CN1889693A - Gamma characteristic correcting method and detecting method for determining equivalent model and parameter thereof - Google Patents

Abstract

A method for calibrating gamma characteristic includes confirming calibration point and setting Na links before calibration point and Np links after calibration point, selecting the first equivalent model of equivalent Na links gamma characteristic and confirming its the first inverse model, selecting the second equivalent model of equivalent Np links gamma characteristic and confirming its the second inverse model, forming calibration link model according to two said inverse models, utilizing calibration link model to confirm calibration signal of the last output signal in Na links and inputting said calibration signal to Np links.

Description

The detection method of the bearing calibration of gamma characteristic and definite equivalent model and parameter thereof

Technical field

The present invention relates to video communication technology, particularly determine the detection method and the bearing calibration of gamma characteristic equivalent model and parameter thereof in video signal transmission and the processing links.

Background technology

It is the problem that ubiquity also needs solution in the multimedia information system that Gamma proofreaies and correct, for the graphics/animation that generates by video camera/camera video captured/rest image, computer of multimedia information system and in communication from the video/rest image of distant terminal etc., in order on display device, to reach high-quality display effect, obtain good user experience (User Experience, perhaps be called Quality ofExperience), must carry out Gamma and proofread and correct.

The situation that multimedia information system related to the present invention relates to the Gamma correction comprises following aspect:

Situation A: the local terminal system, information spinner will be handled on the local terminal, does not relate to communication process.Mainly be local video camera/camera video captured/rest image, the situation that the graphics/animation that computer generates etc. shows on local display screen.Such as PC, PDA (Personal Digital Assistant) or have mobile phone of camera function etc.

Situation B: multiple terminals/terminal-server communication systems, communicate by letter between a plurality of terminals, perhaps communicate by letter between terminal and the server.Mainly be video telephone, have high-end (2.5G/3G, B3G) mobile phone, PDA of MMS (multi-media SMS) or video communication function etc.

Video communication is obtaining at present increasingly extensive application along with developing rapidly of broadband network, at home and in the world, video conference and visual telephone service are becoming the basic service on the NGN (Next Generation Network next generation network).The telecom operators of various countries also pay much attention to this market opportunity, can be expected in the coming years, and video communication service will become the important business growth point of operator.A key issue that develops this type of business is to improve the user experience (UserExperience perhaps is called Quality of Experience) of end-to-end (End-to-end).(packet loss postpones, shake except the QoS of network in the user experience, the R factor etc.) outside the parameter, for video, because the Gamma nonlinear problem that each link causes, causing the distortion (Distortion) for luminance signal, also is to influence the key factor that the end user experiences.But at present, mainly concentrate on the assurance network QoS pre-process and post-process (Pre-processing relevant for method that improves end-to-end user experience and technology with video compression coding, Post-processing) aspect, and the brightness distortion problem that causes for the Gamma characteristic, the solution that lacks concern and system, but the seriousness of this problem has caused concerns of some international big telecom operators.France Telecom (France Telecom) has just proposed to consider the influence that the Gamma characteristic is experienced for communication user in the recent period at the ITU-T of International Telecommunications Union in video communication, and the suggestion that this type of problem is solved.

In a multi-media information terminal (hereinafter referred terminal), the light signal of extraneous scene (personage, background, file etc.) enters into video camera/camera, converts data image signal to through A/D, then:

Under situation A, directly deliver on the display device and show, finally become light signal again by the human eye perception.Certainly may handle through some in the middle of, such as for the figure image intensifying that promotes image effect etc.;

Under situation B, data image signal is through condensing encoder (Encoder) compression, send out the arrival distant terminal by network (wired, wireless, circuit switching, packet switching network can) then, go compression (decompression) decoding to be reduced to data image signal at distant terminal through decoder (Decoder), on display device, show again, finally become light signal again by the human eye perception.

As shown in Figure 1, Fig. 1 is the model schematic diagram of link Gamma characteristic, no matter situation A or situation B, (Luminance is a kind of luminance signal of broad sense to image brightness signal in this process here, and promptly light signal at the beginning is to the signal of telecommunication, arrive digitized image brightness/grey scale signal again, the signal in each stage all contains the information of luminance signal, and therefore in broad terms, luminance signal has been passed through a plurality of links) passed through a plurality of links.According to definition, the Gamma characteristic just is meant that the luminance signal input-output relation of a link is not linear, but a kind of non-linear, shown in the curve among Figure 10.

The influence of Gamma nonlinear element distortion as shown in Figure 2, top delegation's gray scale square brightness is linear increment, from 0.1 to 1.0, below delegation distort through the Gamma nonlinear element, brightness increases progressively according to the power function rule.

In practice, Gamma is non-linear is caused that by different reasons for CRT (Cathod Ray Tube, cathode ray tube) display devices such as television set, PC displays, the Gamma characteristic satisfies formula 1 under ideal state:

L _out＝L _in ^2.2 (1)

Formula 1 is a power function (Power Function).Need to prove that the input and output luminance signal here all is to have carried out normalization (Normalized), i.e. 0≤L in coordinate space separately _Out≤ 1,0≤L _In≤ 1.

And for the display of other types such as mobile phone, PDA such as liquid crystal etc., the form of its Gamma function or different in form or parameter difference.

As shown in Figure 3, Fig. 3 is the model schematic diagram of a plurality of link cascades (cascading or the be called series connection) link of getting up Gamma characteristic, and total Gamma characteristic equals compound (composition) of each link Gamma function, satisfies formula 2:

G _CT(.)＝G ⁽¹⁾(.)оG ⁽²⁾(.)оG ⁽³⁾(.)……G ^(n-1)(.)оG ⁽ⁿ⁾(.)

l _out＝GCT(l _in)＝G(n)(G ^(n-1)(G ^(n-2)(……G ⁽²⁾(G ⁽¹⁾(l _in))))) (2)

The compound operation of " о " representative function.CT represents cascaded total, i.e. the meaning of the total Gamma of cascade.

For situation A: a plurality of Gamma links that relate to mainly contain as shown in Figure 4:

1, video camera/camera Gamma is expressed as G _Cam(.);

General video camera all has the Gamma characteristic, except image device non-linear such as CCD self, video camera has been introduced artificial non-linear, its objective is the Gamma characteristic that allows the Gamma characteristic of video camera just compensate display, makes that total Gamma characteristic is linear.If the desirable Gamma of display is: L _Out=Lin2.2; The desirable Gamma of video camera is so: L _Out=L _In ^0.45

Therefore in theory, the Gamma characteristic of video camera is by the Gamma characteristic decision of display.But because terminal system is complicated day by day, there are a plurality of links between video camera and the display, its number is indefinite, Gamma characteristic separately is also unknown, even the Gamma of video camera and display just in time mates and can compensate mutually like this, but because intermediate link exists, this compensation is invalid generally speaking.And type of display is numerous, for example: displays such as CRT and liquid crystal, plasma, its Gamma characteristic is mutually far short of what is expected, and the Gamma characteristic of cheap camera its desirable Gamma of substantial deviation often.

2, storage file Gamma is expressed as G _Fil(.);

File may pass through processing from video camera, and compressed encoding also experienced a plurality of Gamma links, so file itself has carried the Gamma characteristic.

3, display frame is deposited Gamma, is expressed as G _FBuf(.);

Early stage display is because show that the color depth of storage is not enough, such as can only supporting 4,8,16 color depths, rather than 24 desirable true color, equaled to compress the dynamic range of input luminance signal, therefore also introduced the Gamma characteristic.In addition, because under pattern under the non-true color, the palette of use (Palette) color map technology or shake (Dither) technology etc. all can be introduced non-linear Gamma.

4, show the Gamma that tables look-up, be expressed as G _LUT(.);

Some display device in order to compensate the non-linear of display, has artificially been introduced Gamma, and this Gamma shows as a LUT (Look-Up Table), and the brightness data of reading from frame is deposited will be through the LUT conversion, just remove driving display.

5, display Gamma is expressed as G _Disp(.).

It is non-linear that general display has very strong Gamma.

For situation B: a plurality of Gamma links that relate to mainly contain as shown in Figure 5:

1, video camera/camera Gamma is expressed as G _Cam(.);

2, storage file Gamma is expressed as G _Fil(.);

3, display frame is deposited Gamma, is expressed as G _FBuf(.);

4, show the Gamma that tables look-up, be expressed as G _LUT(.);

5, display Gamma is expressed as G _Disp(.);

6, encoder Gamma is expressed as G _Enc(.);

Because the Gamma that DCT (the Discrete Cosine Transform) conversion in the compression, quantification cause.

7, decoder Gamma is expressed as G _Dec(.).

Because the DCT inverse transformation in decompressing, the Gamma that inverse quantization causes.

For situation B, even more serious is, local video/image, far-end video and the Gamma link of encircling video (being used for specific purposes such as fault diagnosis etc.) process certainly are different, in addition, no matter situation A or B, the Gamma link that relates in truth may be more, so situation is complicated more.

Ideal situation is that input optical signal is exported light signal from entering camera to finally showing at display screen, has linear relationship between the input and output luminance signal, that is: L _Out=L _In, like this scenery seen of people just and original just the same, user experience is best.

Obtain linear relationship, must carry out Gamma correction (Gamma Correction) for having non-linear Gamma characteristic link.As shown in Figure 6, for a link, its Gamma characteristic is given, can carry out cascade with another one correction link and it so, make that total Gamma characteristic is called real linear relationship after the cascade, fall the nonlinear purpose of given link thereby reached compensation, the model of correction link is the inversion model of Gamma characteristic equivalent model, if equivalent model can represent that then the functional relation of inversion model is its inverse function with functional relation.Obviously, G _g(.) and G _c(.) be inverse function each other.Generally speaking, for a function, obtain its inverse function and not necessarily separate (, also can't obtain) with Calculation Method even perhaps separate existence.

In the practical application more situation as shown in Figure 7, correction link need be inserted between former and later two given links, this moment G _c(.) situation is complicated more, G _c(.) and G _a(.) or G _p(.) no longer is simple inverse function relation.

The implementation method of correction link mainly contains following two kinds in the prior art:

Prior art one: the Gamma characteristic that relies on video camera/camera or demonstration LUT is fully proofreaied and correct display Gamma characteristic

Suppose under the perfect condition: G _Cam(.): L _Out=L _In ^0.45G _LUT(.): L _Out=L _In ^0.45G _Disp(.): L _Out=L _In ^2.2

Then: G _Camо G _Disp(.) becomes: L _Out=L _In, the linear relationship of formation standard; G _LUTо G _Disp(.) becomes: L _Out=L _In, the linear relationship of formation standard.

But there is following deficiency in above-mentioned technology:

Perfect condition is to be difficult to obtain, and can not guarantee that the Gamma of video camera/camera, LUT just mates fully with display Gamma.And type of display is a lot, and the Gamma of cheap camera is imperfect certainly; If G _Cam(.), G _LUT(.) exists simultaneously, overcompensate then, G _Camо G _LUTо G _Disp(.) becomes: L _Out=L _In ^0.45, departed from linearity on the contrary; The Mathematical Modeling of simulation gamma characteristic is inaccurate, studies show that much the Gamma of display is not simple power function, and more precise analytic model may be the linear combination of power function and constant function, or linear function and power function is compound.

Prior art two:

Between some link,, insert a Gamma correction link and carry out Gamma and proofread and correct such as after the video camera link or before display frame deposits link.In addition, may aspect the Gamma of display characteristic model, adopt precise analytic model more, such as formula 3:

$L_{\mathrm{out}}=\left\{\begin{array}{c}\frac{1}{0.45}{L}_{\mathrm{in}},\mathrm{if}0\&le;L_{\mathrm{in}}\&le;0.081\\ \frac{1}{1.099}{(L_{\mathrm{in}}+0.099)}^{2.2},\mathrm{if}0.081<L_{\mathrm{in}}\&le;1\end{array}\right.---\left(3\right)$

Accordingly, the Gamma of video camera is considered to mate fully with display Gamma, such as formula 4:

$L_{\mathrm{out}}=\left\{\begin{array}{c}0.45L_{\mathrm{in}},\mathrm{if}0\&le;L_{\mathrm{in}}\&le;0.081\\ 1.099{L_{\mathrm{in}}}^{0.45}-0.099,\mathrm{if}0.081<L_{\mathrm{in}}\&le;1\end{array}\right.---\left(4\right)$

Prior art two shortcomings are:

Calibration model is single, and the situation of a plurality of links is very complicated, and the Gamma characteristic of a plurality of links fails accurately to obtain, and must cause and can not accurately proofread and correct, and promptly proofreaies and correct the result and remains nonlinear, even accurately some also are not have help to model.Therefore, still can not avoid existing the problem of overcorrect or undercorrection; And this method can only be applied in some particular cases, can't be common to the correction of any a plurality of Gamma links.

Summary of the invention

The invention provides the detection method of a kind of definite gamma equivalent model and parameter thereof, with solve in the prior art that the gamma equivalent model is selected and model parameter measure inaccurate, thereby can't accurately analyze and correction signal is transmitted and processing links in gamma characteristic;

Simultaneously, based on above-mentioned detection method, provide a kind of bearing calibration that is applicable to the gamma characteristic of monocycle joint and too many levels simultaneously, to solve the problem that existing bearing calibration can't be general.

A kind of gamma characteristic correcting method is used for proofreading and correct the N that described multimedia information system or signal specific comprise in described multimedia information system transmission and processing procedure _tIndividual link with gamma characteristic, described method comprises the steps:

Determine check point, this check point is with described N _tIndividual link is divided into and is positioned at this check point N before _aIndividual link and be positioned at N after this check point _pIndividual link, wherein: N _a〉=0, N _p〉=0, N _a+ N _p=N _t

Determine equivalent described N _aFirst equivalent model of individual link gamma characteristic and first inversion model thereof are determined equivalent described N _pSecond equivalent model of individual link gamma characteristic and second inversion model thereof;

According to described first inversion model and the described second inversion model structural correction link model, utilize this correction link model to determine described N _aThe correction signal of individual link final output signal is also imported described N with this correction signal _pIndividual link.

When described equivalent model adopted the function representation form, the representative function of described inversion model was corresponding inverse function.

Described definite described first equivalent model or second equivalent model comprise the steps:

A1, detect N sampled value L respectively with input signal _In(i) N value L of the real output signal of the described link generation of input ^P _Out(i), wherein: 0≤i≤N-1;

A2, in the equivalent model of one group of alternative employing function representation form, select a model execution in step A3 to be measured;

A3, for described model to be measured, select one group of initial parameter;

A4, calculating:

N value L according to the theoretical output signal of described N sampled value correspondence of described Model Calculation to be measured ^M _Out(i), wherein: 0≤i≤N-1; And

Target function value F, described desired value functional value F with each to corresponding L ^P _Out(i) and L ^M _Out(i) difference is relevant;

A5, judge whether described target function value F is equal to or less than the threshold value of setting, if think that then described model to be measured is accepted as final equivalent model, and should change steps A 8 over to after the parameter of target function value F corresponding parameters as this equivalence model of minimum; Otherwise execution in step A6;

Whether the execution number of times of A6, determining step A4 arrives the iterations of qualification, if then select one again as model to be measured and return steps A 3 from other alternative equivalent model that does not detect as yet; Otherwise execution in step A7;

A7, utilize described mathematical optimization method to adjust described model parameter, return steps A 4;

A8, end.

In the described steps A 5, when described target function value F is equal to or less than the threshold value of setting, again according to the cycle-index of setting, utilize described mathematical optimization method to adjust described parameter and calculate described target function value F, target function value F corresponding parameters that will be wherein minimum is as the parameter of described equivalent model then.

Described method also comprises the steps:

Respectively the parameter that measures is brought in the representative function of corresponding inversion model, ask for first inverse function formula of corresponding first inversion model and the second inverse function formula of corresponding second inversion model;

Utilize this first inverse function formula and the second inverse function formula structural correction model.

It is one of following that the building method of described correction link model comprises:

Direct computing method: utilize first inverse function and the second contrafunctional compound function to calculate N in real time _aThe correction signal of the final output signal of individual link;

Two step computing methods: utilize first inverse function to calculate N in real time _aA correction signal of the final output signal of individual link utilizes second inverse function to calculate the secondary correction signal of this correction signal, with this secondary correction signal as described correction signal;

Look-up table: according to described direct computing method or two step computing methods, calculate described N in advance _aThe corrected value of the correspondence of a plurality of sampled values in the interval of the final output signal of individual link, and corresponding relation is kept in the tables of data, determine the corrected value of any value to be corrected then by this tables of data of real-time inquiry.

In the described look-up table: when the value of importing to be corrected is in tables of data, then directly obtain corresponding corrected value by tabling look-up; When the value of importing to be corrected is not in the input of tables of data is listed as, take the average method of linear interpolation to ask for corresponding corrected value.

Described target function value F satisfies following condition:

$F=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-L_{\mathrm{out}}^M\left(i\right))}^2$

The functional relation of described one group of alternative equivalent model and corresponding inverse function thereof comprise:

Described gamma characteristic equivalent model functional relation is: L _Out=pL _In ^α+ (1-p), wherein: the domain of definition of this function is interval [0,1], codomain is interval [(1-p), 1]; Then the inverse function relational expression of this function is:

$L_{\mathrm{out}}={(\frac{1}{p}{L}_{\mathrm{in}}+(1-\frac{1}{p}))}^{\frac{1}{\mathrm{\&alpha;}}};$ Perhaps,

Described gamma characteristic equivalent model functional relation is: $L_{\mathrm{out}}={({\mathrm{qL}}_{\mathrm{in}}+(1-q))}^{\frac{1}{\mathrm{\&beta;}}},$ Wherein: the domain of definition of this function is interval [1-1/q, 1], and codomain is interval [0,1]; Then the inverse function relational expression of this function is:

$L_{\mathrm{out}}=\frac{1}{q}{L_{\mathrm{in}}}^{\mathrm{\&beta;}}+(1-\frac{1}{q});$

Wherein: L _InBe input signal values, L _OutFor being respectively, output signal value, p and α and q and β need the parameter measured; And work as N _a=1 or N _p=1 o'clock: 0＜p≤1, α 〉=1, q 〉=1, β 〉=1; Work as N _a＞1 or N _p＞1 o'clock: 0＜p≤1, α 〉=0, q 〉=1, β 〉=0.

Described mathematical optimization method includes but not limited to one of following method: climbing method; 0.618 method; Steepest descent method; Conjugate gradient method.

The N of a described input signal sampled value is selected in interval [0,1].

When described equivalent model was the tables of data form, corresponding described inversion model was exactly the contrary table of this tables of data.

The detection method of a kind of definite gamma characteristic equivalent model and parameter thereof, described gamma characteristic equivalent model are used for the gamma characteristic of equivalent signal transmission or processing links, comprise the steps:

B1, detect N sampled value L respectively with input signal _In(i) N value L of the real output signal of the described link generation of input ^P _Out(i), wherein: 0≤i≤N-1;

B2, in the equivalent model of one group of alternative employing function representation form, select a model execution in step B3 to be measured;

B3, for described model to be measured, select one group of initial parameter;

B4, calculating:

N value: L according to the theoretical output signal of the described signal N of a described Model Calculation to be measured sampled value correspondence ^M _Out(i), wherein: 0≤i≤N-1; And

Target function value F, described desired value functional value F with each to corresponding L ^P _Out(i) and L ^M _Out(i) difference is relevant;

B5, judge whether described target function value F is equal to or less than the threshold value of setting,, and change step B8 over to after will organizing the parameter of parameter as this equivalence model if think that then described model to be measured is accepted as final equivalent model; Otherwise execution in step B6;

Whether the execution number of times of B6, determining step B4 arrives the iterations of qualification, if then select one again as model to be measured and return step B3 from other alternative equivalent model that does not detect as yet; Otherwise execution in step B7;

B7, utilize described mathematical optimization method to adjust described parameter, return step B4;

B8, end.

Among the described step B5, when described target function value F is equal to or less than the threshold value of setting, again according to the cycle-index of setting, utilize described mathematical optimization method to adjust described parameter and calculate described target function value F, target function value F corresponding parameters that will be wherein minimum is as the parameter of described equivalent model then.

Beneficial effect of the present invention is as follows:

The invention solves selection problem and the detection problem of parameter and the problem of link Gamma characteristic correction of ubiquitous Gamma characteristic model in the multimedia information system, provide a kind of general bearing calibration for the comprehensive Gamma characteristic of any too many levels cascade.Thereby can improve the user experience of multimedia information system greatly.

Description of drawings

Fig. 1 is the universal model of link Gamma characteristic;

The schematic diagram of the luminance signal distortion that Fig. 2 causes for link Gamma characteristic;

Fig. 3 is the universal model of too many levels cascade Gamma characteristic;

Fig. 4 is a plurality of Gamma link schematic diagrames under the situation A;

Fig. 5 is a plurality of Gamma link schematic diagrames under the situation B;

Fig. 6 is for proofreading and correct the Gamma characteristic schematic diagram of single link;

Fig. 7 is for proofreading and correct the Gamma characteristic schematic diagram of a plurality of given links;

Fig. 8 determines the schematic diagram of check point when using bearing calibration of the present invention;

Fig. 9 is for using the schematic diagram that utilizes two sub-links to carry out the front and back correction respectively of the present invention;

Figure 10 is a Gamma characterisitic function curve synoptic diagram;

Figure 11 a and Figure 11 b are respectively Gamma characteristic model function curve schematic diagram;

Figure 12 is a Gamma characteristic model function curve restricted area schematic diagram.

Embodiment

Studies show that, the functional relation of more accurate gamma characteristic equivalent model is the linear combination of power function and constant function, or linear function and power function is compound, in different environment, can obtain concrete functional relation comparatively accurately by the parameter in the actual measurement functional relation.

Embodiment one: the method for measurement of gamma characteristic equivalent model parameter

It is that example is carried out as the universal model of Gamma characteristic that invention selects to adopt following monocycle joint equivalent model

Describe in detail:

First kind Gamma model: L _Out=pL _In ^α+ (1-p) 0＜p≤1, α 〉=1 (5)

Wherein: the domain of definition of function shown in the formula 5 (being the independent variable span) is interval [(1-p), 1] for interval [0,1], codomain (span of functional value), and curve characteristic is shown in Figure 11 a, and obviously for first kind model, if p=1, then there is L α=1 _Out=L _InTherefore when actual measurement, if the parameter p of this first kind model=1, α=1 o'clock, perhaps p fully near 1 and α fully near 1 o'clock, can think that the gamma characteristic of link can be left in the basket, need not proofread and correct.The exemplary of this first kind Gamma characteristic is a CRT monitor.

The second class Gamma model: $L_{\mathrm{out}}={({\mathrm{qL}}_{\mathrm{in}}+(1-q))}^{\frac{1}{\mathrm{\&beta;}}}---q\&GreaterEqual;1,\mathrm{\&beta;}\&GreaterEqual;1---\left(6\right)$

Wherein: the domain of definition of function shown in the formula 6 (being the independent variable span) is interval [1-1/q, 1], and codomain (span of functional value) be interval [(0,1], and curve characteristic is shown in Figure 11 b, and obviously for second class model, if q=1, then there is L β=1 _Out=L _InEqually, when actual measurement, if q=1, β=1 o'clock, perhaps q fully near 1 and β fully near 1 o'clock, can think that the gamma characteristic of link can be left in the basket, need not proofread and correct, and the exemplary of this second class feature is a video camera.

In addition, if q=1/p, α=β, therefore the first kind and second class model inverse function each other can compensate mutually and obtain linear characteristic so.If that is: given link has first kind Gamma characteristic to picture, its correction Gamma characteristic has second class model so; If given link has the second class Gamma characteristic to picture, its correction Gamma characteristic has first kind model so.

After a plurality of Gamma links were carried out cascade, the Gamma characteristic that it is comprehensive from the mathematics generality, no longer had the first kind or second class model of monocycle joint.But we find that under study for action the Mathematical Modeling of the comprehensive Gamma characteristic of too many levels has following characteristics:

1, its functional image be confined to coordinate plane zone [0,1] * [0,1] (here " * " expression two set cartesian product Cartesian Product, perhaps be called direct product) in, as shown in figure 12;

2, monotonic increase;

3, epirelief and protruding down;

Two kinds of situations go up from how much meanings, epirelief be curve to upper left bending, following protruding be that curve is to the bottom right bending.

4 and the Lin axle intersect, perhaps intersect (two kinds of situations) with the Lout axle;

5, curve negotiating (1,1) point;

Therefore, 3,4 situation independently makes up and just forms four kinds of combining cases:

Situation 1 and L _InAxle intersects, epirelief;

Situation 2 and L _InAxle intersects, and is following protruding;

Situation 3 and L _OutAxle intersects, epirelief;

Situation 4 and L _OutAxle intersects, and is following protruding.

According to above qualitative analysis, the comprehensive Gamma characteristic model of a plurality of Gamma link cascade can have following two classes, and functional relation is respectively formula 7 and formula 8:

The first kind: L _Out=pL _In ^α+ (1-p) 0＜p≤1, the corresponding situation 3 in α＞0 and 4 (7)

Second class: $L_{\mathrm{out}}={({\mathrm{qL}}_{\mathrm{in}}+(1-q))}^{\frac{1}{\mathrm{\&beta;}}}$ Q 〉=1, the corresponding situation 1 in β＞0 and 2 (8)

It may be noted that, from in form, too many levels (wherein can have first kind link, the second class link also can be arranged, two class links can be according to any amount and concatenated in order, method of the present invention all is suitable for) first kind of comprehensive Gamma characteristic, second class respectively with the monocycle joint Gamma characteristic first kind, second class is the same.But in the first kind composite characteristic model, according to the empirical value of The qualitative analysis and actual measurement, index α span becomes α＞0, and in the second class composite characteristic model, index β span becomes β＞0.

After having selected the Gamma characteristic model, need measure Gamma characterisitic parameter wherein in concrete applied environment, whether the functional relation that input/output signal satisfies in the last Gamma characteristic model of method of measurement direct relation is accurate.Wherein:, need to determine parameter p and α for first kind model; For second class model, need to determine parameter q and β.

The measurement concrete steps of monocycle joint Gamma characteristic model parameter are as follows:

1, at input luminance signal L _InOn [0,1] interval, select at interval uniform N sampled point: L _In(0), L _In(1), L _In(2) ... L _In(i) ... L _In(N-2), L _In(N-1);

2,, and measure N corresponding value: L of actual output luminance signal with in luminance signal N the sampled value difference input element ^P _Out(0), L ^P _Out(1), L ^P _Out(2) ... L ^P _Out(i) ... L ^P _Out(N-2), L ^P _Out(N-1);

3, the target function of structure match is, the output luminance signal of target function and actual detected is relevant with the difference that the theory of determining by the Gamma characteristic model is exported between the luminance signal, and difference is more little, and the equivalent effect that model is described is more near actual conditions.

The building method of target function is a lot, and comparatively commonly used is following formula 9 or formula 10:

$F_{T1}(p,\mathrm{\&alpha;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{\mathrm{pL}}_{\mathrm{in}}{\left(i\right)}^{\mathrm{\&alpha;}}-(1-p))}^2---\left(9\right)$ Perhaps,

$F_{T2}(q,\mathrm{\&beta;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{({\mathrm{qL}}_{\mathrm{in}}\left(i\right)+(1-q))}^{\frac{1}{\mathrm{\&beta;}}})}^2---\left(10\right)$

4, the thresholding T of target setting functional value and maximum iteration time M utilize the mathematical optimization method to seek optimal parameter group;

At first for the cost function of the first kind $F_{T1}(p,\mathrm{\&alpha;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{\mathrm{pL}}_{\mathrm{in}}{\left(i\right)}^{\mathrm{\&alpha;}}-(1-p))}^2,$ Adopt certain mathematical optimization technology, for example: climbing method, 0.618 method (Hua Luogeng optimum seeking method), steepest descent method or conjugate gradient method etc. are asked for its minimum value;

This process is an iterative process in fact, constantly adjusts parameter p and α in this process, and functional value F, then thinks and found smallest point after functional value drops to less than given thresholding T in continuous decline.This moment, corresponding parameters p and α just thought the real parameter of this applied environment model to should be noted in the discussion above that the span of parameter p and α is respectively: 0＜p≤1, α 〉=1;

If for $F_{T1}(p,\mathrm{\&alpha;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{\mathrm{pL}}_{\mathrm{in}}{\left(i\right)}^{\mathrm{\&alpha;}}-(1-p))}^2$ Through M iteration, can't make function drop to below the thresholding T, think that then Model Selection is not right.Should select second class model, so for $F_{T2}(q,\mathrm{\&beta;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{({\mathrm{qL}}_{\mathrm{in}}\left(i\right)+(1-q))}^{\frac{1}{\mathrm{\&beta;}}})}^2$ Repeat above-mentioned steps 4, obtain corresponding model parameter q and β, should be noted in the discussion above that the span of parameter p and α is respectively: q 〉=1, β 〉=1.

If want to obtain more precise parameters, can be after target function value F drops to below the thresholding T, still iteration is several times again, if target function value F continues to descend, or rise again after descending, or directly rise, no matter which kind of situation of change target function value F is, then select minimum value corresponding parameters wherein can improve the precision of parameter measurement to a certain extent as measurement result.

Can see, types of models determine and the measurement of parameter is carried out simultaneously, in the reality, the type of equivalent model is these two kinds of forms not only, can find only one in the method for all relevant equivalent models by measurement parameter by said method.

Can utilize said method to measure the comprehensive Gamma characteristic model of too many levels parameter equally, concrete steps are as follows:

1, at input luminance signal L _InOn [0,1] interval, select at interval uniform N sampled point: L _In(0), L _In(1), L _In(2) ... L _In(i) ... L _In(N-2), L _In(N-1);

2,, and measure N corresponding value: L of actual output luminance signal with in luminance signal N the sampled value difference input element ^P _Out(0), L ^P _Out(1), L ^P _Out(2) ... L ^P _Out(i) ... L ^P _Out(N-2), L ^P _Out(N-1);

3, the target function of structure match is, the output luminance signal of target function and actual detected is relevant with the difference that the theory of determining by the Gamma characteristic model is exported between the luminance signal, and difference is more little, and the equivalent effect that model is described is more near actual conditions.

The building method of target function is a lot, formula 9 or the formula 10 of remaining comparatively commonly used:

$F_{T1}(p,\mathrm{\&alpha;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{\mathrm{pL}}_{\mathrm{in}}{\left(i\right)}^{\mathrm{\&alpha;}}-(1-p))}^2---\left(9\right)$ Perhaps,

$F_{T2}(q,\mathrm{\&beta;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{({\mathrm{qL}}_{\mathrm{in}}\left(i\right)+(1-q))}^{\frac{1}{\mathrm{\&beta;}}})}^2---\left(10\right)$

4, the thresholding T of target setting functional value and maximum iteration time M utilize the mathematical optimization method to seek optimal parameter group;

At first for the cost function of the first kind $F_{T1}(p,\mathrm{\&alpha;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{\mathrm{pL}}_{\mathrm{in}}{\left(i\right)}^{\mathrm{\&alpha;}}-(1-p))}^2,$ Adopt certain mathematical optimization technology, for example: climbing method, 0.618 method (Hua Luogeng optimum seeking method), steepest descent method or conjugate gradient method etc. are asked for its minimum value;

This process is an iterative process in fact, constantly adjusts parameter p and α in this process, and functional value F, then thinks and found smallest point after functional value drops to less than given thresholding T in continuous decline.This moment, corresponding parameters p and α just thought the real parameter of this applied environment model to should be noted in the discussion above that measuring difference with the monocycle joint is that the span of parameter p and α is respectively: 0＜p≤1, α 〉=0;

Equally, if for $F_{T1}(p,\mathrm{\&alpha;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{\mathrm{pL}}_{\mathrm{in}}{\left(i\right)}^{\mathrm{\&alpha;}}-(1-p))}^2$ Through M iteration, can't make function drop to below the thresholding T, think that then Model Selection is not right.Should select second class model, so for $F_{T2}(q,\mathrm{\&beta;})=\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-{({\mathrm{qL}}_{\mathrm{in}}\left(i\right)+(1-q))}^{\frac{1}{\mathrm{\&beta;}}})}^2$ Repeat above-mentioned steps 4, obtain corresponding model parameter q and β, should be noted in the discussion above that measuring difference with the monocycle joint is that the span of parameter p and α is respectively: q 〉=1, β 〉=0.

Identical with monocycle joint measurement method of parameters, if want to obtain more precise parameters, can be after target function value F drops to below the thresholding T, still iteration is several times again, if target function value F continues to descend, or rises again after descending, or directly rises, no matter which kind of situation of change target function value F is, then select minimum value corresponding parameters wherein can improve the precision of parameter measurement to a certain extent as measurement result.

Identical with monocycle joint measurement method of parameters, determine and the measurement of parameter of types of models are carried out simultaneously, in the reality, the type of equivalent model is these two kinds of forms not only, can find only one in the method for all relevant equivalent models by measurement parameter by said method.

The bearing calibration of embodiment two, too many levels Gamma characteristic:

Method of measurement based on the gamma characteristic equivalent model parameter of embodiment one, can obtain the functional relation of gamma characteristic equivalent model, utilize the inverse function of this function to construct calibration model, just can proofread and correct, be elaborated with the example that has particular application as in the multimedia information system below monocycle joint or too many levels gamma characteristic.

As shown in Figure 9, the ordinary circumstance multimedia information system is that a plurality of Gamma links cascade up, and for different situations, such as situation A and situation B, the link of participating in cascade among the figure is ever-changing.In general, link 1 is video camera/camera, and link N _t(last) is display.In theory, between any two links, can insert a correction link (circuit is realized or software is realized) (be included in the foremost or insert correction link backmost), but actual conditions may be really not so.Deposit the Gamma link and show between the LUT Gamma link such as the display frame that mentions in front and can't insert correction link.Therefore, ordinary circumstance is possible have P correction link insertion point, also is called check point.The inventive method is inserted correction link as long as select one from an above P check point at this point, can realize that whole Gamma proofreaies and correct.

For situation A: one embodiment of the present of invention are to add correction link between video camera and display frame are deposited.

For situation B: one embodiment of the present of invention are:

1,, between depositing, video camera and display frame add correction link for the local terminal video;

2,, between depositing, decoder and display frame add correction link for the far-end video;

3, for encircling video certainly, between video camera and encoder, add correction link, perhaps between decoder and display frame are deposited, add correction link.

From this check point, the link number of front is N _aIndividual (link 1 to link N _a), the link number of back is N _pIndividual (link N _a+ 1 to N _t, the N of relation is arranged _a+ N _p=N _tSet up), following two kinds of special cases are arranged:

1), works as N _aOr N _pWhen equalling zero, in fact at this moment corresponding special case of inserting correction link up front or backmost regarded system as one and had the comprehensive Gamma characteristic of too many levels and proofread and correct;

2), work as N _aOr N _pEqual at 1 o'clock, video camera/camera or display are proofreaied and correct separately.

Make that link 1 is G to the comprehensive Gamma characteristic of the cascade of Na the link of link Na _a(.), link N _a+ 1 to link N _tN _pThe comprehensive Gamma characteristic of the cascade of individual link is G _p(.).Adopt this kind method, can obtain the model of correction link easily.And need explanation, this method is not limited to the model of each single link, and perhaps collective model adopts the first kind of the present invention and second class model.For other model, if can obtain G from mathematics _a(.) and G _p(.) its contrafunctional analytical form (closed form) so all is suitable for subring joint decomposition method of the present invention.And the present invention also is applicable to the model of some other form, such as the model that adopts the tables of data form, for G _a(.), G _p(.) itself do not have the situation of analytical form (realizing that such as adopting look-up method its inverse function has not just had analytical form yet certainly).Exist for the form of model itself with tables of data, its inversion model is exactly the contrary table of this tables of data so, and there are two row in a table, a lot of row, left column (input row) is the sampled value of input signal, be signal value to be corrected, right row (output row) are corresponding output signal values, the signal value after promptly proofreading and correct, line number depends on sampling number, line number is many more accurate more, contrary table be exactly about two row exchange the new data table that obtains, one embodiment of the present of invention are with the realization of tabling look-up.

Concrete bearing calibration comprises the steps:

1, sets up syndrome link G _Ac(.):

If G _a(.) belongs to comprehensive first kind equivalent model:

$L_{\mathrm{out}}=p_o{L_{\mathrm{in}}}^{{\mathrm{\&alpha;}}_n}+(1-p_a)---0<p_a\&le;1,{\mathrm{\&alpha;}}_a>0---\left(13\right)$

G so _Ac(.) model is:

$L_{\mathrm{out}}={(\frac{1}{p_a}{L}_{\mathrm{in}}+(1-\frac{1}{p_a}))}^{\frac{1}{{\mathrm{\&alpha;}}_a}}---p_a\&GreaterEqual;1,{\mathrm{\&alpha;}}_a>0---\left(14\right)$

If G _a(.) belongs to the comprehensive second class equivalent model:

$L_{\mathrm{out}}={(q_a{L}_{\mathrm{in}}+\left(1{-q}_a\right))}^{\frac{1}{{\mathrm{\&beta;}}_a}}---q_a\&GreaterEqual;1,{\mathrm{\&beta;}}_a>0---\left(15\right)$

G so _Ac(.) model is:

$L_{\mathrm{out}}=\frac{1}{q_a}{L_{\mathrm{in}}}^{{\mathrm{\&beta;}}_n}+(1-\frac{1}{q_a})---q_a\&GreaterEqual;1,{\mathrm{\&beta;}}_a>0---\left(16\right)$

2, set up syndrome link G _Pc(.);

If G _p(.) belongs to comprehensive first kind equivalent model:

$L_{\mathrm{out}}=p_p{L_{\mathrm{in}}}^{{\mathrm{\&alpha;}}_p}+(1-p_p)---0<p_p\&le;1,{\mathrm{\&alpha;}}_p>0---\left(17\right)$

G so _Pc(.) model is:

$L_{\mathrm{out}}={(\frac{1}{p_p}{L}_{\mathrm{in}}+(1-\frac{1}{p_p}))}^{\frac{1}{{\mathrm{\&alpha;}}_p}}---p_p\&GreaterEqual;1,{\mathrm{\&alpha;}}_p>0---\left(18\right)$

If G _p(.) belongs to the comprehensive second class equivalent model:

$L_{\mathrm{out}}={(q_p{L}_{\mathrm{in}}+(1-q_p))}^{\frac{1}{{\mathrm{\&beta;}}_p}}---q_p\&GreaterEqual;1,{\mathrm{\&beta;}}_p>0---\left(19\right)$

G so _Pc(.) model is:

$L_{\mathrm{out}}=\frac{1}{q_p}{L_{\mathrm{in}}}^{{\mathrm{\&beta;}}_p}+(1-\frac{1}{q_p})---q_p\&GreaterEqual;1{,\mathrm{\&beta;}}_p>0---\left(20\right)$

3, two syndrome links of cascade constitute correction link;

As shown in Figure 9, correction link G _c(.) is by subring joint G _Ac(.) and subring joint G _Pc(.) is according to G _Ac(.) preceding, G _Pc(.) after concatenated in order obtain.

Set up G _Ac(.) and G _PcBehind (.) model, its both cascade of specific implementation has following method:

1), direct computing method: the definition compound according to function, according to G _Ac(.) and G _PcThe calculation of parameter G of (.) _cThe parameter of (.).Because index α and β be the inverse of integer or integer not necessarily, the broad sense Newton binomial that the functional form after compound relates generally to the non-integer index launches, and contains infinitely multinomial.For convenience of calculation, some items before can only intercepting, calculation of complex, and cause the error of calculation.Calculate correction signal after the output in real time according to the luminance signal of this composite model and input then.

2), two-step method: at first calculate G _AcThe result of the correction of (.), promptly input luminance signal at first passes through G _Ac(.) once proofreaies and correct, and proofreaies and correct the result and is re-used as G _PcThe input of (.) is carried out secondary and is proofreaied and correct G _PcLuminance signal after the secondary of the output of (.) is proofreaied and correct is as last correction result.

3), look-up table: according to 1) or 2), for calculating abundant point on the input luminance signal interval, write down it and proofread and correct the result, table look-up as one.When proofreading and correct, the input signal values for needs are proofreaied and correct obtains to proofread and correct the result by tabling look-up then.The list item number is many more, i.e. sample collection is intensive more, and the effect of tabling look-up is accurate more.

For the big video data of data volume, the amount of calculation of calculating in real time is very big, and tabling look-up is most realistic method.The version of table generally comprises two row and multirows: signal value to be corrected is a left column, proofreaies and correct as a result that respective column is listed as on the right side, and line number depends on and what of sampled value.The method of tabling look-up is, according to signal value to be corrected, in the left column of table, search, if find, directly the right train value of correspondence as the result who tables look-up.If do not find, utilize linear interpolation to calculate, establishing signal value to be corrected is a, it is positioned at b, c, and (between two adjacent list items of left column of c＞a＞b), the right list items of b, c correspondence is d, e, so final checking result $f=\frac{c-a}{c-b}b+\frac{a-b}{c-b}c.$

The beneficial effect that technical solution of the present invention is brought is as follows:

The invention solves the measurement problem and the calibration problem of ubiquitous Gamma characteristic in the multimedia information system, provide a kind of general bearing calibration for the comprehensive Gamma characteristic of any too many levels cascade.Thereby can improve the user experience of multimedia information system greatly, attract the user, promote the market competitiveness of Related product, and promote novel telecommunication service such as video telephone, the quickening of video conference etc. is popularized, for telecom operators provide wide business opportunity.

Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from the spirit and scope of the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.

Claims (19)

1, a kind of gamma characteristic correcting method is used for proofreading and correct the N that multimedia information system comprises _tThe gamma characteristic of individual link is characterized in that, described method comprises the steps:

Determine check point, this check point is with described N _tIndividual link is divided into and is positioned at this check point N before _aIndividual link and be positioned at N after this check point _pIndividual link, wherein: N _a〉=0, N _p〉=0, N _a+ N _p=N _t

Determine equivalent described N _aFirst equivalent model of individual link gamma characteristic and its first inversion model thereof are determined equivalent described N _pSecond equivalent model of individual link gamma characteristic and second inversion model thereof;

According to described first inversion model and the described second inversion model structural correction link model, utilize this correction link model to determine described N _aThe correction signal of individual link final output signal is also imported described N with this correction signal _pIndividual link.

2, the method for claim 1 is characterized in that, when described equivalent model adopted the function representation form, the representative function of described inversion model was corresponding inverse function.

3, method as claimed in claim 2 is characterized in that, described definite first equivalent model or second equivalent model comprise the steps:

A1, detect N sampled value L respectively with input signal _In(i) N value L of the real output signal of the described link generation of input ^P _Out(i), wherein: 0≤i≤N-1;

A2, in the equivalent model of one group of alternative employing function representation form, select a model execution in step A3 to be measured;

A3, for described model to be measured, select one group of initial parameter;

A4, calculating:

N value L according to the theoretical output signal of described N sampled value correspondence of described Model Calculation to be measured ^M _Out(i), wherein: 0≤i≤N-1; And

Target function value F, described desired value functional value F with each to corresponding L ^P _Out(i) and L ^M _Out(i) difference is relevant;

A5, judge whether described target function value F is equal to or less than the threshold value of setting, if think that then described model to be measured is accepted as final equivalent model, and change steps A 8 over to after the parameter of target function value F corresponding parameters with minimum as this equivalence model; Otherwise execution in step A6;

Whether the execution number of times of A6, determining step A4 arrives the iterations of qualification, if then select one again as model to be measured and return steps A 3 from other alternative equivalent model that does not detect as yet; Otherwise execution in step A7;

A7, utilize the mathematical optimization method to adjust described model parameter, return steps A 4;

A8, end.

4, method as claimed in claim 3, it is characterized in that, in the described steps A 5, when described target function value F is equal to or less than the threshold value of setting, again according to the cycle-index of setting, utilize described mathematical optimization method to adjust described parameter and calculate described target function value F, target function value F corresponding parameters that will be wherein minimum is as the parameter of described equivalent model then.

5, as claim 3 or 4 described methods, it is characterized in that, also comprise the steps:

Respectively the parameter that measures is brought in the representative function of corresponding inversion model, ask for first inverse function formula of corresponding first inversion model and the second inverse function formula of corresponding second inversion model;

Utilize this first inverse function formula and the second inverse function formula structural correction model.

6, method as claimed in claim 5 is characterized in that, it is one of following that the building method of described correction link model comprises:

Direct computing method: utilize first inverse function and the second contrafunctional compound function to calculate N in real time _aThe correction signal of the final output signal of individual link;

Two step computing methods: utilize first inverse function to calculate N in real time _aA correction signal of the final output signal of individual link utilizes second inverse function to calculate the secondary correction signal of this correction signal, with this secondary correction signal as described correction signal;

Look-up table: according to described direct computing method or two step computing methods, calculate described N in advance _aThe corrected value of a plurality of sampled value correspondences in the interval of the final output signal of individual link, and corresponding relation is kept in the tables of data, determine the corrected value of any value to be corrected then by this tables of data of real-time inquiry.

7, method as claimed in claim 6 is characterized in that, in the described look-up table:

When the value of importing to be corrected is in tables of data, then directly obtain corresponding corrected value by tabling look-up;

When the value of importing to be corrected is not in the input of tables of data is listed as, take the average method of linear interpolation to ask for corresponding corrected value.

8, method as claimed in claim 3 is characterized in that, described target function value F satisfies following condition:

$F$ $=$ $\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-L_{\mathrm{out}}^M\left(i\right))}^2.$

9, method as claimed in claim 3 is characterized in that, the functional relation of the equivalent model of described one group of alternative employing function representation form and corresponding inverse function relational expression thereof comprise:

Described gamma characteristic equivalent model functional relation is: L _Out=pL _In ^α+ (1-p), wherein: the domain of definition of this function is interval [0,1], codomain is interval [(1-p), 1]; Then the inverse function relational expression of this function is:

$L_{\mathrm{out}}={(\frac{1}{p}{L}_{\mathrm{in}}+(1-\frac{1}{p}))}^{\frac{1}{\mathrm{\&alpha;}}};$ Perhaps,

Described gamma characteristic equivalent model functional relation is: $L_{\mathrm{out}}={({\mathrm{qL}}_{\mathrm{in}}+(1-q))}^{\frac{1}{\mathrm{\&beta;}}},$ Wherein: the domain of definition of this function is interval [1-1/q, 1], and codomain is interval [0,1]; Then the inverse function relational expression of this function is:

$L_{\mathrm{out}}=\frac{1}{q}{L_{\mathrm{in}}}^{\mathrm{\&beta;}}+(1-\frac{1}{q});$

Wherein: L _InBe input signal values, L _OutFor being respectively, output signal value, p and α and q and β need the parameter measured; And work as N _a=1 or N _p=1 o'clock: 0＜p≤1, α 〉=1, q 〉=1, β 〉=1; Work as N _a＞1 or N _p＞1 o'clock: 0＜p≤1, α 〉=0, q 〉=1, β 〉=0.

10, method as claimed in claim 3 is characterized in that, described mathematical optimization method is: climbing method or 0.618 method or steepest descent method or conjugate gradient method.

11, method as claimed in claim 3 is characterized in that, the N of a described input signal sampled value is selected in interval [0,1].

12, the method for claim 1 is characterized in that, when described equivalent model was the tables of data form, corresponding described inversion model was exactly the contrary table of this tables of data.

13, the detection method of a kind of definite gamma characteristic equivalent model and parameter thereof, described gamma characteristic equivalent model are used for the gamma characteristic of equivalent signal transmission or processing links, it is characterized in that, comprise the steps:

B1, detect N sampled value L respectively with input signal _In(i) N value L of the real output signal of the described link generation of input ^P _Out(i), wherein: 0≤i≤N-1;

B2, in the equivalent model of one group of alternative employing function representation form, select a model execution in step B3 to be measured;

B3, for described model to be measured, select one group of initial parameter;

B4, calculating:

N value: L according to the theoretical output signal of the described signal N of a described Model Calculation to be measured sampled value correspondence ^M _Out(i), wherein: 0≤i≤N-1; And

Target function value F, described desired value functional value F with each to corresponding L ^P _Out(i) and L ^M _Out(i) difference is relevant;

B5, judge whether described target function value F is equal to or less than the threshold value of setting,, and change step B8 over to after will organizing the parameter of parameter as this equivalence model if think that then described model to be measured is accepted as final equivalent model; Otherwise execution in step B6;

Whether the execution number of times of B6, determining step B4 arrives the iterations of qualification, if then select one again as model to be measured and return step B3 from other alternative equivalent model that does not detect as yet; Otherwise execution in step B7;

B7, utilize described mathematical optimization method to adjust described parameter, return step B4;

B8, end.

14, method as claimed in claim 13, it is characterized in that, among the described step B5, when described target function value F is equal to or less than the threshold value of setting, again according to the cycle-index of setting, utilize described mathematical optimization method to adjust described parameter and calculate described target function value F, target function value F corresponding parameters that will be wherein minimum is as the parameter of described equivalent model then.

As claim 13 or 14 described methods, it is characterized in that 15, described target function value F satisfies following condition:

$F$ $=$ $\underset{i=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{(L_{\mathrm{out}}^P\left(i\right)-L_{\mathrm{out}}^M\left(i\right))}^2.$

16, method as claimed in claim 15 is characterized in that, described gamma characteristic comprises monocycle joint gamma characteristic or the comprehensive gamma characteristic of too many levels cascade.

17, method as claimed in claim 15 is characterized in that, the functional relation of described one group of alternative equivalent model and corresponding inverse function thereof comprise:

Described gamma characteristic equivalent model functional relation is: L _Out=pL _In ^α+ (1-p), wherein: the domain of definition of this function is interval [0,1], codomain is interval [(1-p), 1]; Then the inverse function relational expression of this function is:

$L_{\mathrm{out}}={(\frac{1}{p}{L}_{\mathrm{in}}+(1-\frac{1}{p}))}^{\frac{1}{\mathrm{\&alpha;}}};$ Perhaps,

Described gamma characteristic equivalent model functional relation is: $L_{\mathrm{out}}={({\mathrm{qL}}_{\mathrm{in}}+(1-q))}^{\frac{1}{\mathrm{\&beta;}}},$ Wherein: the domain of definition of this function is interval [1-1/q, 1], and codomain is interval [0,1]; Then the inverse function relational expression of this function is:

$L_{\mathrm{out}}=\frac{1}{q}{L_{\mathrm{in}}}^{\mathrm{\&beta;}}+(1-\frac{1}{q});$

Wherein: L _InBe input signal values, L _OutFor being respectively, output signal value, p and α and q and β need the parameter measured; And, in the described monocycle joint gamma characteristic: 0＜p≤1, α 〉=1, q 〉=1, β 〉=1; In the described too many levels gamma characteristic: 0＜p≤1, α 〉=0, q 〉=1, β 〉=0.

18, method as claimed in claim 13 is characterized in that, described mathematical optimization method is: climbing method or 0.618 method or steepest descent method or conjugate gradient method.

19, method as claimed in claim 13 is characterized in that, the N of a described input signal sampled value is selected in interval [0,1].

Images