Ref=36, L ₀For the length of GOP is 16, η is an empirical value, generally gets 1.8; To next P frame n=2, can calculate QP _{P_i}=29.24.

Step 2.2.2, calculating QP _{B_i}:

Represent between I frame and the P frame or the B number of frames that comprises between P frame and the P frame with L, the quantization parameter of establishing two two field pictures of B frame reference is respectively QP ₁And QP ₂, can get 2 reference frame QP of B frame according to above calculating ₁=36, QP ₂=29.24, QP then ₂-QP ₁=-6.76; In the present embodiment, L=2 is according to the Δ value, promptly

Can get Δ=2, ${\mathrm{QP}}_{B\_i}={\mathrm{QP}}_1+\mathrm{\Δ}+\max \left\{\min \right\{\frac{({\mathrm{QP}}_2-{\mathrm{QP}}_1)(i-1)}{L-1},2(i-1)\},-2(i-1)\}=38;$

Step 2.3, according to the quantization parameter of each frame, again each frame is encoded, until frame end; If frame does not finish, then jump to step 1;

Step 3, realization macro-block level Rate Control;

The average absolute quantizing distortion of step 3.1, computing macro block (MAQD, Mean Absolute QuantizedDistortion) is designated as MAQD _MB,

${\mathrm{MAQD}}_{\mathrm{MB}}=\left\{\begin{array}{cc}\frac{1}{N}\underset{i=1}{\overset{N-1}{\mathrm{\Σ}}}\left|C_i\right|& \left|C_i\right|\≤\mathrm{\β}\×\mathrm{step}\\ 0& \left|C_i\right|>\mathrm{\β}\×\mathrm{step}\end{array}\right.$

In the following formula, C _iBe i DCT coefficient behind the residual error dct transform of macro block, N is a number of pixels that macro block comprises, and β is an experiment value, generally gets 0.98, and step is the quantization step of macro block;

Step 3.2, the average absolute quantizing distortion of calculating current macro are designated as MAQD _{MB_cur}, obtaining by the MAQD weighting of the macro block of the relevant position of the MAQD of current macro left side adjacent macroblocks and reference frame, the MAQD of the macro block of the relevant position of left side adjacent macroblocks and reference frame is designated as MAQD respectively _{MB_left}And MAQD _{MB_ref}, then:

The target bits of step 3.3, calculating current macro is designated as Bit _{MB_cur},

${\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}=\mathrm{\θ}\×\frac{{\mathrm{MAQD}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(\mathrm{MAQD}\right)}\×\frac{{\mathrm{Bit}}_{\mathrm{cur}\_\mathrm{pic}}}{N}$

In the following formula, θ is an experiment value, determines that by macro block (mb) type and frame type present embodiment gets 0.93;

For P frame, Bit _{Cur_pic}=18339bit, N=352 * 288/16=6336, to a macro block of P frame, the macro block on the left side exists, MAQD _{MB_left}=2.1, MAQD _{MB_ref}=1.9, MAQD _{MB_cur}=3.5, AVG (MAQD)=2.9, then Bit _{MB_cur}=3.2;

For B frame, Bit _{Cur_pic}=7053bit, N=352 * 288/16=6336, to a macro block of B frame, the macro block on the left side exists, MAQD _{MB_left}=1.1, MAQD _{MB_ref}=1.9, MAQD _{MB_cur}=1.1, AVG (MAQD)=1.09, then Bit _{MB_cur}=1.04.

The quantization parameter of step 3.4, calculating current macro is designated as QP _{MB_cur},

QP _{MB_cur}＝λ×QP _{MB_prev}+ΔQP

In the following formula, λ represents the coded-bit Bit of current macro _{MB_cur}Mean value AVG (Bit with the coded-bit of adjacent coded macroblocks _MB) ratio, computing formula is $\mathrm{\λ}=\frac{{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left({\mathrm{Bit}}_{\mathrm{MB}}\right)},$ Wherein

$\mathrm{AVG}\left({\mathrm{Bit}}_{\mathrm{MB}}\right)=\frac{1}{i_{\mathrm{MB}\_\mathrm{left}}+i_{\mathrm{MB}\_\mathrm{upleft}}+i_{\mathrm{MB}\_\mathrm{up}}+i_{\mathrm{MB}\_\mathrm{upright}}}({\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{left}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{upleft}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{up}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{upright}}).$

For P frame: QP _{MB_prev}=32, ω=3.5/2.9=1.2, Δ QP=-1, i _{MB_left}There is not i _{MB_upleft}, i _{MB_up}And i _{MB_upright}Exist, then i _{MB_left}=0, i _{MB_upleft}=1, i _{MB_up}=1, i _{MB_upright}=1, Bit _{MB_left}=0, Bit _{MB_upleft}=3.5, Bit _{MB_up}=2.9, Bit _{MB_upright}=2.0, AVG (Bit _MB)=(3.5+2.9+2)/3=2.8, then $\mathrm{\λ}=\frac{{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left({\mathrm{Bit}}_{\mathrm{MB}}\right)}=3.2/2.8=1.14,$ QP _{MB_cur}＝λ×QP _{MB_prev}+ΔQP＝1.14×32-1＝35.6；

For B frame: QP _{MB_prev}=35, ω=1.1/1.09=1.0, Δ QP=-1, i _{MB_left}, i _{MB_upleft}, i _{MB_up}And i _{MB_upright}Exist, then i _{MB_left}=1, i _{MB_upleft}=1, i _{MB_up}=1, i _{MB_upright}=1, Bit _{MB_left}=3.1, Bit _{MB_upleft}=3.3, Bit _{MB_up}=3.6, Bit _{MB_upright}=2.8, AVG (Bit _MB)=(3.1+3.3+3.6+2.8)/4=3.2, then $\mathrm{\λ}=\frac{{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(B{\mathrm{it}}_{\mathrm{MB}}\right)}=1.2/3.2=0.38,$ QP _{MB_cur}＝λ×QP _{MB_prev}+ΔQP＝0.38×36-1＝12.5；

Step 3.5, adjust quantization parameter, finish the coding of current macro according to the Δ QP of macro-block level;

Step 3.6, judge whether all macroblock codings of present image are finished,, then jump to step 2 if macroblock coding is finished; If macroblock coding is not finished, then jump to step 3.1;

Step 4, upgrade the coded quantization parameter of present image, finish the present frame coding according to step 2; Judge whether the frame coding of two field pictures all among the present image group GOP is finished, if fruit is finished, then continues step 5, otherwise, Bit upgraded _Less, jump to step 2.1;

If the coding of all images group of step 5 present image is all finished, then finish, otherwise jump to step 1.

The embodiment coding result shows that constant code rate control method of the present invention not only has the control precision height, the little advantage of code stream fluctuation of output, and also the coding subjective picture quality is good, and computation complexity is low simultaneously.

Claims (1)

1, a kind of control method of constant code rate, described method realizes Rate Control based on the quantization parameter value of regulating the video image code stream, and described code stream comprises I frame, P frame and B frame, it is characterized in that, and described bit rate control method is realized by following steps:

Step 1, video image is divided into several image sets GOP, each GOP is one group of image continuous in time, and its target bit is designated as Bit _Gop, ${\mathrm{Bit}}_{\mathrm{gop}}=(1+N_P+N_B)\×\left(\frac{R}{F}\right)+{\mathrm{Bit}}_{\mathrm{less}},$ N in the following formula _PAnd N _BBe respectively the B frame among each GOP and the frame number of P two field picture, R is a channel object transmission code check, and F is an output frame rate, Bit _LessBe the remaining bit number in last GOP coding back, Bit _LessInitial value be 0;

Step 2, the control of achieve frame level bit-rate, it further comprises:

The bit number that step 2.1, the current B frame of calculating or P frame can access is Bit _{Cur_pic}=λ _Cur* (Bit _Gop-Bit _{Cur_gop_used}), Bit in the following formula _{Cur_gop_used}Be the already used bit number of current GOP, λ _CurBe the weight coefficient of present image, calculate according to the complexity of present image B frame or P frame;

Step 2.2, calculate in every frame the quantization parameter QP of i P frame or B frame respectively _{P_i}Or QP _{B_i}, i=1,2,3......;

Step 2.2.1, calculating QP _{P_i}:

${\mathrm{QP}}_{P\_i}=\mathrm{\α}\×{\mathrm{QP}}_{\mathrm{ref}}+\mathrm{\η}\×\frac{L_0-2n}{L_0-2},n=\mathrm{1,2},...,L_0-1$

In the following formula, α is the coded-bit Bit of present frame _{Cur_pic}Coded-bit Bit with reference frame _{Ref_pic}Ratio, promptly $\mathrm{\α}=\frac{{\mathrm{Bit}}_{\mathrm{cur}\_\mathrm{pic}}}{{\mathrm{Bit}}_{\mathrm{ref}\_\mathrm{pic}}};$ QP _RefBe the quantization parameter of the reference frame of P frame, L ₀Be the length of GOP, η is an empirical value, gets 1.8;

Step 2.2.2, calculating QP _{B_i}:

Represent between I frame and the P frame or the B number of frames that comprises between P frame and the P frame with L, the quantization parameter of establishing two two field pictures of B frame reference is respectively QP ₁And QP ₂

Then when L greater than 1 the time, ${\mathrm{QP}}_{B\_i}={\mathrm{QP}}_1+\mathrm{\Δ}+\max \left\{\min \right\{\frac{({\mathrm{QP}}_2-{\mathrm{QP}}_1)(i-1)}{L-1},2(i-1)\},-2(i-1)\}$

Set the QP of the 1st I frame among the GOP and the QP value of the 1st P frame and be QP ₁, then

QP _{B_1}＝QP ₁+Δ；

Δ is QP _{B_1}With QP ₁Poor, the Δ value is as follows:

When L equals 1, QP _{B_i}=QP ₁+ 2;

Step 2.3, according to the quantization parameter of each frame, again each frame is encoded, until frame end; If frame does not finish, then jump to step 1;

Step 3, realization macro-block level Rate Control, it further comprises:

The average absolute quantizing distortion MAQD of step 3.1, computing macro block is designated as MAQD _MB,

${\mathrm{MAQD}}_{\mathrm{MB}}=\left\{\begin{array}{cc}\frac{1}{N}\underset{i=0}{\overset{N-1}{\mathrm{\Σ}}}\left|C_i\right|& \left|C_i\right|\≤\mathrm{\β}\×\mathrm{step}\\ 0& \left|C_i\right|>\mathrm{\β}\×\mathrm{step}\end{array}\right.$

In the following formula, C _iBe i DCT coefficient behind the residual error dct transform of macro block, N is a number of pixels that macro block comprises, and β is an experiment value, gets 0.98, and step is the quantization step of macro block;

Step 3.2, the average absolute quantizing distortion of calculating current macro are designated as MAQD _{MB_cur}, obtaining by the MAQD weighting of the macro block of the relevant position of the MAQD of current macro left side adjacent macroblocks and reference frame, the MAQD of the macro block of the relevant position of left side adjacent macroblocks and reference frame is designated as MAQD respectively _{MB_left}And MAQD _{MB_ref}, then:

The target bits of step 3.3, calculating current macro is designated as Bit _{MB_cur},

${\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}=\mathrm{\θ}\×\frac{{\mathrm{MAQD}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(\mathrm{MAQD}\right)}\×\frac{{\mathrm{Bit}}_{\mathrm{cur}\_\mathrm{pic}}}{N}$

In the following formula, θ is an experiment value, is determined by macro block (mb) type and frame type; N is the macroblock number that present frame comprises, and gets 0.93; Bit _{Cur_pic}Bit for the present frame distribution; AVG (MAQD) is the arithmetic average of the MAQD of current encoded macro block;

The quantization parameter of step 3.4, calculating current macro is designated as QP _{MB_cur},

QP _{MB_cur}＝λ×QP _{MB_prev}+ΔQP

In the following formula, λ represents the coded-bit Bit of current macro _{MB_cur}Mean value AVG (Bit with the coded-bit of adjacent coded macroblocks _MB) ratio, computing formula is $\mathrm{\λ}=\frac{{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left({\mathrm{Bit}}_{\mathrm{MB}}\right)},$ Wherein $\mathrm{AVG}\left({\mathrm{Bit}}_{\mathrm{MB}}\right)=\frac{1}{i_{\mathrm{MB}\_\mathrm{left}}+i_{\mathrm{MB}\_\mathrm{upleft}}+i_{\mathrm{MB}\_\mathrm{up}}+i_{\mathrm{MB}\_\mathrm{upright}}}({\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{left}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{upleft}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{up}}+{\mathrm{Bit}}_{\mathrm{MB}\_\mathrm{upright}})$ Bit _{MB_left}, Bit _{MB_upleft}, Bit _{MB_up}And Bit _{MB_upright}Be respectively the left side of macroblock to be encoded, the last left side, top and the bit of going up the adjacent coded macroblocks in the right; i _{MB_left}, i _{MB_upleft}, i _{MB_up}, i _{MB_upright}Value followingly determine:

QP _{MB_prev}Quantization parameter for the previous encoded macro block of current coding macro block; Δ QP is the adjustment amount of quantization parameter, and Δ QP value is definite by the scope of ω, $\mathrm{\ω}=\frac{{\mathrm{MAQD}}_{\mathrm{MB}\_\mathrm{cur}}}{\mathrm{AVG}\left(\mathrm{MAQD}\right)},$

Step 3.5, adjust quantization parameter, finish the coding of current macro according to the Δ QP of macro-block level;

Step 3.6, judge whether all macroblock codings of present image are finished,, then jump to step 2 if macroblock coding is finished; If macroblock coding is not finished, then jump to step 3.1;

Step 4, upgrade the coded quantization parameter of present image, finish the present frame coding according to step 2; Judge whether the frame coding of two field pictures all among the present image group GOP is finished, if fruit is finished, then continues step 5, otherwise, Bit upgraded _Less, jump to step 2.1;

If the coding of all images group of step 5 present image is all finished, then finish, otherwise jump to step 1.