CN100342728C

CN100342728C - Method for controlling quantization degree of video signal coding bit-stream and related device

Info

Publication number: CN100342728C
Application number: CNB2003101188537A
Authority: CN
Inventors: 郭志辉
Original assignee: MediaTek Inc
Current assignee: MediaTek Inc
Priority date: 2003-11-28
Filing date: 2003-11-28
Publication date: 2007-10-10
Anticipated expiration: 2023-11-28
Also published as: CN1622611A

Abstract

The present invention relates to a method for controlling the quantization scale of an image signal in a video signal encoding device, which comprises: the changing frequency of the quantization scale in an encoding process is calculated by the video signal encoding device; a new quantization scale is generated by the video signal encoding device; the changing frequency and a default value are compared to determine whether to replace the original quantization scale with the new quantization scale, wherein when the changing frequency exceeds the default value, the original quantization scale is not replaced by the new quantization scale so as to keep the changing frequency of the quantization scale in the default value in the encoding process.

Description

The method and the relevant apparatus of the quantization degree of control of video signal encoding bit stream

Technical field

The present invention relates to a kind of method and relevant apparatus of quantization scale (Quantization Scale) of control one picture signal, particularly a kind of method and relevant apparatus that is used for the quantization scale of video signal coding apparatus control one picture signal.

Background technology

At moving picture coding (the Moving Picture Coding Experts Group 2 of expert group 2, hereinafter to be referred as MPEG 2 specifications) in, utilize the color sampling, discrete cosine transform (Discrete CosineTransform, DCT) and quantize modes such as (Quantization), remove redundancy (Spatial Redundancies) information on the vision signal image spacing, and by bit shift compensation (MotionCompensation, MC) mode, elimination is because of the temporal redundancy that similitude caused (Temporal Redundancies) information between picture and picture, to reach the effect of compressed video signal.

Generally speaking, having some common characteristics on same vision signal picture perhaps, is on the color, perhaps be on how much, or the further feature value obtains.Remove the mode of the redundant information on these spaces, will identify element important in the picture exactly, and remove repetition and do not have the element that influences.According to experiment, human eye changes responsive for brightness and finds for relative being difficult for of the variation of colourity.Therefore, MPEG 2 specifications adopt the color representation form of brightness and colourity, represent brightness value (Luminance with Y, or be called Luma), represent chromatic value (Chrominance, or be called Chroma) with Cr and Cb, and reduce semaphore in the mode that reduces chroma sampling.Define three kinds of sampling modes of 4:2:0,4:2:2 and 4:4:4 in MPEG 2 specifications, represented three kinds of different chroma sampling frequencies respectively.For example, the 4:2:0 sampling mode is to be in the macro zone block (Macro Block) of 16 * 16 pixels in a size, 4 the Y blocks (each block size is 8 * 8 pixels) of taking a sample, 1 Cr block (size is 8 * 8 pixels) and 1 Cb block (size is 8 * 8 pixels).By the frequency that reduces chroma sampling, can be issued to preferable data reduction effect influencing the less situation of vision.

In addition, the general vision signal of playing is a succession of continuous images sequence in fact, yet because have a kind of phenomenon of the persistence of vision on the human vision, so can produce the illusion of coherent image.And in this kind vision signal, because the time interval between picture is very little, so several indifferences between the picture that faces mutually are the change in location of picture material mostly.Therefore, in moving picture coding expert group 2 (MPEG 2), (Motion Compensation, mode MC) are eliminated the redundant information that similitude caused between picture and picture on the time shaft, reach the purpose of compressed video signal by bit shift compensation.

Above-mentioned video signal compression Methods for Coding is that known technology is known, and below only simply describes its function mode.Figure 1 shows that the schematic diagram of a known video signal encoder 10 (Video Encoder).Video coder 10 includes a discrete cosine transform (Discrete Cosine Transform, DCT) device 12, one dynamic estimation and compensator (Motion Estimator and Compensator) 14, one quantizer (Quantizer) 16, a variable length coder (Variable Length Encoder, VLE) 18 and one bit rate controller (Rate Controller) 20.Video coder 10 is redundant informations of utilizing discrete cosine conversion device 12 and quantizer 16 to eliminate on the vision signal image spacing, and eliminates temporal redundant information with dynamic estimation and compensator 14, with the size of compressed video signal data.At last, data are delivered to a system multi-channel multiplexer (not shown) after encoding through variable length coder 18, with defined transmitting data stream of MPEG 2 specifications (Transport Stream) or program data stream (Program Stream) data are sent.

Discrete cosine transform (DCT) device 12 is for each block (size is 8 * 8 pixels) behind macro zone block process brightness and the chroma sampling, carry out discrete cosine transform respectively, so that video signal data is converted into frequency domain (Frequency Domain) from spatial domain (Spatial Domain).Discrete cosine transform (DCT) is a kind of mathematical computations of completely reversibility.The color-values of block is resulting discrete cosine transform coefficient (DCT Coefficient) after discrete cosine transform, still is one 8 * 8 two-dimensional matrix.Generally speaking, the color in picture does not have very intensive and big fluctuation usually, therefore, represents the value of discrete cosine transform coefficient of higher-spatial frequencies very little usually in the matrix, even is zero.Basically, discrete cosine transform can not reduce data volume, more easily looks for redundant expression pattern but but data can be changed into.

Then, quantizer 16 can quantize discrete cosine transform coefficient, with the size of further compressed video signal data.Say that simply quantizing (Quantization) is exactly the process that reduces the figure place of describing each coefficient, also is about to each coefficient and describes it with more coarse linear module.The action that quantizes mainly contains two functions: the one, and allow former very approaching zero value become zero as far as possible, then be to make the distribution of coefficient of original non-zero diminish in addition, handle to help compression.Quantification is a kind of destructive compression skill, when the data after the quantification restore and original data can be not identical.Therefore, the degree of distortion after the video signal compression mainly is to depend on choosing of quantization scale (Quantization Scale).

The function of the bit rate controller 20 of video coder 10 promptly is a default output bit rate scope during according to video coder 10 design, the quantization scale when adjusting quantizer 16 and carrying out quantification treatment.Bit rate controller 20 is to be unit with the macro zone block when adjusting quantization scale.Anticipate promptly, each block behind macro zone block process brightness and the chroma sampling all has an identical quantization scale.

Finish the quantification treatment of a block when quantizer 16 after, video coder 10 can utilize specific algorithm, one discrete cosine transform coefficient that quantizes in the two-dimensional matrix of back is concatenated into the ordered series of numbers of one dimension, and makes in this one dimension ordered series of numbers that the length of null value is for maximum, to improve compression efficiency continuously.Then, the ordered series of numbers data that variable length coder 18 can these one dimensions of compression to export the data flow (BitStream) after the compression, are called coded bit stream.

Quantizer 16 employed quantization scale are more little, and then the video signal compression ratio is more little, quality video signal is high more, but the bit rate of the coded bit stream of variable length coder 18 outputs also can be high more.Otherwise quantization scale is big more, and then video signal compression is bigger than more, quality video signal is low more, but the bit rate of the coded bit stream of variable length coder 18 outputs also can be low more.In the middle of this coded bit stream, it is the quantization scale that is used for expressing central this block of cataloged procedure that part position (about 7-8 position) is arranged, the usefulness during for this coded bit stream decoding reduction.

According in the above-mentioned running explanation for known video signal encoder 10 as can be known, when the output bit rate of video coder 10 need maintain in the scope of presetting, if it is too frequent that bit rate controller 20 changes the frequency of quantization scale, then in the coded bit stream of video coder 10 coding back output, the figure place that is used for noting down quantization scale will be very considerable.For example, with the NTSC specification, per second will be play 30 frames (Frame), and each frame has 1350 macro zone blocks.In the case, if the output bit rate of video coder 10 is 2Mbps, then average each macro zone block only can use 49.3 positions approximately after overcompression.If bit rate controller 20 changes the quantization scale of each macro zone block, then exporting has 14% (7/49.3=0.14) to consume on these quantization scale of record approximately in the bit rate.

Hence one can see that, in known video signal encoder 10, if it is high more that bit rate controller 20 changes the frequency of quantization scale, then have many more bit spaces in the coded bit stream of last output and be used to note down these quantization scale, the feasible recorded video signal section data that really can be used for is restricted, and then reduces the quality of video signal coding.

Summary of the invention

Therefore, when one of purpose of the present invention is to be provided at encoded video signal control quantization scale method and relevant apparatus, by the restriction quantization scale change frequency to address the above problem.

The invention provides a kind of method of controlling quantization scale (Quantization Scale) when coding image signal, this method includes (a) and obtains the change frequency of this quantization scale in a period; (b) produce a new quantization scale; And (c) relatively this change frequency and a preset value, whether a decision replaces an original quantization scale with this new quantization scale.Wherein, when this change frequency surpasses this preset value, then keep this original quantization step.

Advantage of the present invention is to limit the change frequency of the quantization scale in the middle of the cataloged procedure, and making has in the middle of the coded bit stream more multidigit to use for video data, and then the quality of lifting video coding.

The accompanying drawing summary

Fig. 1 is the schematic diagram of known video coder.

Fig. 2 is the schematic diagram of video coder of the present invention.

Fig. 3 controls the flow chart of method of the quantization scale of a picture signal for the present invention.

The reference numeral explanation

10,100 video coders

12 discrete cosine conversion devices

14 dynamic estimation and compensator

16 quantizers

18 variable length coders

20,120 bit rate controllers

Embodiment

Please refer to Fig. 2.Fig. 2 quantizes the video coder 100 of rank, position change frequency for may command of the present invention.Video coder 100 is very similar with video coder 10, and therefore identical assembly is to be numbered with identical number in two video coders.Redundancy (SpatialRedundancies) information that video coder 100 utilizes discrete cosine conversion device 12 and quantizer 16 to eliminate on the vision signal image spacing equally, and eliminate temporal redundancy (TemporalRedundancies) information with dynamic estimation and compensator 14, with the size of compressed video signal data.At last, data are through variable length coder 18 coding back outputs a becoming coded bit stream.

As previously mentioned, quantizer 16 is that the quantization scale that determined according to bit rate controller 120 is to quantize a block.Yet in the present invention, bit rate controller 120 can be according to the change frequency of the quantization scale in the coding course, and whether decision will replace an original quantization scale with a new quantization scale.Next with the detailed operation situation of flowchart text bit rate controller 120 of the present invention.

Please refer to Fig. 3.Fig. 3 is flow process Figure 200 of the quantization scale of video coder 100 controls one picture signal of the present invention, and it includes following steps:

Step 202: beginning.

Step 204: required parameter when setting 120 runnings of bit rate controller, for example, initial value, the higher limit TH of quantization scale change frequency and critical value QTH of quantization scale amplitude of variation of the change frequency Freq_Q of the P of chronomere of calculating quantization scale change frequency, quantization scale or the like.

Step 206: 120 generations one of bit rate controller correspond to the new quantization scale Q of next macro zone block _New

Step 208: bit rate controller 120 is this new quantization scale Q relatively _NewWhether equal an original quantization scale Q _OldIf,, then carry out step 216; If not, then carry out step 210.

Step 210: this new quantization scale Q relatively _NewWith this original quantization scale Q _OldBetween amplitude of variation, and check whether this amplitude of variation equals or exceeds this critical value QTH, if then carry out step 214; If not, then carry out step 212.

Step 212: relatively whether the change frequency Freq_Q of quantization scale equals or exceeds this higher limit TH, if then carry out step 216; If not, then carry out step 214.

Step 214: bit rate controller 120 will this new quantization scale Q _NewBe sent to quantizer 16, to replace this original quantization scale Q _Old, the foundation that next quantizes as quantizer 16.

Step 216: bit rate controller 120 can be with this new quantization scale Q _NewReplace this original quantization scale Q in the quantizer 16 _OldSo next quantizer 16 continues with this original quantization scale Q _OldAs the foundation that quantizes.

Step 218: bit rate controller 120 upgrades the value of quantization scale change frequency Freq_Q.

Step 220: judge whether to also have the macro zone block that needs processing, if then carry out step 206; If not, then carry out step 222.

Step 222: finish.

During the function mode of following description of flow diagram 200, be that hypothesis quantizer 16 carries out quantification treatment over against a macro zone block L at present, and the quantization scale that quantizes this macro zone block L time institute foundation is Q _Old

In a preferred embodiment of the present invention, video coder 100 is understood parameter required when step 204 is set 120 runnings of bit rate controller.For example, the time basis P that calculates the quantization scale change frequency is set at 20 macro zone blocks, it is unit that expression is finished the time that quantizes 20 macro zone blocks with quantizer 16, calculate during this section in the change number of times of quantization scale; The higher limit TH of quantization scale change frequency is set at 10, and expression quantizer 16 was finished within the time that quantizes 20 macro zone blocks, was limited to 10 times on the change number of times of quantization scale; The initial value of the change frequency Freq_Q of quantization scale is set at 0, and expression bit rate controller 120 is since the 0 change number of times that calculates quantization scale; With the critical value setting of quantization scale amplitude of variation is QTH.

In step 206, for being next to after the macro zone block L, be about to be undertaken next macro zone block M of quantification treatment by quantizer 16, bit rate controller 120 can produce corresponding one new quantization scale Q _New

Then, bit rate controller 120 can be checked the quantization scale Q that this is new in step 208 _NewWith this original quantization scale Q _OldWhether identical, if both value is identical, then carry out step 216, bit rate controller 120 can inform that quantizer 16 keeps this original quantization scale Q _Old, and can not transmit and this original quantization scale Q _OldThe quantization scale Q that identical this is new _NewTo quantizer 16.Therefore, quantizer 16 can be according to this original quantization scale Q _OldM quantizes to this macro zone block.Otherwise, if this new quantization scale Q _NewWith this original quantization scale Q _OldAnd unequal, then carry out step 210, further check the quantization scale Q that this is new _NewWith this original quantization scale Q _OldBetween amplitude of variation.

In a preferred embodiment of the present invention, bit rate controller 120 can be checked this new quantization scale Q in step 210 _NewWith this original quantization scale Q _OldAmplitude of variation, whether surpass this preset range QTH, that is check | Q _New-Q _Old| whether 〉=QTH sets up.If | Q _New-Q _Old| 〉=QTH sets up, and then represents for these front and back two macro zone block L and M its corresponding quantization scale Q _NewWith Q _OldBetween amplitude of variation greater than this preset range.Because the quantization scale size that bit rate controller 120 is calculated is the complexity of this macro zone block of expression with regard to a certain degree.Therefore, when the quantization scale intensity of variation of these front and back two macro zone block L and M is very big, but use same quantization scale (that is Q _Old) this macro zone block M is quantized, the macro zone block M that may significantly improve this relative complex after quantification distortion level or may consume multidigit at simple relatively this macro zone block M.So, in a preferred embodiment of the present invention, as the quantization scale Q of these front and back two macro zone block L and M _OldWith Q _NewIntensity of variation when surpassing this preset range QTH, bit rate controller 120 can transmit these new quantization scale Q _NewTo quantizer 16, next to quantize the foundation of this macro zone block M as quantizer 16, the value of the change frequency Freq_Q of quantization scale limits and be not subjected at this moment.Relatively, in step 210, if the quantization scale intensity of variation of these front and back two macro zone block L and M does not reach critical value, that is | Q _New-Q _Old| whether during＜QTH, then bit rate controller 120 can carry out step 212, to determine with this new quantization scale Q _NewReplace this original quantization scale Q _Old

In step 212, bit rate controller 120 can check whether the value of the change frequency Freq_Q of quantization scale this moment is equal to or greater than this default value 10.If this moment, the value of change frequency Freq_Q of quantization scale was more than 10 or 10, then bit rate controller 120 can carry out step 216, informs that quantizer 16 keeps this original quantization scale Q _Old, and quantization scale Q that can this is not new _NewBe sent to quantizer 16.Therefore, quantizer 16 next still can be according to this original quantization scale Q _OldQuantize this macro zone block M.At last, in the coded bit stream of output, do not change quantization scale when only noting down quantizer 16 this macro zone block of quantification M behind variable length coder 18 codings, the quantization scale of a macro zone block is continued to use in representative.Repeat to note down this quantization scale Q and need not take 7-8 position _Old

Otherwise if the value of the change frequency Freq_Q of quantization scale is less than 10 at this moment, then bit rate controller 120 can carry out step 214.In step 214,120 meetings of bit rate controller will this new quantization scale Q _NewBe sent to quantizer 16, make quantizer 16 next according to this new quantization scale Q _NewQuantize this macro zone block M.At last, in the coded bit stream of behind variable length coder 18 codings, exporting, employed this new quantization scale Q in the time of can using 7-8 position record quantizer 16 to quantize this macro zone block M _New

In step 218, no matter whether bit rate controller 120 is with this new quantization scale Q _NewReplace this original quantization scale Q _Old, bit rate controller 120 all can upgrade the value of quantization scale change frequency Freq_Q.Please note, because the time basis P of the calculating quantization scale change frequency that sets in step 204 is 20 macro zone blocks, therefore, when bit rate controller 120 upgrades the value of Freq_Q, can weed out this macro zone block M front the 20th macro zone block (just with present processing in macro zone block M at the coded program middle distance farthest, the oldest previous macro zone block), and with in this macro zone block M adding computer capacity, add up in the quantizing process of this macro zone block M 19 macro zone blocks preceding (totally 20 macro zone blocks) with it, the change frequency of quantization scale is with the value as the Freq_Q after upgrading.

Whether the purpose of step 220 is to judge after this macro zone block M, also have other macro zone block to quantize.If this macro zone block M is last macro zone block, then finish flow process of the present invention; If also have other macro zone block to need to quantize after this macro zone block M, then bit rate controller 120 can carry out step 206, and generation one corresponds to the quantization scale of next macro zone block N of this macro zone block M, repeats above-mentioned steps then.

By finding in the explanation of above-mentioned flow process Figure 200 that step 204 can be put upside down with the order of step 206, and can not cause harmful effect to enforcement of the present invention.In addition, be that time basis P that hypothesis is calculated the quantization scale change frequency is that the higher limit TH of 20 macro zone blocks, quantization scale change frequency is 10 in the explanation of above-mentioned steps 204, the initial value of the change frequency Freq_Q of quantization scale is 0, be QTH with the critical value of quantization scale amplitude of variation.Then, the above-mentioned parameter value, only be for convenience of description for example, and non-limiting parameter value used in the present invention needs as previously mentioned.Hereat, each parameter value that sets in the step 204 can be adjusted according to need.For example, the time basis P that calculates the quantization scale change frequency also can be set at the time of a length-specific, or calculates benchmark At All Other Times.And the upper limit TH of quantization scale change frequency can be set at any number greater than 0 (integer or non-integer all can).The initial value of the change frequency Freq_Q of quantization scale can be set at 0, also can be set at any number of non-0.Critical value as for the quantization scale amplitude of variation can be set at a special value or a particular percentile or the like.

Even step 204 also can further be omitted, as long as write needed these parameter values also effect of attainable cost invention in the firmware of control bit rate controller 120 runnings in advance.

In a preferred embodiment of the present invention, if this new quantization scale Q _NewWith this original quantization scale Q _OldWhen unequal, then need to look this new quantization scale Q _NewWith this original quantization scale Q _OldBetween amplitude of variation, and relatively this moment quantization scale change frequency Freq_Q value and this implied value T H, whether to determine with this new quantization scale Q _NewReplace this original quantization scale Q _OldAnd in another embodiment of the present invention, the step 210 among flow process Figure 200 can be omitted.That is, if this new quantization scale Q _NewWith this original quantization scale Q _OldWhen unequal, only relatively this moment quantization scale change frequency Freq_Q value and this implied value T H, can determine whether with this new quantization scale Q _NewReplace this original quantization scale Q _Old

Compared to known technology, an important technology of bit rate controller 120 of the present invention is characterised in that, can limit the change frequency of the central quantization scale of cataloged procedure, to be used for noting down the position amount of quantization scale in the coded bit stream that reduces last output, use for video signal data and vacate more multidigit, and then promote the quality of video signal coding.

More than be to be that example illustrates a preferred embodiment of the present invention with a video coder 100, in fact, video coder 100 also can be a computer system, and squares such as the discrete cosine conversion device 12 among Fig. 2, dynamic estimation and compensator 14, quantizer 16, variable length coder 18 and bit rate controller 120 all can be respectively corresponding program or be integrated into a video signal coding program.At this moment, flow process Figure 200 of Fig. 2 just is equivalent to the algorithm of this video signal coding program, and its detailed procedure code can be write by any program language, under the situation that does not influence technology exposure of the present invention, does not repeat them here.

The above only is preferred embodiment of the present invention, and all equalizations of being done according to the present patent application claim change and modify, and all should belong to the covering scope of patent of the present invention.

Claims

One kind when coding image signal control quantization scale method, this method includes:

A obtains the change frequency of this quantization scale in a period;

B produces a new quantization scale; And

C is this change frequency and a preset value relatively, whether replaces original quantization scale with this new quantization scale with decision;

Wherein, when this change frequency surpasses this preset value, then keep this original quantization scale.
2. the method for claim 1, wherein this preset value is the specific change number of times of this quantization scale in this period.
3. the method for claim 1, wherein this preset value is a specific change number of times of quantization scale in the quantification course of action of macro zone block of a predetermined number in carrying out this picture signal.
4. the method for claim 1, it comprises in addition and uses a Bit-Rate Control Algorithm device with execution in step a and step b.
5. the method for claim 1, it comprises in addition by carrying out a video coding program to carry out step a and step b.
6. the method for claim 1, it includes relatively this new quantization scale and this original quantization scale in addition, when if the difference of this new quantization scale and this original quantization scale surpasses a preset range, no matter whether the change frequency of this quantization scale surpasses this preset value, all replace original quantization scale with this new quantization scale.
7. the method for claim 1, it includes relatively this new quantization scale and this original quantization scale in addition, when if the difference of this new quantization scale and this original quantization scale is no more than a preset range, relatively whether the change frequency of this quantization scale and this preset value replace this original quantization scale with this new quantization scale with decision.
8. the method for claim 1, wherein this period is a Preset Time length, or quantizes the required special time of macro zone block of a predetermined number in this picture signal.
9. a video quantizing device is used for a picture signal is quantized, and to produce a matrix after quantizing, this video quantizing device includes:

One Bit-Rate Control Algorithm device is used to provide quantization scale; And

One quantizer is connected in this Bit-Rate Control Algorithm device, is used for according to this quantization scale this picture signal being quantized;

Wherein, this Bit-Rate Control Algorithm device can be obtained the change frequency of this quantization scale when coding image signal, and compare with a preset value, when this Bit-Rate Control Algorithm device produces a new quantization scale, if this change frequency surpasses this preset value, then this Bit-Rate Control Algorithm device can not replace the original quantization scale of this quantizer in using with this new quantization scale.
10. quantization device as claimed in claim 9, wherein, this preset value is the specific change number of times of this quantization scale in this period.
11. quantization device as claimed in claim 9, wherein, this preset value is a specific change number of times of quantization scale in the quantification course of action of macro zone block of a predetermined number in carrying out this picture signal.
12. quantization device as claimed in claim 9, wherein, this Bit-Rate Control Algorithm device is this new quantization scale and this original quantization scale poor relatively, when surpassing a preset range as this difference, no matter whether the change frequency of this quantization scale surpasses this preset value, this Bit-Rate Control Algorithm device all can provide this new quantization scale to give this quantizer, to replace this original quantization scale.
13. quantization device as claimed in claim 9, wherein, this Bit-Rate Control Algorithm device is this new quantization scale and this original quantization scale poor relatively, when if this difference is no more than a preset range, this Bit-Rate Control Algorithm device is change frequency and this preset value of this quantization scale relatively, whether replaces this original quantization scale with this new quantization scale with decision.
14. the method for a coded digital picture signal, this data image signal includes a plurality of macro zone blocks, and this method includes:

A utilizes the one first quantization scale target macro block of encoding;

B obtains the quantization scale change frequency of quantization scale; And

C judges whether to use instead one second quantization scale according to the quantization scale change frequency of this quantization scale and replaces this first quantization scale next macro zone block of encoding.
15. method as claimed in claim 14, wherein, step b is by the change frequency of this quantization scale of counting in a scheduled time slot, or carries out at the change frequency of the macro zone block that quantizes a predetermined number in the required period.
16. method as claimed in claim 15, wherein, step c is undertaken by a number of times and a preset value that this quantization scale relatively changes in a scheduled time slot.
17. method as claimed in claim 14, wherein, step c comprises poor according to the quantization scale change frequency of this quantization scale and this first, second quantization scale in addition, judges whether to use instead these rank, second amount position and replaces this first quantization scale next macro zone block of encoding.
18. a code device, being used to encode comprises the data image signal of a plurality of macro zone blocks, and this code device includes:

One Bit-Rate Control Algorithm device is used to provide a quantization scale; And

One quantizer is electrically connected on this Bit-Rate Control Algorithm device, is used to receive this quantization scale so that this each macro zone block is quantized;

Wherein, whether this quantization scale will replace with new numerical value, is the change frequency that depends on the quantization scale of the macro zone block that is before quantized.
19. code device as claimed in claim 18, wherein, when the change frequency of the quantization scale of the previous macro zone block that quantizes surpassed one first critical value, this quantizer can not change the quantization scale that is adopted.
20. code device as claimed in claim 18, wherein, when the difference of this new numerical value and last numerical value surpasses a preset level, no matter whether the change frequency of the quantization scale of the previous macro zone block that quantizes surpasses this first critical value, this quantizer all can be used this new numerical value instead as quantization scale.
21. code device as claimed in claim 18, wherein, the mode of the change frequency of the quantization scale of the previous macro zone block that quantizes of appraisal, be by the change frequency of this quantization scale of counting in a scheduled time slot, or carry out at the change frequency of the macro zone block that quantizes a predetermined amount in the required period.