CN1245840C - Variable position-element rate control device for dynamic image compression - Google Patents

Abstract

The present invention relates to a position variable element rate control device for dynamic image compression. According to the present invention, the total activating quantity of the current picture to be encoded in a group of continuous pictures is calculated firstly; then, the complexity of the current picture is estimated on the basis of the total activating quantity and the proportion of the activating quantity to the complexity of the previous encoding picture in the same type in the group of continuous pictures, and the instant complexity and the statistic complexity of the group of continuous pictures are updated by the complexity; then, the instant position element rate is calculated; a target position element rated value is allocated to the current picture according to the complexity of the picture, the instant complexity and the instant position element rate of the group of continuous pictures. After the picture is encoded, the proportion of the activating quantity to the complexity of the current picture is worked out according to the total activating quantity of the current picture, the number of actually consumed position elements and the average value of actually used quantizing position steps.

Description

The variable position-element rate control device of dynamic image compression

Technical field

The present invention is meant especially that relevant for the field of data compression a kind of video coding system that is applicable to is to carry out the device of variable position-element rate control.

Background technology

As everyone knows, the telecommunications world is just experiencing a digital revolution, is to be close to inerrably with the major advantage of numeral information to store, regenerate, pass receipts, handle and the data bit flow filament (bitstream) of utilization.For example, the color video image per second of NTSC has 29.97 pictures, every about 480 visible scanning linears of picture, every scanning linear needs 480 pixels (pixel) that present with red, green, blue three looks approximately, but if every kind of colour component is with 8 codings, the about 168 megabit units (Mbits/s) of bit rate (bitrate) per second that is then produced so the bit rate that its uncompressed of various video format is handled is very high and uneconomical, therefore are unsuitable for most application.

Digital audio and the video integrated with computer, communication network, consumer products, advancing of stimulus information revolution more, the core of this revolution then be look, the digital compression technology of audio frequency.Many compression standards comprise the algorithm based on the common marrow of compress technique, as: ITU-T (predecessor is CCITT) recommendation H.261 and H.263, and the MPEG-1 of ISO/IEC, MPEG-2 and MPEG-4 standard.The algorithm of MPEG is by (the Moving Picture Experts Group of dynamic image expert group, MPEG) develop out, this dynamic image expert group is (the InternationalStandards Organization of International Standards Organization, ISO) and (the InternationalElectrotechnical Commission of International Electrotechnical Commission (IEC), IEC) joint technical committee, being devoted to development looks, the compression of audio frequency and multichannel pass different manifestation mode, these standard codes the compressed-bit flow filament grammer (syntax) and the decoding method, but for the employed algorithm of encoder, but novelty and the diversity for coding kept the considerable degree of freedom.

According to MPEG, a series of video pictures (picture) be divided into a sequence picture group (groupof picture, GOP), wherein every group of GOP begins with the I-picture, the arrangement of P-picture and B-picture is followed in the back, and Fig. 1 illustrates one group of typical GOP with the order that shows.The coding of I-picture needn't be with reference to previous or picture in the future, and the P-picture is then encoded with reference to immediate in time I-picture or P-picture in the video pictures of consecutive, and the B-picture intersperses among between I-picture and the P-picture.And utilize next-door neighbour I-picture and P-picture previous, genetic system in the future to encode.Though and then several B-pictures can occur continuously, must not estimate other picture with the B-picture.

Every picture has three kinds of compositions: brightness value (luminance represents with Y), red difference (redcolor difference represents with Cr), and blue difference (blue color difference represents with Cb).For the form of the 4:2:0 of MPEG, the sampling point of each Cr and Cb composition has only half of Y composition in level and vertical direction.As shown in Figure 2, its basic comprising square of the picture of a MPEG is big block (macroblock represents with MB).Video with 4:2:0 is an example, and each MB comprises the array of Y composition 16 * 16 sampling points, and the block of two Cr and Cb composition 8 * 8 sampling points, and wherein in fact the array of Y composition 16 * 16 sampling points is made up of the block of 48 * 8 sampling points.

The effect of encoder is to determine which kind of picture coding kenel and which kind of predictive mode for best.To every I-picture, among the MB each 8 * 8 squares all pass through discrete cosine transform (discrete cosinetransform, DCT) form 8 * 8 conversion coefficient arrays, conversion coefficient then quantizes with a quantization matrix, use zigzag (zig-zag) to scan the quantized result of DCT coefficient and obtain a series of DCT coefficient then, and (variable length code's this DCT coefficient sequence VLC) encodes with variable-length code (VLC).The P-picture must determine each MB is encoded with I type MB or P type MB, the coding of I type MB is to carry out in above-mentioned mode, and for each P type MB, then need obtain the prediction that this MB is done with previous picture, this prediction is to obtain by a kind of motion-vector (motion vector), motion-vector is signifying to be about in the present picture MB of coding and is formerly estimating change between the MB in the picture, estimates that MB then scans with DCT, quantification, zigzag with predicated error between present MB and VLC encodes.

Which kind of during treatments B-picture, must determine with following MB pattern encode each MB:I pattern, F pattern, B pattern and FB pattern.The I pattern is not by motion compensation encode (as I type MB) with MB itself; The F pattern is unidirectional predictive coding forward, is estimate (as the P type MB) that obtains motion compensation with previous picture; Otherwise the B pattern is unidirectional predictive coding backward, is to obtain estimating of motion compensation with afterwards picture.FB pattern specifically, it is two-way predictive coding, uses forward and motion compensation backward to estimate both and does interpolation (interpolation) and the motion compensation that obtains the FB pattern is estimated.For F, B and FB pattern, predicated error can use DCT, quantification, zigzag to scan and VLC encodes.

Video encoder is necessary for that whole picture chosen quantization step so that the visible distortion of control under given bit rate, yet, with actual bit number that picture is used of quantization step coding of choosing, must be after real coding, just can learn, do not exist one to reverse function in the reality, can under the quantization step that given hope reaches, determine an actual bit number that picture is used in advance.The encoder the very important point is the control of bit rate.Important video encoder restriction among the MPEG: i.e. every restriction that the used bit number of picture changes, particularly in the fixing situation of bit rate running, this restriction is that (Video Buffer Verifier VBV) implements through the video buffer proofer.If for every picture, the input data rate of VBV is all identical, and then video can be said so, and (Constant Bitrate CBR) encodes with fixing bit rate; Otherwise be variable position-element rate (Variable Bitrate, VBR) coding.VBV is a kind of virtual bumper and is input buffer model in decoder end, and with fixing bit rate, the bit amount that encoder is distributed to a picture must make the VBV buffer can spill-over or scarcity; And for variable position-element rate, the bit element flow of encoding then be with specific maximum bit rate enter the VBV buffer saturated up to it, do not have till the input of many bit again, this may be interpreted as the bit rate that enters the VBV buffer and is actually variable and can reaches above-mentioned maximum bit rate.So for variable position-element rate, as long as prevent that then VBV is deficient.

The purpose of variable position-element rate control is to improve the quality of decoding rear video as much as possible and can also keep output bit rate in allowed limits.Different with the mode of fixing bit rate, variable position-element rate has bigger elasticity and comes that the outer bit number of allocation is complicated to scenery, the more violent picture of change.But, traditional variable position-element rate control architecture is but quite complicated, need carry out just finishing several times the processing of video coding usually.

Summary of the invention

The variable position-element rate control device that the purpose of this invention is to provide a kind of dynamic image compression of novelty, can be used for single, instant video encoder, moreover also expectation can provide a kind of and be fit to be performed in integrated circuit, have the dynamic image compression set of variable position-element rate control.

According to the present invention, each opens the distribution of its target bit of picture, be based on coding result and its activity of analyzing in advance of present picture of previous picture to being about to be encoded, moreover, relation between the activity of analyzing in advance by present picture and the actual complex degree of previous coding picture, the complexity of picture can be got by estimation at present.Because the complexity of picture contains the quality information after this picture coding, so the present picture complexity of estimating is very useful to video encoder, can distribute target bit amount to give every picture more accurately, use and reach in the video sequence vision quality of homogeneous between the different pictures.

Present invention is directed to a kind of variable position-element rate control device of dynamic image compression.According to the present invention, this device comprises and moves evaluation unit, a complexity estimator, a target bit allocator and a parameter update device.The present picture that is about to be encoded in one group of continuous pictures in the mobile evaluation unit receiver, video sequence is in order to calculate the full activity of this present picture.Based on together its activity of previous coding picture of kenel is to the ratio of complexity in the full activity of present picture and this group continuous pictures, the complexity estimator can be used to estimate the complexity of present picture.The complexity of estimating with present picture, target bit allocator upgrades the moment complexity of this group continuous pictures, and based on the set average bit rate of this group continuous pictures, statistics complexity and moment complexity calculate the moment bit rate of this group continuous pictures, and according to the complexity of present picture, this group continuous pictures moment complexity and moment the bit rate, target bit allocator is distributed to present picture with target bit amount.And the parameter update device according to full activity, the actual consumption bit number of present picture and the mean value of the actual used quantization step of present picture of encoding, calculates the ratio of its activity of present picture to complexity after present picture coding; Wherein, the activity of picture is directly proportional with the full activity of present picture to the ratio of complexity at present, and is inversely proportional to the actual consumption bit number of present picture and the mean value of the actual used quantization step of present picture.In addition, the parameter update device is that aforesaid statistics complexity is calculated on the basis with the complexity of present picture also.

On the other hand, the present invention also discloses a kind of target bit allocator, comprises bit rate calculator, a quota calculator and a limit unit in a flash.The complexity that the present picture that is about to be encoded in one group of continuous pictures in the moment bit rate calculator receiver, video sequence is contained and the statistics complexity of this group continuous pictures, and with this complexity is the moment complexity that this group continuous pictures is upgraded on the basis, and calculate the moment bit rate of this group continuous pictures according to a function, wherein this function depend on this group continuous pictures set average bit rate, add up complexity and moment complexity.The quota calculator then based on the complexity of present picture, this group continuous pictures moment the bit rate and moment complexity, target bit amount is distributed to present picture, limit unit then is used for this target bit amount is limited under the set upper limit.

Description of drawings

Fig. 1 presents typical picture group (GOP) with the order that shows.

Fig. 2 is the big block of MPEG.

Fig. 3 is the block schematic diagram of its preferred embodiment of variable position-element rate control device of the present invention.

Fig. 4 is an operational flowchart of the present invention.

Fig. 5 is the block schematic diagram of the target bit allocator among Fig. 3.

The figure number explanation

300...... dynamic image compression set

310...... mobile evaluation unit

312...... signal line

320...... complexity estimator

330...... target bit allocator

340...... parameter update device

350...... image coder

510...... moment bit rate calculator

520......... quota calculator

530......... limit unit

P ⁽ⁱ⁾... the .. picture

A ⁽ⁱ⁾... the full activity of ..

C _Est ⁽ⁱ⁾... .. estimates complexity

ACR ⁽ⁱ⁾... activity is to the ratio of complexity

EB............ the super bit amount of using

STAT_C..... add up complexity

TB ⁽ⁱ⁾... target bit amount

AQ ⁽ⁱ⁾... the mean value of actual used quantization step

UB ⁽ⁱ⁾... actual consumption bit number

CD ... the data behind the coding

INST_R...... moment the bit rate

INST_C...... moment complexity

Embodiment

For above-mentioned purpose of the present invention, feature and advantage can be become apparent, a preferred embodiment cited below particularly, and cooperate appended graphicly, be described in detail below:

At the beginning, the I-picture of prediction usefulness and P-picture must be decoded in mpeg encoder earlier, utilize root mean square (the root mean square) error of calculating between raw frames and decoded pictures, can obtain the objective measurement of its quality of decoded picture.With the distortion factor of root-mean-square error as decoded picture, then the complexity of decoded picture (complex)-C is defined as follows:

C＝r×d

Wherein, r is the required bit rate of this picture of coding, and d is the distortion factor of decoded pictures.In theory, be a proportionate relationship between the average quantization step (quantization step size) and the distortion factor, so the complexity of picture may be defined to:

$C\≈\frac{r\×q}{2}$

Wherein, q is the average quantization step-length of this picture.After picture coding, can learn bit rate (actual consumption bit number) and average quantization step-length, and therefore the complexity of this picture obtains also.

The complexity of picture look closely picture essence and the coding form and decide.The complexity of the nearest coded picture of prior art utilization is represented the complexity of present picture, and keep its complexity to alleviate the influence of different coding form respectively for I-, P-and B-picture, if its content of the picture of video sequence is smooth-going variation, then this mode can reach pretty good effect.Yet because inconsistent between the target bit of picture coding unit's quota and its actual consumption bit number, if the content of video sequence significantly changes, aforesaid way also is not suitable for the vision quality that is used for reaching an agreement.

According to the present invention, analyze the relation between the actual complex degree of the activity (activity) of present picture and previous coding picture in advance, the complexity of picture can be estimated by this at present.Activity be a kind of behind picture coding its bit rate and quality measure trial, suppose that the full activity A of one whole picture is proportional to complexity C, then

$A=k\×C\≈\frac{k}{2}\×r\×q=k^{\′}\×r\×q$

Wherein, k is a proportionality constant.If k ' be activity to the ratio of complexity (activity-to-complexityratio, ACR), then activity A can also following formula represent entirely:

A＝ACR×r×q

According to the present invention, can estimate ACR from the previous coding picture of identical kenel.In idea, before picture of coding, calculate its full activity A earlier, then, the picture P that is about to be encoded ⁽ⁱ⁾Complexity C _Est ⁽ⁱ⁾Estimation as follows:

$C_{\mathrm{est}}^{\left(i\right)}=\frac{A^{\left(i\right)}}{{\mathrm{ACR}}^{(i-1)}}$

Wherein, A ⁽ⁱ⁾Be picture P ⁽ⁱ) full activity, ACR ^(i-1)Be with its activity of previous coding picture of kenel ratio to complexity.Below, the subscript in symbol or parameter ⁽ⁱ⁾Represent that this symbol or variable are relevant with the present picture that is about to be encoded; In like manner, the subscript in symbol or variable ^(i-1)Then relevant with the previous coding picture.

The complexity C that estimates _Est ⁽ⁱ⁾Can be used to its complexity of frame updating, so its complexity of the picture of three kinds of kenels C for suitable kenel _I, C _PAnd C _BTo store respectively in order to follow-up processing.The moment complexity of one group of continuous pictures (instantaneous complexity) is also upgraded as follows:

INST_C＝N _I×C _I+N _P×C _P+N _B×C _B

Wherein, N _I, N _PAnd N _BBe respectively the number of I-, P-and B-picture in this group continuous pictures, and one group of continuous pictures of indication comprise one group of GOP at least herein.Except C _I, C _PAnd C _BOutside, also deposit the complexity statistic SC of I-, P-and B-picture respectively _I, SC _PAnd SC _B, and get the complexity C of present picture _Est ⁽ⁱ⁾Upgrade the complexity statistic with the kenel picture, then the statistics complexity of one group of continuous pictures is:

STAT_C＝N _I×SC _I+N _P×SC _P+N _B×SC _B

If maximum, minimum bit rate is expected the coding bit rate that reaches and is determined that in advance so, the moment bit rate of one group of continuous pictures can be calculated and be got by following formula according to every group of continuous pictures:

$\mathrm{INST}\_R={\mathrm{ER}}_{\mathrm{mean}}\·(1+\mathrm{SF}\·(\frac{\mathrm{INST}\_C}{\mathrm{STAT}\_C}-1))$

Wherein, ER _MeanBe the set average bit rate of this group continuous pictures, SF is a zoom factor, and its scope is between 0 to 1; Moment bit rate INST_R also must be limited between the set maximum of every group of continuous pictures, the minimum bit rate further:

ER _min≤INST_R≤ER _max

In case moment complexity INST_C and moment bit rate INST_R determine, then present picture P ⁽ⁱ⁾Target bit amount TB ⁽ⁱ⁾For

${\mathrm{TB}}^{\left(i\right)}=\frac{C_{\mathrm{est}}^{\left(i\right)}}{\mathrm{INST}\_C}\×\mathrm{INST}\_R$

By following formula as can be known, target bit amount TB ⁽ⁱ⁾With complexity C _Est ⁽ⁱ⁾And moment bit rate INST_R is directly proportional, and is inversely proportional to moment complexity INST_C.

All MPEG-2 data bit flow filaments all must be followed the VBV rule of Moving Picture Experts Group-2, and the target amount of distribution must be limited so that the VBV buffer can spill-over (overflow) or deficient (underflow).After the target bit amount of present picture determines, just can utilize the image coder of many different kenels, similarly be the test prototype 5 (Test Model 5) that Moving Picture Experts Group-2 is described, finish the compression of picture according to target bit amount.

Picture is after finishing coding at present, and its actual complexity can be learnt, and the full activity of present picture and the relation between the actual complex degree, ACR ⁽ⁱ⁾Can calculate from following

${\mathrm{ACR}}^{\left(i\right)}=\frac{A^{\left(i\right)}}{A{Q}^{\left(i\right)}\×U{B}^{\left(i\right)}}$

Wherein, ACR ⁽ⁱ⁾With full activity A ⁽ⁱ⁾Be directly proportional, and count UB with the actual consumption bit of present picture ⁽ⁱ⁾And average quantization step-length AQ ⁽ⁱ⁾Be inversely proportional to.This ACR ⁽ⁱ⁾Can bring and estimate next complexity with the picture of kenel.ACR ⁽ⁱ⁾Can and ACR ^(i-1)Do that linear junction is incompatible avoids being subjected to the influence that those are rich in the noise picture.

Single of the present invention (single-pass) video coding can be by Fig. 3 preferred embodiment and cooperate the operational flowchart of Fig. 4 to explain.As shown in Figure 3, dynamic image compression set 300 comprises and moves evaluation unit 310, a complexity estimator 320, a target bit allocator 330, a parameter update device 340 and an image coder 350.The present picture P that is about to be encoded in one group of continuous pictures in the mobile evaluation unit 310 receiver, video sequences ⁽ⁱ⁾, be used for when carrying out the block coupling that moves estimation, calculating this present picture P ⁽ⁱ⁾Full activity A ⁽ⁱ⁾(step S410).Based on full activity A ⁽ⁱ⁾And in this group continuous pictures with the ACR of the previous coding picture of kenel ^(i-1), complexity estimator 320 can be used to estimate the complexity C of present picture _Est ⁽ⁱ⁾(step S420).Target bit allocator 330 complexity C to estimate _Est ⁽ⁱ⁾Upgrade this group continuous pictures its moment complexity INST_C (step S430), simultaneously, target bit allocator 330 receives the statistics complexity STAT_C that this organizes continuous pictures from parameter update device 340, and organizes the set average bit rate ER of continuous pictures according to this _Mean, statistics complexity STAT_C, moment complexity INST_C, calculate the moment bit rate INST_R (step S440) of this group continuous pictures.

Target bit allocator 330 is then with moment bit rate INST_R, complexity C _Est ⁽ⁱ⁾And moment complexity INTC is that the basis is with target bit amount TB ⁽ⁱ⁾Distribute to present picture P ⁽ⁱ⁾(step S450).Therefore, image coder 350 is according to target bit amount TB ⁽ⁱ⁾And determine present picture P ⁽ⁱ⁾All quantization steps that use, these quantization steps are as picture P ⁽ⁱ⁾In each area of space: the usefulness of the coding of big block (MB).Image coder 350 is according to these quantization steps present picture P that encodes ⁽ⁱ⁾(step S460).After picture coding was finished at present, image coder 350 was with picture P ⁽ⁱ⁾The actual consumption bit count UB ⁽ⁱ⁾And the mean value AQ of actual used quantization step ⁽ⁱ⁾ Parameter update device 340 is given in repayment.Then with full activity A ⁽ⁱ⁾, actual consumption figure place UB ⁽ⁱ⁾Mean value AQ with the used quantization step of reality ⁽ⁱ⁾Be the basis, parameter update device 340 is present picture P ⁽ⁱ⁾Calculate the ratio ACR of its activity to complexity ⁽ⁱ⁾(step S470).

In the equivalence, the preferred embodiment of Fig. 3 can be considered with hardware and/or software and realize.According to the present invention, other assembly among mobile evaluation unit 310 and Fig. 3 can pipeline (pipeline) mode operating, before complexity estimator 320 begins to calculate the complexity of present picture, mobile evaluation unit 310 must be finished the calculating of present its full activity of picture that is about to be encoded and the estimation of motion-vector earlier, and, when mobile evaluation unit 310 was prepared for a following picture, complexity estimator 320 still just had been busy with finishing all and the relevant running of picture at present with other assembly.Next will describe each assembly in the preferred embodiment in detail.

One of main purpose of mobile evaluation unit 310 for decision with encode each MB in the picture of which kind of predictive mode, if necessary, also carry out forward and moving projection backward, also can from the block matching operation, extract the information of picture activity.At first, calculate the own activity amount (intra-activity) of each MB, with the pixel intensity of 48 * 8 brightness value blocks among the MB with Y _{M, n}, m=0 ..., 7, n=0 ..., 7 represent, and each its mean value of 8 * 8 blocks is Y, then the interior activity IntraAct of each MB:

$\mathrm{IntraAct}=\underset{k=0}{\overset{3}{\mathrm{\Σ}}}{\mathrm{\σ}}_k$

Wherein

${\mathrm{\σ}}_k=\sqrt{\underset{m=0}{\overset{7}{\mathrm{\Σ}}}\underset{n=0}{\overset{7}{\mathrm{\Σ}}}{(Y_{m,n}-\stackrel{\‾}{Y})}^2}$

If need lower computation complexity, the absolute difference of Y replaces relatively:

$\mathrm{IntraAct}=\underset{k=0}{\overset{3}{\mathrm{\Σ}}}{\mathrm{\Δ}}_k$

Wherein

${\mathrm{\Δ}}_k=\underset{m=0}{\overset{7}{\mathrm{\Σ}}}\underset{n=0}{\overset{7}{\mathrm{\Σ}}}|Y_{m,n}-\stackrel{\‾}{Y}|$

Because the MB in the I-picture only has a kind of pattern: the I pattern, so IntraAct is the activity of each MB of I-picture.

If picture is P-or B-picture, then need move estimation.The most normal making is used for finding that the technology of best motion-vector is the block coupling.For non-self coding (nonintra coding) (as P-and B-picture), utilize distortion criterion such as variance or distortion absolute value and minimize, select forward, backward, two-way prediction or do not need motion compensation.In case the MB pattern of each MB decision in P-or the B-picture, the variance of 48 * 8 blocks also can be tried to achieve in the difference MB of each motion compensation, and the difference MB of motion compensation is pending MB and estimates poor to pixel of pixel between MB; Therefore distortion absolute value and have preferable computational efficiency usually can be used to replace variance.With the variance of 48 * 8 blocks or distortion absolute value and addition mutual activity (inter-activity) in the hope of each MB in the coded picture in non-, InterAct, then the IntraAct of each MB in the coded picture in non-and InterAct thereof are brought and make comparisons, to judge whether InterAct is less, if, then with the activity of InterAct as this MB, and with mutual pattern (inter-mode) this MB that encodes; Otherwise with the activity of IntraAct, and with self pattern (intra-mode) this MB that encodes as this MB.At last, to present I-, P-or B-picture, with the activity addition of all MB and full activity A ⁽ⁱ⁾, mobile evaluation unit 310 is again full activity A ⁽ⁱ⁾Send complexity estimator 320 and parameter update device 340 to.

Then, complexity estimator 320 is the present picture P of certain kenel ⁽ⁱ⁾Estimate its complexity, and, introduce weight coefficient to the complexity C that estimates according to three kinds of picture kenels _Est ⁽ⁱ⁾Because the B-picture must not be used to estimate other picture, keep more bit to distribute less bit to give the B-picture to I-and P-picture so can reduce the weight coefficient of B-picture; Generally speaking, coding I-picture can produce maximum bits, so the weight coefficient of P-picture is again less than the weight coefficient of I-picture.Complexity estimator 320 upgrades complexity C according to the kenel of present picture _I, C _POr C _BThe three one of them, present picture P ⁽ⁱ⁾Complexity C _Est ⁽ⁱ⁾Estimation as follows:

If (I-picture)

$C_{\mathrm{est}}^{\left(i\right)}=C_I=K_I\×\frac{A^{\left(i\right)}}{{\mathrm{ACR}}_I^{(i-1)}}$

Else if (P-picture)

$C_{\mathrm{est}}^{\left(i\right)}=C_P=K_P\×\frac{A^{\left(i\right)}}{{\mathrm{ACR}}_P^{(i-1)}}$

Else if (B-picture)

$C_{\mathrm{est}}^{\left(i\right)}=C_B=K_B\×\frac{A^{\left(i\right)}}{{\mathrm{ACR}}_B^{(i-1)}}$

Wherein, ACR _I ^(i-1), ACR _P ^(i-1)And ACR _B ^(i-1)Be in one group of continuous pictures, the previous coding picture activity separately of I, P and Type B attitude is to the ratio of complexity.These ACR _I ^(i-1), ACR _P ^(i-1)And ACR _B ^(i-1)Leave among the parameter update device 340, and complexity estimator 320 can read the ratio of corresponding activity to complexity for the previous picture of suitable kenel.K _I, K _PAnd K _BBe respectively the weight coefficient of I-, P-and B-picture, its scope generally is between 0 to 1.0.As for the I-picture, preferred embodiment can be adopted K _I=1.0.

Target bit allocator 330 can be three kinds of kenels in one group of continuous pictures: I-picture, P-picture and B-picture store its complexity C respectively _I, complexity C _PWith complexity C _BWhen receiving C _Est ⁽ⁱ⁾The time, look present picture P ⁽ⁱ⁾Kenel, only upgrade complexity C at every turn _I, C _P, C _BThe three one of them, as follows:

If (I-picture)

$C_I=C_{\mathrm{est}}^{\left(i\right)}$

Else if (P-picture)

$C_P=C_{\mathrm{est}}^{\left(i\right)}$

Else if (B-picture)

$C_B=C_{\mathrm{est}}^{\left(i\right)}$

Preferred embodiment as shown in Figure 5, target bit allocator 330 by moment bit rate calculator 510, quota calculator 520 and limit unit 530 formed.Moment, bit rate calculator 510 was for this group continuous pictures upgrades its moment complexity INST_C, and is as follows:

INST_C＝N _I×C _I+N _P×C _P+N _B×C _B

Wherein, N _I, N _PAnd N _BIt is respectively the number of I-, P-and B-picture in one group of continuous pictures.At the same time, moment bit rate calculator 510 also read the statistics complexity STAT_C of this group continuous pictures from parameter update device 340, and calculate the moment bit rate INST_R of one group of continuous pictures according to following formula:

${\mathrm{INST}\_R=\mathrm{ER}}_{\mathrm{mean}}\·(1+\mathrm{SF}\·\left(\frac{\mathrm{INST}\_C}{\mathrm{STAT}\_C}\right)-1)$

Wherein, ER _MeanBe the set average bit rate of this group continuous pictures, SF is a zoom factor, and its scope is between 0 to 1; And moment bit rate INST_R preferably is limited in maximum, the minimum bit rate ER of this group continuous pictures _MaxAnd ER _MinBetween.

Quota calculator 520 then distributes target bit amount TB ⁽ⁱ⁾Give present picture P ⁽ⁱ⁾, and target bit amount TB ⁽ⁱ⁾It is basis

${\mathrm{TB}}^{\left(i\right)}=\frac{C_{\mathrm{est}}^{(i)}}{\mathrm{INST}\_C}\×\mathrm{INST}\_R$

But, actual consumption bit number can't just equate with the bit amount of expectation, therefore develop and a kind ofly feedback the bit number that strategy makes the picture actual consumption and approach target bit amount, in preferred embodiment, be encoding to the super bit amount of using so far, can be by the present picture amortization some of encoding:

TB ⁽ⁱ⁾＝TB ⁽ⁱ⁾-AR×EB

Wherein, EB is that parameter update device 340 transmits, and super to use bit amount, AR then be set amortization factor, and its scope generally is between 0.05 to 0.2.According to the formula of front, target bit amount TB ⁽ⁱ⁾Be complexity C with present picture _Est ⁽ⁱ⁾And the moment bit rate INST_R of one group of continuous pictures is directly proportional, and is inversely proportional to the moment complexity INST_C of one group of continuous pictures.530 of limit unit can be used for adjusting target bit amount TB ⁽ⁱ⁾To meet the VBV standard, the upper limit (U_bound) and the lower limit (L_bound) of amount have therefore been fixed in advance.With regard to the operation of fixing bit rate (CBR), the target bit amount of distributing to a picture must make the VBV buffer can spill-over or scarcity, so limit unit 530 is with target bit amount TB ⁽ⁱ⁾Be limited in the upper and lower limit scope:

if(TB ⁽ⁱ⁾＞U_bound)?then?TB ⁽ⁱ⁾＝U_bound

if(TB ⁽ⁱ⁾＜L_bound)?then?TB ⁽ⁱ⁾＝L_bound

If variable bit rate (VBR), as long as prevent that then VBV is deficient, so:

if(TB ⁽ⁱ⁾＞U_bound)?then?TB ⁽ⁱ⁾＝U_bound

Then, target bit amount TB ⁽ⁱ⁾Can be sent to parameter update device 340 and image coder 350.

Get back to Fig. 3, image coder 350 sees through signal line 312 receives picture data and each MB from mobile evaluation unit 310 motion-vector and MB pattern.In order to make picture reach best in quality as much as possible, image coder 350 is according to target bit amount TB ⁽ⁱ⁾Decide the used quantization step of each MB coding of present picture, based on these data, image coder 350 begins present picture P ⁽ⁱ⁾Encode and output encoder after data CD.At present picture P ⁽ⁱ⁾After the coding, image coder 350 is done used quantization step average and is calculated the actual consumption bit and count UB ⁽ⁱ⁾, repay the UB of present picture again ⁽ⁱ⁾And the mean value AQ of actual used quantization step ⁽ⁱ⁾Give parameter update device 340.

Parameter update device 340 can be three kinds of kenels in one group of continuous pictures: I-picture, P-picture and B-picture store its complexity statistic SC respectively _I, SC _PWith SC _B, and get the complexity C of picture _Est ⁽ⁱ⁾Upgrade in these three kinds of kenels with the complexity statistic of kenel picture, as follows:

$S{C}_X=\frac{S{C}_X\·(W_X-1)+C_{\mathrm{est}}}{W_X}$

Wherein, the I of subscript x representative, P and Type B attitude are looked closely present picture P ⁽ⁱ⁾Kenel and decide W _I, W _PAnd W _BIt is respectively the set weight coefficient of I-picture, P-picture and B-picture.In other words, according to present picture P ⁽ⁱ⁾Kenel, complexity statistic SC _I, SC _POr SC _BHave only one of them to be updated at every turn:

If (I-picture)

$S{C}_I=\frac{S{C}_I\·(W_I-1)+C_{\mathrm{est}}^{\left(i\right)}}{W_I}$

Else if (P-picture)

$S{C}_P=\frac{S{C}_P\·(W_P-1)+C_{\mathrm{est}}^{\left(i\right)}}{W_P}$

Else if (B-picture)

$S{C}_B=\frac{S{C}_B\·(W_B-1)+C_{\mathrm{est}}^{\left(i\right)}}{W_B}$

Wherein, W _I, W _PAnd W _BScope generally be between 200 to 2000.Parameter update device 340 then calculates the statistics complexity STAT_C of one group of continuous pictures with following equation,

STAT_C＝N _I×SC _I+N _P×SC _P+N _B×SC _B

Wherein, N _I, N _PAnd N _BIt is respectively the number of I-, P-and B-picture in every group of continuous pictures.Picture P ⁽ⁱ⁾Target bit amount TB ⁽ⁱ⁾And the actual consumption bit is counted UB ⁽ⁱ⁾Between gap, can by parameter update device 340 do accumulative total so that at picture P ⁽ⁱ⁾Obtain super behind the coding with bit amount EB:

EB＝EB×(1-AR)+UB ⁽ⁱ⁾-TB ⁽ⁱ⁾

Wherein, AR is set amortization factor.Therefore, the relation between activity and the actual complex degree, ACR ⁽ⁱ⁾, can calculate from following

$\mathrm{AC}{R}^{\left(i\right)}=\frac{A^{\left(i\right)}}{A{Q}^{\left(i\right)}\×U{B}^{\left(i\right)}}$

This has just been encoded and has finished picture P ⁽ⁱ⁾Affiliated ACR ⁽ⁱ⁾Can be used to estimate next complexity with the kenel picture.For making ACR ⁽ⁱ⁾Can not be subjected to being rich in the influence of noise picture, preferred embodiment has utilized ACR ⁽ⁱ⁾And ACR ^(i-l)Linear combination, for example:

If (I-picture)

$\mathrm{AC}{R}_I^{\left(i\right)}=\mathrm{AC}{R}_I^{(i-1)}\×(1-\mathrm{CW})+\mathrm{AC}{R}^{\left(i\right)}\×\mathrm{CW}$

Else if (P-picture)

$\mathrm{AC}{R}_P^{\left(i\right)}=\mathrm{AC}{R}_P^{(i-1)}\×(1-\mathrm{CW})+\mathrm{AC}{R}^{\left(i\right)}\×\mathrm{CW}$

Else if (B-picture)

$\mathrm{AC}{R}_B^{\left(i\right)}=\mathrm{AC}{R}_B^{\left(i\right)}\×(1-\mathrm{CW})+\mathrm{AC}{R}^{\left(i\right)}\×\mathrm{CW}$

Wherein, CW is the set weight coefficient of linear combination.The ACR that upgraded _I ⁽ⁱ⁾, ACR _P ⁽ⁱ⁾And ACR _B ⁽ⁱ⁾Can be sent to complexity estimator 320,, in addition, surpass with bit amount EB and statistics complexity STAT_C and then deliver to the usefulness of target bit allocator 330 as the position distribution so that estimate its complexity for next picture of opening suitable kenel.

Though the present invention discloses as above with a specific embodiment; so it is only in order to be easy to illustrate technology contents of the present invention; and be not with narrow sense of the present invention be defined in this embodiment; anyly have the knack of this skill person; without departing from the spirit and scope of the present invention; when can doing a little change and retouching, so protection scope of the present invention is as the criterion when looking defining of claim.

Claims (10)

1. the variable position-element rate control device of a dynamic image compression is characterized in that, comprises at least:

One moves evaluation unit, receives a present picture that is about to be encoded in one group of continuous pictures in the video sequence, to calculate a full activity of this present picture;

One complexity estimator, based on this full activity of this present picture and should the group continuous pictures in together its activity of previous coding picture of kenel the ratio of complexity is estimated a complexity of this present picture;

One target bit allocator, upgrade the complexity in a flash of this group continuous pictures with this complexity of this present picture, and based on a set average bit rate of this group continuous pictures, a statistics complexity and this moment complexity calculate the rate of bit in a flash of this group continuous pictures, and according to this complexity of this present picture, this group continuous pictures this moment complexity and this moment the bit rate, distribute a target bit amount to give this present picture; And

One parameter update device, according to this full activity of this present picture, an actual consumption bit number of this present picture and a mean value of the actual used quantization step of this present picture of encoding, calculate the ratio of this its activity of present picture, and calculate this statistics complexity of this group continuous pictures based on this complexity of this present picture complexity;

Wherein, this activity of this present picture is directly proportional with this full activity of this present picture to the ratio of complexity, and is inversely proportional to this actual consumption bit number of this present picture and this mean value of the actual used quantization step of this present picture.

2. the variable position-element rate control device of dynamic image compression according to claim 1, it is characterized in that, also comprise an image coder at least, determine the employed a plurality of quantization steps of above-mentioned present picture according to the above-mentioned target bit amount of above-mentioned present picture, described quantization step is the usefulness as each the area of space coding in the above-mentioned present picture, this image coder is according to the quantization step above-mentioned present picture of encoding, and the above-mentioned actual consumption bit number of above-mentioned present picture and the above-mentioned mean value of the actual used quantization step of above-mentioned present picture are repaid to the above-mentioned parameter renovator.

3. the variable position-element rate control device of dynamic image compression according to claim 1 is characterized in that, the above-mentioned complexity of the estimated above-mentioned present picture that obtains of above-mentioned complexity estimator is according to an equation: $C_{\mathrm{est}}=K\×\frac{A}{\mathrm{ACR}}$

Wherein, C _EstBe the above-mentioned complexity that above-mentioned present picture is estimated, K is a set weighted value, and its scope is between 0 to 1, and A is the above-mentioned full activity of above-mentioned present picture, and ACR is the ratio of above-mentioned its activity of previous coding picture with kenel to complexity.

4. the variable position-element rate control device of dynamic image compression according to claim 2 is characterized in that the above-mentioned parameter renovator calculates the ratio of above-mentioned its activity of present picture to complexity, is foundation:

$\mathrm{ACR}=\frac{A}{\mathrm{AQ}\×\mathrm{UB}}$

Wherein, ACR represents the ratio of above-mentioned its activity of present picture to complexity, and A is the above-mentioned full activity of above-mentioned present picture, and AQ is the above-mentioned mean value of the actual used quantization step of above-mentioned present picture, and UB is the above-mentioned actual consumption bit number of above-mentioned present picture.

5. the variable position-element rate control device of dynamic image compression according to claim 1 is characterized in that above-mentioned target bit allocator is three kinds of kenels in above-mentioned one group of continuous pictures: I-picture, P-picture and B-picture store its complexity C respectively _I, complexity C _PWith complexity C _B, and the above-mentioned complexity of getting above-mentioned present picture is upgraded in these three kinds of kenels the complexity with the kenel picture.

6. the variable position-element rate control device of dynamic image compression according to claim 5 is characterized in that above-mentioned target bit allocator comprises at least:

Bit rate calculator in a flash, above-mentioned complexity based on above-mentioned present picture, upgrade above-mentioned moment complexity of above-mentioned one group of continuous pictures, and calculate above-mentioned moment bit rate of above-mentioned one group of continuous pictures according to a function, wherein this function depends on the above-mentioned set average bit rate of above-mentioned one group of continuous pictures, above-mentioned moment complexity and above-mentioned statistics complexity;

One quota calculator, based on the above-mentioned complexity of above-mentioned present picture, above-mentioned one group of continuous pictures above-mentioned moment the bit rate and above-mentioned moment complexity, distribute above-mentioned target bit amount;

And a limit unit, be used for above-mentioned target bit amount is limited under the set upper limit.

7. the variable position-element rate control device of dynamic image according to claim 6 compression is characterized in that, above-mentioned moment bit rate calculator above-mentioned one group of continuous pictures of trying to achieve above-mentioned moment complexity, be basis: INST_C=N _I* C _I+ N _P* C _P+ N _B* C _B

Wherein, INST_C represents above-mentioned moment complexity, N _I, N _PAnd N _BIt is respectively the number of above-mentioned I-, P-and B-picture among above-mentioned one group of continuous pictures.

8. the variable position-element rate control device of dynamic image according to claim 6 compression is characterized in that, above-mentioned moment bit rate calculator calculate above-mentioned moment bit rate of above-mentioned one group of continuous pictures, be basis: $\mathrm{INST}\_R=E{R}_{\mathrm{mean}}\·(1+\mathrm{SF}\·(\frac{\mathrm{INST}\_C}{\mathrm{STAT}\_C}-1))$

Wherein, ER _MeanIt is the above-mentioned set average bit rate of above-mentioned one group of continuous pictures, SF is a zoom factor, its scope is between 0 to 1, and INST_C, STAT_C are respectively above-mentioned moment complexities and above-mentioned statistics complexity of above-mentioned one group of continuous pictures, and INST_R represents above-mentioned moment bit rate.

9. the variable position-element rate control device of dynamic image compression according to claim 1 is characterized in that the above-mentioned parameter renovator is three kinds of kenels in above-mentioned one group of continuous pictures: I-picture, P-picture and B-picture store its complexity statistic SC respectively _I, complexity statistic SC _PWith complexity statistic SC _B, and the above-mentioned complexity of getting above-mentioned present picture upgrades in these three kinds of kenels with the complexity statistic of kenel picture, as follows: ${\mathrm{SC}}_X=\frac{{\mathrm{SC}}_X\·(W_X-1)+C_{\mathrm{est}}}{W_X}$

Wherein, I, the P of this subscript x representative and Type B attitude are looked closely the kenel of above-mentioned present picture and are decided W _I, W _PAnd W _BBe respectively the set weight coefficient of I-picture, P-picture and B-picture, C _EstIt is the above-mentioned complexity of above-mentioned present picture.

10. the variable position-element rate control device of dynamic image compression according to claim 9 is characterized in that the above-mentioned parameter renovator calculates the above-mentioned statistics complexity of above-mentioned one group of continuous pictures, is basis:

STAT_C＝N _I×SC _I+N _P×SC _P+N _B×SC _B

Wherein, STAT_C represents above-mentioned statistics complexity, N _I, N _PAnd N _BIt is respectively the number of above-mentioned I-, P-and B-picture among above-mentioned one group of continuous pictures.

Images