CN1131638C

CN1131638C - Video signal encoding method and appartus employing adaptive quantization technique

Info

Publication number: CN1131638C
Application number: CN 98100935
Authority: CN
Inventors: 菅原隆幸; 上田基晴; 中川裕史; 西谷胜义; 藤原光章
Original assignee: Victor Company of Japan Ltd
Current assignee: JVCKenwood Corp
Priority date: 1998-03-19
Filing date: 1998-03-19
Publication date: 2003-12-17
Anticipated expiration: 2018-03-19
Also published as: CN1230080A

Abstract

The present invention relates to a method adopted in an encoding device, which is used for allocating bit to each picture of a video signal according to quantization relevant signals and quantization irrelevant signals. Firstly, the quantization relevant signals are quantified according to fixed quantization parameters; secondly, quantized data and the quantization irrelevant signals are statistically encoded to generate first data and second data thereby. The encoded bit target amount of each picture and the encoded bit target amount of each block are determined according to a target bit rate, the amount of the first data and the amount of the second data. The encoded data amount of the pictures and the encoded data amount of the blocks are controlled in a self-adaptive mode according to the encoded bit target amount by adjustably changing the quantization parameters according to the difference between the encoded bit target amount and the amount of generated encoded data.

Description

Adopt a kind of video Signal encoding method and equipment of adaptive quantizing technology

Technical field

The present invention relates to a kind of video Signal encoding method and equipment; More particularly, relate to a kind ofly be used for making vision signal pass through two (promptly first and second) coding paths and come this vision signal is carried out the method and apparatus of adaptive coding according to the bit distribution information of estimating, wherein the first coding path provide regulate second encode in the path quantizing process the time bit distribution information that will use.

Background technology

A kind of coding method and equipment have been proposed, be used for the vision signal of importing being quantized in first and second paths and adjustable length coding according to for example MPEG (dynamic menu expert group) standard, wherein first path obtains the required bit distribution information of cataloged procedure of carrying out in second path, the bit distribution information of resulting each picture constitutes a sequence of video signals and each equal-sized (for example 16 * 16 pixels) in each picture, referring to the unsettled 7-284097 publication application of Japan.

In the video signal coding technology such as MPEG of routine, when if vision signal is variable length code, particularly this vision signal is after being quantized with a fixing quantization parameter (Qp) or quantum step size (Qs), and amount of coded data just increases along with the complexity of vision signal and dynamic compensation error.Utilize this feature, the encoding scheme of prior art distribution or the bit distribution information of fixing Qp inspection in the data that are encoded in first path; And based on this bit that can use in second path allocation, these bits are determined in the predicted value of the target bit rate of picture and interblock, so that uniform and high-quality picture to be provided.According to coming allocation bit according to distributed data between picture, just can provide the time to go up uniform image quality.Similarly, by the interblock allocation bit in each picture, just can obtain uniform image quality on the space.

The Qp in first path is set as little than those second paths, so that the data volume that produces in first path is controlled, makes it bigger than the data volume that is produced in second path.By use the less Qp of amplitude in first path, the radio-frequency component in the vision signal just can be detected more completely, and the characteristic of vision signal can check more accurately that also this effective Bit Allocation in Discrete for second path is essential.

In a kind of typical prior art scheme, the Bit Allocation in Discrete in second path is controlled with about inversely prroportional relationship of be encoded data and Qp or Qs according to bit distribution information and a predetermined target bit rate from first path.In other words, the distributive law Be Controlled of the coded-bit of the second path picture and piece is kept a roughly the same ratio so that the coded-bit of those pictures in first path and piece is distributed.The control of the data volume that is encoded that second path is produced is undertaken by changing Qp.

But, even the amount of some coded data also remains unchanged when Qp changes.Those be add or with quantize irrelevant codified data, as header information among the MPEG and motion vector, they are the variable length codes that are not quantized.

The amount of this additional data depends on the size of picture.For example, for the picture that 720 * 480 pixels are arranged with MPEG 2 schemes coding, the additional data of average generation 20-50K bit.

In Fig. 7 A and 7B, the data volume that is produced by first and second paths for each picture has been described respectively, wherein each bar is represented the data total amount that produces for picture, and dash area is wherein represented to add or with quantize irrelevant data volume.

And quantizing the irrelevant data volume and the beguine of data total amount has nothing in common with each other between picture according to the actual distribution of the bit that is distributed.For the transfer rate of 6Mbps, in the data total amount that is produced by first path, account for 20% with quantizing irrelevant data, shown in Fig. 7 A.But, because second path with than the low average transfer rate in first path to video signal coding, in second path, just become big with the occupancy that quantizes irrelevant data.For example, to account for by target bit rate be 30% of the data total amount that produced of second path of 4Mbps with quantizing irrelevant data volume, shown in Fig. 7 B.

With reference to figure 8A and 8B, they have illustrated the amount of coded data that is produced for each piece by first path and second path respectively, and wherein each root stick is expressed as a data volume that piece produced, and its dash area is represented in the piece and quantize irrelevant data volume.Similarly, in first path, the occupancy of the additional data in some piece in total data is greater than 50%, shown in Fig. 8 A; Be lower than first path because the average transmission rate in second path is controlled in, additional amount of data has taken 70% of the data total amount that produced by second path in some piece, shown in Fig. 8 B.

But, in the Bit distribution method of above-mentioned routine, because being based on by the definite bit distribution information of the total amount of the data that are encoded (comprising relevant with quantification and irrelevant data), this Bit Allocation in Discrete carries out, the position distribution result do not reflect Qp and and quantize inverse relation between the related data amount, cause the local shortage that is assigned with bit of some picture or piece, this may cause picture quality degradation.

Summary of the invention

Therefore, one object of the present invention a kind ofly can be improved image quality and not have the method and apparatus that is assigned with the bit shortage problem with regard to providing by allocation bit.

According to an aspect of the present invention, a kind of encoding video signal method to input is provided, the vision signal that this method comprises the steps: to handle input according to predetermined way is with the signal after providing to handle thus, processed signal comprise one with quantize relevant signal and one and the signal that has nothing to do of quantification; Quantize to provide first to quantize the back data thus according to a fixing Qp pair of signal relevant with quantification; Carry out variable length code to the first quantification back data with the data that quantize to have nothing to do, so that first and second data to be provided respectively thus; First data volume and second data volume are stored; Determine the aim parameter of the coded-bit of each video signal unit according to predetermined target bitrate and first, second data volume; According to quantizing the signal relevant with quantification based on the determined one or more Qp of coded-bit aim parameter, so that the data that are encoded to be provided thus, wherein second quantized data is adjusted according to the difference of the target code bit quantity and the data volume that is encoded.

The video signal unit that its coded-bit aim parameter is determined can be picture and/or piece.

Another aspect provides a kind of video signal encoding device according to the preferred embodiment of the invention, comprises one first code device, a memory, a delayer and one second code device.

First code device comprises one first signal processor, is used for handling by predetermined way the vision signal of input, with produce one comprise with quantized signal irrelevant and with quantized signal relevant processed video signal; One first quantizer; One first variable length code (VLC) circuit; One first counter.First quantizer quantizes according to a fixing Qp and quantizes relevant signal, with the data after providing first to quantize thus.A VLC circuit is encoded to first quantized data with the quantification irrelevant signal respectively with the VLC technology and is produced first and second data then.At first counter, first and second data volumes are counted respectively, and first data volume is relevant with the amplitude of Qp, and the amplitude of second data volume and Qp is irrelevant.

Memory stores first and second data amount separately; Delayer is then synchronous to carry out to the video signal delay of input.In other words, the vision signal of input is delayed device and postpones to give second code device then till first and second data volumes are stored into memory.

Second code device comprises a secondary signal processor, and a target bits is determined circuit, a bit controller, one second quantizer, one the 2nd VLC circuit and one second counter.The incoming video signal of second processor after according to the mode processing delay identical with first processor is to produce the processed video signal.Target bits determines that circuit determines the aim parameter of the coded-bit of each video signal unit according to predetermined target bitrate and first and second data amount separately.The bit controller is determined the one or more new Qp of each unit according to the aim parameter of coded-bit.Second quantizer quantizes according to the new Qp that determines and quantizes relevant signal to produce second quantized data.Then, the data after the 2nd VLC circuit quantizes second and with quantize irrelevant signal encoding to produce the 3rd data and the 4th data respectively.Two counters are to each amount counting of third and fourth data; The bit controller calculate the be encoded aim parameter and third and fourth data volume and between deviation, thus by be controlled at the Qs of second quantizer adaptively according to this bias adjustment Qp.

According to said method of the present invention and equipment, the vision signal of input is quantized with fixing Qp and variable length code in first code device, then, the data of variable length code deposit in the memory respectively according to first data volume (depending on Qp) and second data volume (irrelevant with Qp).Then, at second code device, determine the aim parameter of each video signal unit coded-bit according to the amount that deposits in of first and second data.Because at the Qp of second code device is to adjust according to the deviation between the actual amount of the aim parameter of coded-bit and the length varying coded data that produced by second code device, thereby approaching between the actual amount of the data that are encoded and the aim parameter of coded-bit, consider with second data that quantize to have nothing to do and the aim parameter of definite coded-bit can be in second path optimized distribution.

According to another preferred embodiment of the invention, video signal encoding device be by with public part (i.e. the signal processor of first and second code devices, quantizer, VLC circuit and bit counter among first embodiment) thus shared minimizing part count realize.In this case, just constitute the equipment of second embodiment, it has a signal processor, a quantizer, a VLC circuit, a bit counter, a memory, one determines that according to predetermined target bitrate and first and second data amount separately the target bits of each cell encoding bit aim parameter of vision signal determines circuit, one determines initial Q p and further finds out between the actual amount of coded-bit aim parameter and length varying coded data deviation to control the bit controller of Qp (it is the Qs of adaptively modifying quantizer successively) thus adjustably according to the coded-bit aim parameter, be used to provide the fixedly Q of Qp _FixCircuit and a switching circuit take place.

In the structure of the second embodiment equipment, at first, fixing Qp is added in the quantizer by switching circuit, makes the incoming video signal of handling in signal processor quantize by the fixing determined Qs of Qp thus; The data volume that is encoded that is produced is counted and stores.Then, the stored data volume that is encoded is given target bits and is determined circuit, to determine the aim parameter of each cell encoding bit.The aim parameter of coded-bit offers the bit controller again, and Qp is adjusted the Qs with the control quantizer adaptively there.

Brief description of drawings

Above-mentioned purpose of the present invention, feature and other purpose, feature are passed through will be apparent below in conjunction with the accompanying drawing description of a preferred embodiment.In the accompanying drawings,

Shown in Figure 1 is block diagram according to the video signal encoding device of first preferred embodiment of the invention;

Fig. 2 is the block diagram of first code device shown in Fig. 1;

Fig. 3 is the block diagram of bit counter shown in Fig. 2;

Fig. 4 is the example of the first code device data memory format shown in Fig. 2;

Fig. 5 is the block diagram of second code device shown in Fig. 1;

Fig. 6 is the block diagram according to the video signal encoding device of second preferred embodiment of the invention;

Fig. 7 A and 7B provide respectively picture and quantize related data and and quantize to carry out between extraneous data before and after the Bit Allocation in Discrete the example that concerns between two kinds of data;

Fig. 8 A and 8B provided respectively piece and quantize related data and and quantize to carry out between extraneous data before and after the Bit Allocation in Discrete the example that concerns between two kinds of data.

Fig. 9 is the block diagram of the 3rd embodiment of device of the present invention.

Preferred embodiment

With reference to figure 1, it has described the block diagram according to the video signal encoding device of first preferred embodiment of the invention.Of the present invention first and other preferred embodiment in this explanation adopt the mpeg encoded method.

In MPEG, sequence of video signals is divided into picture group (GOP).As everyone knows, picture can be represented with the frame or the field of vision signal; And term " picture " is more more commonly used than term " frame " or " field " in technology.GOP is a group of pictures adjacent on display order, is minimum coding unit, and it can be decoded independently in a sequence.GOP comprises the precoding picture (P-picture) and the bi-directional predictive coding picture (B-picture) of at least one interior coded picture (I-picture) and arbitrary number.Minimum coding unit is the piece of 8 * 8 pixels.In 4: 2: 0 forms, four luminance block of believing and two chrominance block (each expression and four corresponding dissimilar luminance signals of luminance block) have constituted a macro block, and it is the minimum unit of estimation and compensation; Several macro blocks constitute a picture or one page.About the detailed content of MPEG, please refer to the ISO-IEC11172-2 of International Organization for Standardization and the ITU-TH.262/ISO-IEC13818-2 of International Telecommunication Union.

Give first code device 1 and the delay buffer 3 that constitutes one first path with reference to figure 1, one input digital video signal again.At first code device 1, vision signal quantizes with fixed quantisation parameter (Qp) and encodes with statistical technique.Then, be included in being counted respectively in each vision signal scheduled unit coded data with the quantification related data with quantification extraneous data amount.Offer memory 2 and therein preservation as bit distribution information from first code device 1 with the quantification related data with the counting that quantizes extraneous data.

Vision signal postpones a period of times by delay buffer 3, this time with obtain bit information and deposit in the memory 2 it in needed time consistent, send into second code device 4 in formation second path again.

At second code device 4, the coded-bit aim parameter of each picture is according to determining with a kind of algorithm that the back will illustrate that from the target bit rate and the bit distribution information of memory 2 aim parameter of the macroblock encoding bit that comprises is then determined according to the bit distribution of each picture and the aim parameter of coded-bit in each picture.After this, the quantization step of being determined with picture and macroblock encoding bit aim parameter by the vision signal of delay buffer 3 (Qs) quantizes.

When the aim parameter of the coded-bit of determining picture, the predetermined collections of pictures (as GOP) that comprises picture is come total aim parameter of the calculation code bit first time with respect to the ratio of unit frame numbers in the time according to frame numbers in target bit rate and the set.The data total amount with quantizing to have nothing to do of picture is just deducted from the catalogue scalar of coded-bit in the set then, so that the coded-bit aim parameter relevant with quantification of set to be provided thus.Picture with quantize relevant coded-bit aim parameter according to set with quantize correlative coding bit aim parameter and picture with quantize the related data amount with respect to set with the calculating recently that quantizes the related data total amount.The coded-bit aim parameter of final picture by with picture with quantize the related data amount and with quantize correlative coding bit total amount and add and obtain.

Similarly, the macroblock encoding bit aim parameter in a picture is determined as follows.At first, macro block with quantize relevant coded-bit aim parameter according to picture with quantize correlative coding bit aim parameter and macro block with quantize the related data amount with respect to picture with the calculating recently that quantizes the related data total amount.Then by with macro block with quantize correlative coding bit aim parameter and macro block with the quantification extraneous data amount coded-bit aim parameter of Calais's computing macro block mutually.

According to the present invention, as considering that the additional data relevant with quantification determine the result of coded-bit aim parameter, the bit distribution between picture and macro block can be finished under the situation that is not subjected to the bit local defect.

First preferred embodiment of the present invention will illustrate in greater detail referring to figs. 2 to Fig. 5.

With reference to figure 2, this figure has illustrated the details of first code device 1 shown in Fig. 1.According to the type of picture in the writing circuit (not shown), one comprises that the digital image signal of luminance and chrominance information is recorded by a coded sequence.The vision signal that is recorded is given subtracter 10 and estimation and motion compensation (ME ﹠amp as the digitized video of an input; MC) circuit 11.At subtracter 10, by from the digitized video of input, deducting ME ﹠amp; The motion compensation signal that MC circuit 11 produces and produce a deviation signal.This deviation signal offers converter 12, it is that the basis is transformed with piece one (block-by-block) that deviation signal is used for example discrete cosine transform (DCT) technology therein, each piece of deviation signal is transformed to one group of conversion coefficient, comprises AC (interchange) coefficient of the high fdrequency component of pixel in DC (direct current) coefficient and a plurality of (for example 63) the expression piece of average pixel value of an expression piece.All conversion coefficients are quantized in follow-up cataloged procedure, and therefore, they are relevant with quantification.

Fixing Qp quantization transform coefficient of one quantizer (Q), 13 usefulness is to provide coefficient after the quantification to 14 and statistics devices of inverse quantizer (IQ) (VLC) 18.As in common knowledge, the Qs of the quantization parameter in quantizer determines that amplitude is based on that Qp and quantization matrix calculate; Therefore between piece, add a unique Qs, but can reach 64 at the number of the inside of piece Qs.

Thereafter, because I picture and P picture as ME ﹠amp; Quilt re-quantization and the inverse transformation successively in IQ14 and inverse converter (IT) 15 respectively of reference picture among the MC, quantization parameter correspondingly.One adder 16 is with ME﹠amp; The inverse transformation coefficient addition that motion compensation signal that MC circuit 11 produces and IT 15 produce produces the vision signal of a reorganization thus, and is identical with the signal of receiving terminal decoding.Chong Zu vision signal is existed in the picture memory 17 then.It should be noted that the corresponding motion compensation signal in inside with an I picture or piece is a null value.

At ME ﹠amp; MC circuit 11, the reorganization reference signal on the basis of for example macro block-macro block (macroblock-by-macroblock) in the reference picture store 17 is carried out the estimation and the compensation of the input digital video signal of I-or B-picture with conventional estimation and compensation technique.Each macro block of input digital video signal is at ME ﹠amp; The output of MC circuit 11 is to deliver to after the motion compensation of going in subtracter 10 and the adder 16 signal and deliver to the motion vector prediction mode signal that goes in the statistics device 18.This motion vector information can comprise two motion vectors that are used for B-picture macro block.Indicating this macro block with representative is the still predictive mode signal of the information of mutual coding (inter-coded) of interior coding (intra-coded).If this predictive mode signal can comprise also when B-picture macro block is had to a motion vector that this macro block is an information of making estimation forward or backward.These additional datas that comprise motion vector information and predictive mode signal are not quantized, so its coded-bit amount is irrelevant with quantification.

Statistics device 18 will be from coefficient and ME ﹠amp after the quantification of quantizer 13 order inputs according to for example conventional variable length code technology (will distribute to the normal input data that occur than short code word); The motion vector information of MC circuit 11 and predictive mode signal encoding.

Know altogether in the technical staff with reference to the function and the feature of the described parts 10 to 18 of Fig. 2 above, except the fixing Qp of basis in quantizer 13 carries out the quantification of conversion coefficient, at MPEG SupraIn more detailed description is all arranged.Also have, according to the present invention, statistics device 18 provides and quantizes relevant data to bit counter 20, promptly by statistical conversion coefficient, with additional with quantize irrelevant data, comprise by statistical motion vector information, predictive mode signal and other control bit as beginning code, sequence header or the like.In the output of static coding device 18, data such as statistical motion vector information, the predictive mode signal behind this static coding that relates to macro block and be used to comprise that the be encoded control bit etc. of macro block of block graphics (coded-block-pattern) is sent to bit counter 20 according to the basis of macro block-macro block.

With reference to figure 3, the detailed block diagram that it has illustrated bit counter 20, comprise data counts circuit 22 that a switch 21, is relevant with quantizing (Q) and with quantize (Q) extraneous data counting circuit 23.Switch 21 in to its input, discern with quantize related data and with quantize extraneous data, and the former with the latter is offered Q-relevant data counting circuit 22 and Q-extraneous data counting circuit 23 respectively.In

counting circuit

22 and 23, their input is counted respectively.

By bit counter 20 that obtain with quantize relevant and be admitted to memory shown in Figure 12 with quantification extraneous data amount separately, and store therein as bit distribution information.Memory 2 can be realized with a hard dick or any other high speed storing medium.

Can preserve by its former state the amount that each macro block produces, or divided by a predetermined number, as 256, before storage, to reduce its precision.Also have, suppose that the data volume that whole image produces is with 2 ^NRepresent that the data volume that each macro block produces can be with 2 ^NCome standardization so that preserve, N is a positive number.Some advantages have been brought on the calculation rate of standardization to given bit aim parameter in second code device 4 shown in Figure 1, because denominator can shift operation.

In the preferred embodiment that here provides, 2 ^NBe assumed to 2 ¹¹, promptly 2048.

The data volume that each macro block produces is expressed as 5, that is to say that its value is 0 to 31.Because the data volume that needed information is macro block to be produced is the order sum in code device, if the data volume that macro block produces greater than 31/2048, that macro block promptly is saved as maximum 31, remainder just is added in the data volume that its next macro block produces.The data volume that each macro block produces is according to storing respectively with the quantification related data with quantification extraneous data amount.

Similarly, the size of code that each picture produces can be preserved and not change by its former state, or scheduled number for example 2 ⁸Remove (=256), to reduce its precision before preserving.At this in a flash, suppose that the maximum data total amount that a picture produces is 1Mb, need 20 memory, storing this data volume with the lower accuracy after moving 8 only needs 12.Also similar under the situation of macro block, each picture with quantize the related data amount and with quantize the extraneous data amount and also be saved respectively.

When the storage data of each picture was standardized data, the data volume that each picture produces also must store.In another example, the data volume of picture does not need to be stored, because add and can calculate this data volume by the data volume with piece in the picture.

With reference to figure 4, it has illustrated the storage format of the data volume that produces.At first, each picture with quantize related data amount (" PIC-ACDC ") and with quantize extraneous data amount (" PIC-ETC ") respectively with 12 preservations.Each thereafter with 5 preservations be in for example 1350 pieces that in picture, comprise each piece with quantize related data amount (" BLOCK-ACDC ") and with quantize extraneous data amount (" BLOCK-ETC ").

With reference to figure 5, it is a more detailed block diagram of second code device 4 shown in Figure 1.Subtracter 30, estimation and compensation (the ME ﹠amp that second code device 4 is included; MC) function of circuit 31, converter 32, quantizer (Q), inverse quantizer (IQ) 34, inverse converter (IT) 35, adder 36, picture memory 37, statistics device 38 is identical with first code device, 1 corresponding part shown in Figure 2 with feature; Therefore, detailed description thereof will not illustrated for succinct consideration.

According to the present invention, second code device 4 comprises that also a target bits determines circuit 40, is used for determining each picture and macroblock encoding bit aim parameter according to target bit rate and bit distribution information (being stored in memory shown in Figure 1 2); One bit counter 41 is used for the statistical data amount counting to statistics device 38; One quantizer (Q) 33; And one transmit buffer 39.The amount of coded data of the bit counter 41 of bit control circuit 42 each picture of calculating and macro block and the deviation between the coded-bit aim parameter, and adjust Qp in view of the above to change the Qs in the quantizer 33, control the coded-bit amount that produces according to the coded-bit aim parameter thus.

A kind of algorithm that is used for determining in target bits the preferred embodiment of piece 40 definite target bits amounts will be described below.According to fixing Qp, the bit distribution information that Qfix produces is according to PIC_ACDC by first code device 1, PIC_ETC, and BLOCK_ACDC, the form of BLOCK_ETC exists in the memory shown in Figure 12.Qfix preferably is made as minimum value Qp, and for example " 1 ", it is not more than any possible Qp that uses in second code device, accurately to estimate the complexity of video and motion compensation deviation.

In order to determine the aim parameter of a picture coding bit, the coded-bit aim parameter of the picture group (for example, GOP) that each is predetermined is at first determined.The aim parameter of each group can according to target bit rate and in a group ratio between the sampling rate of number of pictures and vision signal calculate.

The coded-bit aim parameter TARGET_BUDGET of one group of pictures distributes according to the bit distribution information of picture.In the bit distribution information of group, PIC_ETC is relevant with quantification.Therefore, for the aim parameter of the coded-bit of organizing relevant with quantification, ACDC_TARGET_BUDGET is determined by following formula:

ACDC_TARGET_BUDGET＝TARGET_BUDGET_PETC(i)(1)

The PIC_ETC value of i picture in PETC (i) the representative group wherein, i equals 1 to I, and I is the number of picture in the group.

Distribute the ACDC_TARGET_BUDGET of each picture then according to the ratio of the value of each picture PIC_ACDC and class mean in the group, be defined as:

PIC_ACDC_TARGET_BUDGET(i)(j)＝

ACDC_TARGET_BUDGET(i)×PACDC(j)/∑PACDC(i)(2)

Wherein PIC_ACDC_TARGET (i) represent i picture with the aim parameter that quantizes the correlative coding bit, PACDC (i) represents the value of the PIC_ACDC of i picture.

At last, the coded-bit aim parameter of a picture can by will with quantize the extraneous data amount and with quantize the addition of correlative coding bit aim parameter and obtain, be defined as:

PIC_TARGET_BUDGET(i)

＝PIC_ACDC_TARGET_BUDGET(i)+PETC(i) (3)

Wherein PIC_TARGET_BUDGET (i) represents the aim parameter of i picture coding bit.

Similarly, i picture and the aim parameter PIC_ACDC_TARGET_BUDGET (i) that quantizes relevant coded-bit are that each macroblock allocation in the picture is as follows according to the ratio of the value of the BLOCK_ACDC of each piece:

BLOCK_ACDC_TARGET_BUDGET(i)(j)

＝PIC_ACDC_TARGET_BUDGET(i)×BACDC(j)/2048 (4)

Wherein BLOCK_ACDC_TARGET_BUDGET (i) j macro block (j) representing i picture with quantize correlative coding bit aim parameter, and BACDC (j) is the value of BLOCK_ACDC of j piece of i picture, j is 1 to J, and J is the piece number of i picture.

J piece of i picture (j) can followingly calculate with the irrelevant coded-bit aim parameter BLOCK_ETC_TARGET_BUDGET (i) of quantification:

BLOCK_ETC_TARGET_BUDGET(i)(j)

＝PETC(i)×BETC(j)/2048 (5)

Wherein BETC (j) is the value of BLOCK_ETC of j piece of i picture.

At last, the coded-bit aim parameter of j piece of i picture can followingly calculate:

BLOCK_TARGET_BUDGET(i)(j)

＝BLOCK_ACDC_TARGET_BUDGET(i)(j)+

BLOCK_ETC_TARGET_BUDGET(i)(j) (6)

The bit control algolithm of carrying out in the bit control circuit 42 shown in Figure 5 is as follows: for the actual generation amount of coded data according to i picture of coded-bit aim parameter (being PIC_TARGET_BUDGET (i)) control, the Qp of the current macro of picture is by determining the amount of coded data sum of the actual generation of previous coding macro block and the deviation or the difference feedback of its coded-bit aim parameter sum.The poor d (j) of j macro block of i picture is defined as:

d(j)＝d(0)+Bits(j-1)-TGT(j-1) (7)

Wherein d (0) is the initial occupancy of virtual bumper; Bits (j-1) is the summation by the amount of coded data of first to j-1 actual generation of macro block of picture; And TGT (j-1) is (j) sum of coded-bit aim parameter BLOCK_TARGET_BUDGET (i) of j-1 piece of i picture.

For first macro block in the picture, Bits (j-1) and TGT (j-1) are made as null value respectively, therefore, and d (1)=d (0).

The Qp of j macro block is that Q (j) is defined as:

Q(j)＝d(j)×31/r (8)

Wherein r is a parameter of determining the response speed of feedback.

Parameter r and bit rate/sampling rate are proportional; And be defined as:

r＝2×Bit_rate/picture_rate (9)

Wherein Bit_rate is the target numbers of coded data each second, i.e. target bit rate, and picture_rate is the number of per second input picture.

The initial Q p that determines the initial occupancy d (0) of virtual bumper is that Q (0) calculates according to the inverse relation between the data volume of Qp and generation.Q (0) may be defined as:

Q(0)＝Qfix×PACDC/PIC_ACDC_TARGET_BUDGET(i)(10)

And d (0) may be defined as:

d(0)＝Q(0)×r/31 (11)

Wherein the meaning of Qfix, PACDC, PIC_ACDC_TARGET_BUDGET (i) is with top identical.

The Qp that each macro block is determined is sent to quantizer 33 in a manner described, and it is according to the conversion coefficient of the quantization step quantification of being determined by Qp from each macro block of converter 32.Coefficient after this quantification is delivered to IQ 34 and statistics device 38.At statistics device 38, from ME ﹠amp; The output of MC circuit 31 is for example adopted, and the VLC technology is encoded.Coded data from statistics device 38 offers bit counter 41 and transmits buffer 19, wherein the coded data of irregular bit rate is temporarily preserved, and sends to the conveyer (not shown) that is used to send as encoded video signal with a constant bit rate.

With reference to figure 6, it is the block diagram of explanation according to the video signal encoding device 6 of second embodiment of the invention.The equipment 6 of second embodiment is realized by the common unit of first code device 1 shown in Figure 2 and second code device 4 shown in Figure 5 among shared first embodiment.The label 50 to 58 and 62 elements that identified of common component and equipment shown in Figure 66 are corresponding, and therefore, its detailed content no longer repeats for succinct consideration.

Also have, the transmission buffer 59 of equipment 6, target bits are determined the same in circuit 67 and bit control circuit 68 and second code device 4 shown in Figure 5; And memory 69 is the same with the function of memory 2 among first embodiment.Equipment 6 also comprises a path control circuit 60, switch 63-65, and a Qfix produces a circuit 61 and a data storage 66.Path control circuit 60 sends a switch controlling signal to switch 63-65.By means of switch controlling signal among the switch 63-65 each is switched to a end or b end, like this type of action of equipment 6 with regard to respectively with the first embodiment equipment in first code device 1 identical with second code device 4.

Qfix produces circuit provides a fixedly Qp who is scheduled to and Qfix to quantizer 53 by switch 65.Data storage 66 has a capacity bigger than memory 69, and it recovers the stored data of memory 69 and restore data is preserved.

In order according to second embodiment of the invention input digital video signal to be encoded, at first, path control circuit 60 sends one first switch controlling signal and gives each switch 63-65.To this response, switch 63-65 switches to a end, the a end of switch 64 is zero or opens end, the output of bit counter 62 is coupled by switch 63 and memory, and conversion coefficient quantizes according to the determined Qs of Qfix that sends here according to Qfix generation circuit 61 at quantizer 53 by switch 65.

The vision signal of input is to encode with reference to figure 2 illustrated first code device, 1 the same modes together; With quantize the relevant transform coefficient data and offer bit counter 62 with macro block-macro block mode from statistics device 58 of being encoded with quantification related data such as motion vector and predictive mode signal.

In bit counter 62, and quantize relevant and be counted and offer memory 69 with each data volume that quantizes to have nothing to do, be kept at wherein as bit distribution information.The bit distribution information of being preserved at memory 69 is kept at data storage 66 then.In the bit distribution information access process, the data that are encoded that obtained by statistics device 58 do not offer transmission buffer 59.

One obtains the bit distribution information of predetermined picture group, just begins bit allocation procedures by producing a second switch signal from path control circuit 60, and for responding this second switch signal, each all becomes b with its link by a among the switch 63-65.According to this switching manipulation, bit counter 62, bit control circuit 68 and memory 69 determine that with bit control circuit 68, quantizer 53 and target bits circuit 67 links to each other respectively, thereby provide equipment 6, it has identical structure with second code device 4 shown in Figure 5, and quantizer 53 is handled according to carried out adaptive quantizing by bit control circuit 68 determined Qp.

In adaptive quantizing is handled process to video signal coding, to the input of equipment 6 be with the bit distribution information access process in used identical input digital video signal.This input digital video signal can be by providing in storage medium (not shown) such as input buffer or the last tape of installing of VTR (video recorder).For example, if tape is that in case the bit distribution information access process is finished, tape offers equipment 6 by playing tape once more with identical input digital video signal promptly by rewinding as storage medium.

After this, cataloged procedure and second code device 4 shown in Figure 5 is the same.

Below the third embodiment of the present invention is described.

Fig. 9 is the block diagram of the third embodiment of the present invention.When reference Fig. 9, picture intelligence offers quadrature transformer 102 from input terminal 101.The pictorial data of each continuous block is carried out Hadamard transform or the such orthogonal transform of discrete cosine transform by quadrature transformer 102.102 pairs of each continuous blocks of quadrature transformer, output comprises a plurality of orthogonal transform coefficient of flip-flop (DC coefficient) and alternating component (AC coefficient).

The orthogonal transform coefficient of being exported from quadrature transformer 102 is provided for normalizer 103, carries out standardization by normalizer 103.

Normalizer 103 is carried out standardization to input DC coefficients and is become the DC coefficient of standard.Normalizer 103 is carried out standardization to input AC coefficients and is become the AC coefficient of standard.The standardization of the standardization of DC coefficient and AC coefficient separates to be carried out.Standard DC coefficient and standard A C coefficient are exported from lead-out terminal respectively.Normalizer 103 generates the standardized information (data) of representing level value and change value, and output.

The standardized method of input DC coefficient is different mutually with the standardized method of input AC coefficient.Because the DC coefficient is corresponding to the mean value in the piece, thereby, with the similar method of standardization of general sampling value, use maximum and minimum to come the DC coefficient is carried out standardization.Because the mean value of AC coefficient approaches zero, then standardization is carried out according to the absolute value of AC coefficient.

The minimum of the DC coefficient of each piece of input and maximum are detected respectively continuously.

Pass through the minimum and the maximum that are detected, deduct minimum, obtain maximum and minimizing difference value d1 from maximum.

Calculate the absolute value of input AC coefficient, calculated absolute on duty 2.Is absolute on duty 2 reason to consider that peak value is approximately 1/2nd such facts and carries out when absolute value is calculated.

Detect the maximum d2 that the AC coefficient is taken advantage of 2 absolute value.

Difference value d1 that is associated with the DC coefficient and the maximum d2 that is associated with the AC coefficient equally represent departure (deviate), and therefore, value d1 and d2 can handle in the back equally.The bigger value of selective value d1 and d2, then making selected value is deviate d3.

Deviate d3 and minimum (representative level value) are provided for not shown vector quantizer, quantize to reduce amount of information.

Never representative level value that is quantized that illustrated vector quantizer is exported and the deviate that is quantized are represented standardized information D.The representative level value that is quantized is transformed to the minimum relevant with the decrease of information (representative level value) m and deviate d with the deviate that is quantized.

Deduct minimum and generate new DC coefficient a from the DC coefficient.Its result, the minimum value of new DC coefficient a equals zero, and the scope of DC coefficient moves, and thus, has reduced amount of information.

Divided by new DC coefficient, thus, the DC coefficient is carried out standardization with deviate d, and is created on the standardized DC coefficient that exists between 0 and 1, and exports to regular length memory 104.

The AC coefficient is carried out delay, and becomes the AC coefficient b that is delayed, and divided by AC coefficient b, thus, the AC coefficient is carried out standardization with deviate d, and is created on the standardized AC coefficient that exists between 0 and 1, and exports to regular length memory 104.

Output signal from normalizer 103 is delayed preset time by the regular length memory, and this scheduled time is set to compensation by feedfoward control caused time lag.

The signal that is delayed of AC and DC coefficient offers adaptive quantizer 105 from regular length memory 104.Adaptive quantizer 105 comes AC and DC coefficient are quantized with the variable quantization stride Sq that each block is determined.Dateout from adaptive quantizer 105 is encoded as the such variable length code of Huffman code by entropy coder 106.

When bigger quantization step Sq was used by adaptive quantizer 105, expression was less from the bit number of the coded data of entropy coder 106.Otherwise when less quantization step Sq was used by adaptive quantizer 105, expression was bigger from the bit number of the coded data of entropy coder 106.In the method, the outputting encoded data amount from entropy coder 106 depends on by adaptive quantizer 105 employed quantization step Sq.Outputting encoded data from entropy coder 106 temporarily is stored in the variable-length buffer storage 107.Coded data is read from variable-length buffer storage 107, and transmits with predetermined constant transmissions rate through lead-out terminal 108.

Under the situation of using the Huffman code, the data of level " 0 " are corresponding to the shortest code word, and the data of bigger absolute level are corresponding to longer code word.Because the major part of AC orthogonal transform coefficient is present near 0, then the major part of data is encoded by entropy coder 106 and is become short code word.Like this, can realize the abundant compression of data by the coding that entropy coder carried out.

The standardized information signal D that is exported from normalizer 103 postpones by regular length memory 104.The standardized information signal is read by regular length memory 104, transmits through lead-out terminal 109.Because the standardized information signal data has fixed word length, prepares buffer storage with regard to not needing for the standardized information signal.

As described above, the outputting encoded data amount from entropy coder 106 depends on by adaptive quantizer 105 employed quantization step Sq.Control part A controls outputting encoded data amount from entropy coder 106 by adjusting by adaptive quantizer 105 employed quantization step Sq.This control is designed to like this: in each resembles such scheduled period image duration, the outputting encoded data amount is remained on the predetermined fixed value approx.The delay standardized information that control part A is provided according to the non-delay standardized information that is provided by normalizer 103, by regular length the memory 104 and represent signal that is provided by entropy coder 106 of the outputting encoded data amount Db of reality to decide quantization step Sq.

Control part A has data volume fallout predictor 110 and quantizes step setting apparatus 112.Non-delay standardized information is provided for data volume fallout predictor 110 by normalizer 103.The delay standardized information is provided for quantification step setting apparatus 112 by regular length memory 104.Control part A carries out feedfoward control according to non-delay standardized information and delay standardized information.

Data volume fallout predictor 110 has ROM, this ROM keep the predetermined piece corresponding to standardized information the representative coded data and in the relative predicted value of the bit quantity of necessity.As described above, data volume fallout predictor 110 is subjected to the signal D of replacement table standardized information from normalizer 103.Data volume fallout predictor 110 uses the standardized information signal as the address signal relative with ROM, and from reading predicted value by the field of the represented ROM of address signal, thus, decides the predicted value relative with bit quantity according to the standardized information of being got.The predicted value relative with data volume changes with the rugosity that quantizes or the degree of precision, and therefore, these predicted values are selected so that corresponding to by adaptive quantizer 105 employed quantization step.The predicted value that is determined is exported to threshold setting device 111 from data volume fallout predictor 110.Threshold setting device 111 comes setting threshold according to the predicted value that is provided by data volume fallout predictor 110.Threshold value is provided for quantification step setting apparatus 112 by threshold setting device 111.

Quantize step setting apparatus 112 and got the delay standardized information from regular length memory 104, delay standardized information and each piece right threshold value compare, thus, decide the grade of piece.

Generate initial quantization step Sq according to grade.Have not shown ROM, be assigned to the predetermined value of the quantization step Psqs of each address with storage.

Generate predicted data amount PDb according to grade.The predicted data amount PDb that is generated is consistent with the target data amount.Have not shown ROM, be assigned to the predetermined value of the predicted data amount PDb of each address with storage.

Owing to decide the address according to the piece grade, then select the predetermined value of predicted data amount PDb according to the piece grade.Be predetermined predicted data amount PDb value process be predetermined by the data volume fallout predictor 110 employed every processes of predicted value of bit quantity identical.The mean value that calculates is used as the value of predicted data amount PDb with adopting statistics.

Entropy coder 106 generates the signal of the actual amount Db of the coded data that representative exported from entropy coder 106 during corresponding to a piece.The actual amount of data signal is exported to subtracter 113 from entropy coder 106.The data volume difference (error) of subtracter 113 generation forecast data volume PDb and actual amount of data Db.The data volume error offers the not shown adder of accumulating the adder 114 from subtracter 113.

Accumulation adder 114 has the not shown piece holding circuit that keeps accumulation additional calculation value K.Adder is carried out add operation to the data volume error with accumulation additional calculation value K, and generates new accumulation additional calculation value, and new accumulation additional calculation value is exported to not shown piece holding circuit.In the piece holding circuit, accumulation additional calculation value K is replaced by new accumulation additional calculation value, upgrades accumulated value like this.Accumulation additional calculation value K is offered by not shown piece holding circuit and quantizes the step setting apparatus.

Wherein, accumulation additional calculation value K is transformed to shift signal DSq.Initial quantization step PSq is carried out add operation, generate the final quantization step Sq that offers adaptive quantizer 105.

As described above, in the data volume fallout predictor of controller A, be predetermined the function that the predicted value relative with the data volume of piece is used as standardized information.Predicted value relative with the data volume of piece and the relation between the standardized information are selected as: the outputting encoded data amount from entropy coder 106 that guarantees the scheduled period is remained predetermined constant basis approx.The mean value that calculates is used as the predicted value relative with the data volume of piece with adopting statistics.

Threshold setting device 111 is predicted the data volume of exporting in the scheduled period according to the predicted data amount of each piece that is provided continuously by data volume fallout predictor 110.In threshold setting device 111, select or determine to be the employed threshold value of classification of each piece, so that make the premeasuring of the data of exporting in such scheduled period of the image duration equal predefined aim parameter.

Because each grade corresponding to piece decides quantization step, then the deviation in the threshold value has been brought the variation in the quantization step.Like this, the actual amount of the data of exporting in such scheduled period of the image duration just can equal predefined aim parameter.

Particularly, threshold setting device 111 in such scheduled period image duration, the predicted data amount of each piece that is provided continuously by data volume fallout predictor 110 is provided, thus, and the predicted data amount of being exported in calculating during predetermined (a for example frame).In data volume fallout predictor 110, prepare many group predetermined threshold value.Estimate the dateout actual amount by the premeasuring and respectively the organizing of predetermined threshold value of using the data of in the scheduled period, being exported.

In the method, can be consistent for the evaluation of estimate of the actual amount of dateout with respectively organizing of predetermined threshold value.Evaluation of estimate is by being compared with predetermined target value, and selects to equal desired value or less than desired value and nearest evaluation of estimate one.From the group of threshold setting device 111 outputs with the corresponding to threshold value of selected evaluation of estimate.

The function of threshold setting device 111 is consistent with following processing.For its total that obtains in such scheduled period image duration, being generated, come from data volume fallout predictor 110 continuously the premeasuring of the output encoders of input carry out add operation.The total of premeasuring is compared by the aim parameter with the output encoder that is generated in such scheduled period image duration.According to the total of premeasuring and the comparative result of aim parameter decide with the scheduled period in the relative threshold value of continuous piece.The decision of threshold value is configured to the actual amount of output encoder is maintained on the aim parameter of output encoder.The threshold value that is determined is exported to continuously quantizes step setting apparatus 112.

Under the situation of general picture intelligence, the accumulated value of predicted data amount PDb slightly equals the accumulated value of actual amount of data Db, and accumulation additional calculation value K approaches zero.Under the situation that the actual amount Db of reality is biased, accumulation additional calculation value K obviously is a negative or positive, quantizes the shift signal DSq that step shift unit 119 generates the biasing that can compensate actual amount of data Db.

As described above, quantize step setting apparatus 112 accumulation additional calculation value K is transformed to shift signal DSq.Basically decide mutually conversion characteristics independently in the characteristic of coding, and accumulation additional calculation value K is set to and is no more than preset upper limit Kmax to the accumulation additional calculation value of shift signal DSq.

As from seeing above-mentioned, carry out feedfoward control by using predicted data amount.Obtain predicated error and as the result of feedfoward control.Feedfoward control is carried out according to predicated error.During during corresponding, lacking like this, carry out feedfoward control and FEEDBACK CONTROL repeatedly with several.Go up the data volume error is maintained below the upper limit Kmax finish time that is controlled at each image duration of above-mentioned digital coding amount.The threshold value that is determined in feedfoward control portion is relevant with the target data amount and need have the allowance of higher limit Kmax.

Although the present invention with preferred embodiment as an illustration, under the prerequisite that does not deviate from the spirit and scope of the invention that is indicated in the appended claims, various modifications and variations can be arranged.

Claims

One kind to contain with quantize coherent signal and with the method for the encoding video signal that quantizes irrelevant signal, the steps include:

(a) quantize and the quantification coherent signal with a fixed quantisation parameter, to produce first quantized data;

(b) carry out the statistics coding to first quantized data with the signal that quantizes to have nothing to do respectively, so that first and second data to be provided respectively;

(c) summation of picture integral body is represented with 2 Nth power,, and the value after the standardization is stored so that carry out standardization;

(d) determine the aim parameter of the coded-bit of each video signal unit according to predetermined target bitrate and first, second data volume;

(e) determine one or more quantization parameters according to the coded-bit aim parameter;

(f) quantize and quantize related data according to the quantization parameter of determining to produce second quantized data; With

(g) carry out the statistics coding to second quantized data with the quantification irrelevant signal, to produce the 3rd data and the 4th data coded data thus respectively as vision signal, wherein, according to the difference between coded-bit aim parameter and the amount of coded data, quantization parameter is determined adaptively.
2. according to the process of claim 1 wherein that each unit is corresponding with a picture of vision signal.
3. according to the process of claim 1 wherein that each unit is corresponding with a macro block of vision signal.
4. according to the process of claim 1 wherein that each picture and macro block to vision signal determine the coded-bit aim parameter.
5. according to the process of claim 1 wherein that described determining step (d) comprises the steps:

(d1) calculate one according to the sampling rate of frame numbers in target bit rate, the picture group and vision signal and be scheduled to the coded-bit aim parameter of picture group;

(d2) to this picture group, detect this group by second data that from the coded-bit aim parameter of this group, deduct this group with the aim parameter that quantizes the correlative coding bit;

(d3) find with the picture group in corresponding first data volume of each picture with respect to the ratio of corresponding first data volume of this picture group;

(d4) with this picture group multiply by the resulting ratio of step (d3) with quantizing correlative coding bit aim parameter, obtain each picture thus with quantize correlative coding bit aim parameter;

(d5) with obtain in the step (d4) with quantize of the second data volume addition of correlative coding bit aim parameter with each picture, with the coded-bit aim parameter of definite each picture.
6. according to the method for claim 5, wherein said determining step (d) also comprises the steps:

(d6) calculate corresponding to first data volume of each macro block in the picture ratio with respect to first data volume of this picture;

(d7) this picture that in step (d4), obtains multiply by the ratio that obtains in (d6) with quantizing correlative coding bit aim parameter, with each macro block is provided thus with quantize correlative coding bit aim parameter; With

(d8) (d7) provided with quantize of the second data volume addition of correlative coding bit aim parameter with each macro block, thereby definite each macroblock encoding bit aim parameter.
7. according to the method for claim 5, wherein the amount of each picture second data adds by each amount of second data with a plurality of macro blocks in each picture and obtains, and first data volume of each picture is by adding a plurality of macro blocks contained in each picture first data volume separately and obtaining.
One kind to contain with quantize coherent signal and with the equipment of the encoding video signal that quantizes irrelevant signal, it is characterized in that comprising:

First code device, it has:

First quantifying unit is used for quantizing and quantizing coherent signal according to a fixed quantisation parameter, producing first quantized data,

The first statistics coding unit, be used for to first quantized data and with quantize the irrelevant signal coding, so that first data and second data to be provided respectively,

First counting unit, be used for to first and second data count respectively and

Subtracter, and discrete cosine transformer;

Storage device is represented the summation of picture integral body with 2 Nth power, so that carry out standardization, and the value after the standardization is stored in the described storage device;

With

Second code device, it has:

Be used for determining the parts of each predetermined video signal cell encoding bit aim parameter,

Produce the parts of one or more quantization parameters according to the coded-bit aim parameter for each unit,

Second quantifying unit is used for quantizing and quantizing coherent signal according to determined quantization parameter, producing second quantized data,

The second statistics coding unit is used for encoding to second quantized data with quantizing irrelevant signal, so that the 3rd data and the 4th data to be provided respectively, and as the coded data of vision signal,

Second counting unit is used for third and fourth data are counted respectively.
9. equipment according to Claim 8 is characterized in that the parts of described generation quantization parameter comprise:

Detect the parts of difference between coded-bit aim parameter and the amount of coded data; With

Regulate the parts of quantization parameter adaptively according to described difference.
10. equipment according to Claim 8 is characterized in that each predetermined video signal unit is corresponding with a picture of vision signal.
11. equipment according to Claim 8 is characterized in that each predetermined video signal unit is corresponding with the macro block of vision signal.
12. equipment according to Claim 8 is characterized in that the aim parameter that each picture and the macro block of vision signal are determined coded-bit.
13. equipment according to Claim 8 is characterized in that each picture and the macro block of vision signal are stored its first and second data volume.
14. equipment according to Claim 8 is characterized in that also comprising a deferred mount, is used for depositing video signal delay in storage device until first and second data volumes, it gives second code device with vision signal then.
15. one be used for to contain with quantize coherent signal and with the device of the video signal coding that quantizes irrelevant signal, it is characterized in that comprising:

To quantizing so that the parts of first quantized data to be provided with the quantification coherent signal;

Carry out the statistics coding to produce the parts of first data and second data respectively to first quantized data with the quantification irrelevant signal;

Counter block is used to obtain first data volume and second data volume;

Memory unit is represented the summation of picture integral body with 2 Nth power, so that carry out standardization, and the value after the standardization is stored in the described memory unit;

Target bits is determined parts, is used for determining according to target bit rate and the first and second stored data volumes coded-bit aim parameter of each scheduled unit of vision signal;

The bit control assembly is used for determining the initial quantization parameter according to the coded-bit aim parameter, and the difference of the amount of coded data that produces according to coded-bit aim parameter and statistics addressable part further determines quantization parameter, produces determined quantization parameter with this;

Be used to be provided with the parts of fixed quantisation parameter; With

Path control member, be used for fixed quantisation parameter is coupled to the quantification parts, and import first and second data volumes to memory unit, after this, the output of the first and second stored data volumes, counter block and definite quantization parameter are offered target bits respectively determine parts, bit control assembly and quantize parts, in quantizing parts, control quantization step adaptively thus.
16., it is characterized in that each predetermined video signal unit is a picture of vision signal according to the device of claim 15.
17., it is characterized in that each predetermined video signal unit is a macro block of vision signal according to the device of claim 15.
18., it is characterized in that each picture of vision signal and macro block are determined the aim parameter of coded-bit according to the device of claim 15.