WO2007003086A1

WO2007003086A1 - Method for gamma character equivalent model and parameter determination and gamma correction

Info

Publication number: WO2007003086A1
Application number: PCT/CN2006/000710
Authority: WO
Inventors: Zhong Luo
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2005-06-30
Filing date: 2006-04-18
Publication date: 2007-01-11
Also published as: CN100527856C; CN1889693A

Abstract

The invention discloses a method for gamma correction,for correcting gamma character of Nt links. Said invention comprises the steps of: determining the insertion point for correction and dividing the Nt links into Na links which are located before the insertion point and Np links which are located after said point, wherein Na≥0, Np≥0 and Na +Np =Nt ; determining the gamma character equivalent model and the inverse gamma model of said Na links and the gamma character equivalent model and the inverse gamma model of said Np links; configuring the correction link model according to the inverse gamma character model of Na and the inverse gamma character model of Np, determining the correction signal output of the Na links by the correction link model and then inputting the correction signal into the Np links.The invention solves such problem as the inaccurate precise in link gamma character correction and is suitable for the general gamma character correction in single link and multiple links connection.

Description

Method for determining gamma characteristic equivalent model and parameters and gamma characteristic correction method

The invention relates to the field of multimedia information technology, in particular to a method for determining a gamma characteristic equivalent model and its parameters and a method for correcting gamma characteristics in a multimedia signal transmission and processing link. Background of the invention

At present, with the development of broadband networks, video communication has become more and more widely used, and video conferencing and videophone services are becoming one of the basic services in Next Generation Network (NGN). In video communication, the distortion of the luminance signal due to the gamma nonlinearity of each link is an important factor affecting the signal transmission quality and the end user experience.

In a multimedia information system terminal (hereinafter referred to as a terminal), an optical signal of an external scene enters a camera or a camera, and then undergoes analog to digital conversion (A/D conversion) into a digital image signal, and then the following entry and B appear. Kind of situation.

Case A mainly involves local terminal systems, such as PCs, personal digital assistants (PDAs) or mobile phones with camera functions. The information is mainly processed on the local terminal and does not involve the communication process. It is mainly the case where the video/still image captured by the local camera/camera or the graphics/animation generated by the computer (PC) is displayed on the local screen.

In case A, the digital image signal is sent directly to the display device for display, which in turn becomes an optical signal that is received and perceived by the human eye. Of course, there may be some processing in the middle, such as image enhancement for improving the image effect.

Case B mainly relates to a multi-terminal-server communication system, communication between a plurality of terminals, or communication between a terminal and a server. Such terminals are, for example, video conferencing, video telephony, high-end (2.5G/3G, 3G) mobile phones with multimedia information (MMS) or video communication functions, PDAs, and the like. In case B, the digital image signal is compressed by a compression encoder (Encoder), then transmitted to the other terminal through a network (wired, wireless, circuit-switched, packet-switched network, etc.), and compressed by a decoder (Decoder) at the other terminal ( Decompression) Decodes and restores the digital image signal, which is then displayed on the display device, and finally becomes an optical signal that is perceived by the human eye.

Figure 1 is a schematic diagram of the model of the link Gamma. Whether in case A or in case B, the image brightness signal (Luminance) has gone through multiple steps. Here, the luminance signal refers to a generalized luminance signal, and the optical signal, the electrical signal, and the digitized image luminance/gray signal all contain information of the luminance signal. Therefore, in a broad sense, the luminance signal passes through a plurality of links. By definition, the Gamma characteristic refers to the brightness of one or more links. The input-output relationship is not a linear relationship, but a nonlinear relationship. The curve in Figure 2 gives an example of a typical Gamma characteristic. The abscissa in Figure 2 is the normalized input luminance information, and the ordinate is the normalized output luminance information.

The influence of the link Gamma nonlinear distortion is shown in Fig. 3. In Fig. 3, the brightness of the upper row of gray squares is linearly increased from 0.1 to 1.0, and the brightness of the next row of gray squares is nonlinearly distorted by the link Gamma. Take the law of power function as an example.

In practice, Gamma nonlinearity is caused by different reasons. For cathode ray tube (CRT) display devices, the Gamma characteristic satisfies the formula (1) under ideal conditions:

Among them, L. _Ut is the output luminance signal, L _in is the input luminance signal, and equation (1) is a power function. It should be noted that the input and output luminance signals here are normalized in their respective coordinate spaces, that is, 0 ≤ L. _{ut ≤l, 0≤L} _in _≤l.

For other types of displays such as mobile phones and PDAs, such as liquid crystal displays, their Gamma characteristic functions may be different in form or different from the formula (1). Referring to FIG. 4, FIG. 4 is a schematic diagram of a gamma-characterized model of cascading. The total gamma characteristic is equal to the composition of the Gamma characteristic function of each link, and satisfies the formula (2):

G _CT Q

.

L _m , G _cr (L = " ₍ (?("- ² ) ( ...·.··„ , brain

Among them, ".," represents the compound operation of the function. The subscript CT indicates the meaning of cascaded total Gamma.

For case A, the multiple Gamma links involved are shown in Figure 5. The main ones are:

1. Camera/camera Gamma, denoted as G _Cam (-). General cameras have Gamma characteristics. In addition to the nonlinearity of imaging devices such as the CCD itself, the camera also introduces artificial nonlinearity. The purpose is to make the camera's Gamma characteristics just compensate for the Gamma characteristics of the display, so that the total input-output characteristics are Linear. If the display's ideal Gamma is L. _Ui = L _in ² ' ² , then the camera's ideal Gamma is L. _u corpse L _in ⁴⁵ .

Therefore, the Gamma characteristics of the camera are theoretically determined by the Gamma characteristics of the display. However, because the terminal system is increasingly complicated, there are a number of different links between the camera and the display, and the Gamma characteristics of each link are unknown, so that even if the Gamma characteristics of the camera and the display are able to compensate each other, because of the existence of the intermediate link, This compensation is usually invalid. And there are many types of displays, such as CRT, liquid crystal, and plasma. The Gamma characteristics are quite different, and the Gamma characteristics of ordinary cameras often deviate significantly from the ideal Gamma characteristics described above, so the camera/camera always has Gamma characteristics.

2. Store the file Gamma as G _Fil (.). The file may come from the camera. Because it has been processed, compressed and encoded, it also carries the Gamma feature.

3. The indicator frame is stored as Gamma, denoted as G _FBuf (.). Some monitors because of the display storage The color depth is not enough. For example, it can only support 4-bit, 8-bit or 16-bit color depth instead of the ideal 24-bit true color. It is equivalent to compressing the dynamic range of the input luminance signal during display, so the Gamma feature is also introduced. In addition, the Gamma feature is also introduced in the palette (Palette) color mapping technology or the Dither technology used in the non-true color mode.

4, display check table Gamma, expressed as ( _{3⁄4 υ τ} (.). Some display devices in order to compensate for the non-linearity of the display, artificially introduced Gamma, the Gamma appears as a look-up table (LUT), from the frame The brightness data read out is first converted by the LUT and then driven to the display.

5. Display Gamma, denoted as G _Disp (.). A typical display has a strong gamma nonlinearity.

For Case B, several examples are given in Figure 6, such as local video/image represented by L1, self-loop video/image represented by L2, and far-end video/image represented by L3. As shown in Figure 6, the multiple Gamma links involved in Case B are:

1. Camera/camera Gamma, denoted as G _Cam (.);

2. Store the file Gamma (not shown), denoted as G _Fil (.);

3. Display the frame to store Gamma, denoted as G _FBuf (.);

4. Display the checklist Gamma, denoted as G _L UT(');

5. Display Gamma, denoted as G _D i _Sp (.);

6. The encoder Gamma is denoted as G _Enc (.), which is mainly due to the discrete cosine transform (DCT transform) in compression and the Gamma characteristic caused by quantization;

7. The decoder gamma, denoted as G _Dee (.), is mainly due to the gamma characteristic caused by inverse DCT and inverse quantization in decompression.

For Case B, it is more serious that the local video/image, far-end video/image and self-loop video/image (for special purposes such as troubleshooting) pass through the Gamma link. In addition, regardless of case A or case B, there may be more Gamma links involved in the actual environment, and the situation is more complicated.

In the ideal case, there is a linear relationship between the luminance signal input to the camera and the final output luminance signal on the display, ie: L. _Ut = L _in , so that the scene displayed on the display is exactly the same as the original, and the user experience is also the best.

To obtain a linear relationship between input and output, Gamma Correction must be performed on a link with Gamma characteristics. As shown in Fig. 7, for the case of a link, the gamma characteristic is given, then another correction link can be used to cascade it, so that the total gamma characteristic after the cascading becomes a true linear relationship, thereby achieving Compensate for the non-linearity of the Gamma link. The model of the correction link is the inverse model of the Gamma characteristic equivalent model. If the equivalent model can be expressed by a function relation, the function relation of the inverse model is its inverse function, namely G _g (.) and G _c (. ) Mutual inverse function. In general, the inverse function of a function does not necessarily have a solution. Even if there is a solution, it is generally difficult to obtain it by calculation.

More cases in the actual application are shown in Figure 8. The correction link needs to be inserted between the two Gamma links. The situation of G _c (.) is more complicated, G _c (.) and G _a (.) or G _p (.) is no longer a simple inverse function relationship.

There are two main methods for implementing the correction link in the prior art.

Prior Art One: It relies entirely on the camera/camera or the Gamma characteristic of the display LUT to correct the Gamma characteristics of the display.

Assume that under ideal conditions: G _Cam (.) is L. L _in ⁰ - ⁴⁵ , G _LUT (.) is L. _Ut = L _in ^M5 and G _Disp (.) is L. _Ut = L _in ² ' ² . Then, G _Cara . G _Disp (.) makes L. _Ut =L _in . ' ^45x2 ' ² =L _in , that is, the input luminance signal forms a standard linear relationship with the output luminance signal; likewise, G _UT . (? _Disp (.) makes

L. _{_{^{^{ut = L in 45x2 '2 =}}}} L in, i.e., between the input luminance signal and the output luminance signal forming a linear relationship criteria.

However, the above techniques have the following disadvantages: It is often difficult to obtain an ideal state and cannot be guaranteed. The camera/camera, the LUT's Gamma is exactly matched to the monitor Gamma. And the display type is more than 4, and the gamma characteristic of the ordinary camera is usually not ideal; if G _CAM (.) and GL _UT (.) exist at the same time, ^ ^^ ^^ makes! ^^ !^ ⁴⁵ , that is, the compensation is excessive, but deviates from the linearity; the mathematical model of the simulated gamma characteristic in the prior art 1 is also inaccurate, and the model and parameters of the link Gamma characteristic cannot be accurately determined.

Prior Art 2: Gamma correction is performed by inserting a Gamma characteristic correction link between certain links, for example, after the camera link or before displaying the frame storage link. In addition, it is also possible to use a more accurate model in the Gamma characteristic equivalent model of the display, such as the model shown in equation (3): ,, (3)

81 < , < 1

Accordingly, the camera's Gamma characteristics are considered to exactly match the Gamma characteristics of the display, such as the model shown in equation (4):

However, in the prior art 2, the following disadvantages exist: The calibration model is single, and the situation of multiple links is very complicated, and the Gamma characteristic of multiple links cannot be accurately obtained, which inevitably leads to inaccurate correction, that is, the calibration result is still nonlinear. Therefore, there is still no problem of overcorrection or undercorrection; and this method can only be applied to some specific cases and cannot be applied to the correction of any multiple Gamma links, that is, the versatility is poor. Summary of the invention

In view of this, the present invention proposes a method for determining a gamma equivalent model and its parameters to solve the problem of gamma equivalent model selection and inaccurate model parameters in the prior art, and thus It provides the basis for analyzing and correcting gamma characteristics in signal transmission and processing.

Another object of the present invention is to provide a method of correcting gamma characteristics to improve the accuracy of gamma characteristic correction. A further object of the present invention is to make the method suitable for the correction of gamma characteristics of single-link and multi-link simultaneously to solve the problem that the existing correction method cannot be universal.

In accordance with the above objects, the present invention provides a method of determining a Gamma characteristic equivalent model and its parameters, the Gamma characteristic equivalent model being used for the Gamma characteristic in an equivalent link, the method comprising the steps of:

51. Input N sample values L _in (i) of the input signal into the link, and detect N actual output signal values L ^p generated by the link. _Ut (i), where N is an integer greater than or equal to 1, and i is an integer greater than or equal to 0 and less than or equal to N-1;

52. Select one of a set of alternative equivalent models as the equivalent model to be tested;

53. For the equivalent model to be tested, selecting a set of initial parameters and an objective function, and setting a threshold of the objective function value and a maximum number of iterations;

54. Calculate N theoretical output signal values L ^M corresponding to the N sample values according to the equivalent model to be tested. _Ut (i), and according to the calculation expression of the objective function and L ^p . _Ut (i) and L ^M . _Ut (i) calculate the objective function value;

55. Determine whether the target function value is equal to or smaller than the threshold, and if yes, adopt the equivalent model to be tested as a final equivalent model, and use a parameter corresponding to the smallest objective function value as the equivalent model The parameters, the result of the process; otherwise, step S6;

56. Determine whether the number of executions of the step S4 reaches the maximum number of iterations, and if so, select one of the other unrecognized alternative equivalent models as the equivalent model to be tested, and return to step S3; otherwise, perform step S7;

57. Adjust the parameters using a mathematical optimization method and return to step S4.

The N sample values in step S1 are evenly spaced or unevenly spaced. In the step S5, after the adopting the equivalent model to be tested is the final equivalent model, the method further includes: adjusting the parameter and calculating the target function value by using a mathematical optimization method within a preset number of cycles, Get the parameters corresponding to the smallest objective function value. The objective function is F (-(I.

The Gamma characteristic is a single-link Gamma characteristic or a multi-step cascaded integrated Gamma characteristic.

The set of alternative equivalent models includes: the functional relationship of the Gamma characteristic equivalent model is 4„, = ^„" + (1-, where the domain of the function is the interval [0, 1], The range is the interval [(Ι-ρ), Ι], and the inverse function of the function is ., =(丄4 + (1-丄)) ; and / or ρ Ρ

J_

The functional relationship of the Gamma characteristic equivalent model is L _out = (qL _in +(lg)Y , where the domain of the function is the interval [ll/q, l], and the range is the interval [0, 1]; The inverse function of the function is =丄^/+(1-丄); where L _in is the input signal value, L. _ut is the output signal value, pqq

And α and q and β are parameters respectively; and when N _a = l or N _p = 1, 0 < ρ ≤ 1, α ≥ 1, q > l, β >1; when N _a > l or Ν _ρ > At 1 o'clock, 0 < ρ ≤ 1, α > 0, q ≥ l, β ≥ 0.

The mathematical optimization method is a hill climbing method, a 0.618 method, a steepest descent method or a conjugate gradient method. The sample values of the input signal in step S1 are located in the interval [0, 1].

The present invention also provides a gamma characteristic correction method for correcting gamma characteristics of N _t links, wherein N _t is an integer greater than or equal to 1, the method comprising the following steps:

A. determining a correction link insertion point, dividing the N _t links into N _a links before the insertion point of the correction link and N _p links after the insertion point of the correction link, where N _a and N _p are An integer greater than or equal to 0, and N _a +N _p =N _t ;

B. Determine an equivalent model equivalent to the >^ link Gamma characteristics and correspond to it An inverse model, and determining an equivalent model equivalent to the N _p links Gamma characteristics and an inverse model corresponding thereto;

Inverse model correction part configured to model the inverse model of Gamma characteristic of the cascaded C. and the N _a segment of N _p the Gamma characteristics of the link, using the correction model determines a correction signal segment last output signal of the part of the N _a And inputting the correction signal into the _Np links.

The link the N _t Gamma characteristic multimedia information system links the N _t Gamma characteristics; the signal is a luminance signal.

It is determined in step A that the correction link insertion point is located between the imaging device and the display frame memory, or between the imaging device and the encoder or between the decoder and the display frame memory.

The equivalent model takes the form of a data table, and the inverse model adopts an inverse table form corresponding to the data table.

The equivalent model takes the form of a function, and the inverse model takes the form of an inverse function corresponding to the function of the equivalent model.

The steps of determining the equivalent model described in step B include:

B1. Input N sample values L _in (i) of the input signal into the link, and detect N actual output signal values L ^p generated by the link. _Ut (i), where N is an integer greater than or equal to 1, and i is an integer greater than or equal to 0 and less than or equal to N-1;

B2. Select one of a set of alternative equivalent models as the equivalent model to be tested;

B3. For the equivalent model to be tested, select a set of initial parameters and an objective function, and set a threshold of the objective function value and a maximum number of iterations;

B4. Calculate N theoretical output signal values L ^M corresponding to the N sample values according to the equivalent model to be tested. _Ut (i), and according to the calculation expression of the objective function and L ^p . _Ut (i) and L ^M . _Ut (i) calculate the objective function value;

B5. determining whether the value of the objective function is equal to or smaller than the threshold, and if so, Using the equivalent model to be tested as the final equivalent model, and taking the parameter corresponding to the minimum objective function value as the parameter of the equivalent model, the result is the flow; otherwise, step B6 is performed;

B6. It is determined whether the number of executions of step B4 reaches the maximum number of iterations, and if so, one of the other alternative model models that have not been detected is selected as the equivalent model to be tested, and returns to step B3; otherwise, step B7 is performed;

B7. Adjust the parameters using the mathematical optimization method and return to step B4.

The N sample values in step B1 are evenly spaced or unevenly spaced.

In the step B5, after the adopting the equivalent model to be tested is the final equivalent model, the method further includes: adjusting the parameter and calculating the target function value by using a mathematical optimization method within a preset number of cycles, Get the parameters corresponding to the smallest objective function value.

The step of determining the inverse model corresponding to the equivalent model in step B includes: bringing the parameters of the equivalent model into the corresponding inverse function to obtain a function of the inverse model.

In step C, a direct calculation method, a two-step calculation method or a table look-up method is used to construct a correction link model.

In step C, a look-up table method is used to construct a correction link model, and in the look-up table method: when the input value to be corrected is in the data table used in the look-up table method, the corresponding correction is directly obtained by looking up the table. Value; When the input value to be corrected is not in the data table, a linear interpolation method is used according to other values to be corrected in the data table to calculate a correction value corresponding to the input value to be corrected.

N-1 N-1

The objective function is F = ∑(L ^p _out ( - ( ) ² or ^∑l ( - (I. The set of alternative equivalent models includes: the functional relationship of the Gamma characteristic equivalent model is „ ₍ = Ρ " +(Ι- , where the domain of the function is the interval [ο, ι], the range is the interval [(Ι-ρ), Ι], and the inverse function of the function is 4^ = (丄丄) ^ ; and / or pp The functional relationship of the Gamma characteristic equivalent model is L _l = (qL _in + (l - q) , where the domain of the function is the interval [ll/q, l], and the range is the interval [0, 1]; Then the inverse function of the function is =丄丄); where 1^ is the input signal value, L. _ut is the output signal value, pqq

And α and q and β are parameters respectively; and when N _a = l or N _p = 1, 0 < _p ≤ l, ≥ 1, q > l, β >1; when N _a > l or Ν _ρ > 1 When 0 < ρ ≤ 1, α ≥ 0, q ≥ l, β > 0 „

The mathematical optimization method is a hill climbing method, a 0.618 method, a steepest descent method or a conjugate gradient method. The sample values of the input signal described in step B1 are located in the interval [0, 1].

The invention solves the problem that the selection problem of the gamma characteristic equivalent model which is ubiquitous in the multimedia information system and the parameter inaccurate detection thereof, and also solves the problem that the accuracy of the link gamma characteristic correction is poor, for the single link and the multi-link level The integrated Gamma feature provides a universal correction method that improves the quality of signal transmission and greatly enhances the user experience of multimedia information systems. Since the present invention improves the accuracy of the gamma characteristic correction so that the multimedia information displayed at the receiving end is consistent with the multimedia information input by the transmitting end, the present invention is also advantageous for the widespread use of multimedia information services such as videophones, video conferencing, and the like. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a general model of the link Gamma characteristics;

Figure 2 is a graph of a typical Gamma characteristic;

Figure 3 is a schematic diagram of luminance signal distortion caused by the link Gamma characteristic;

Figure 4 is a schematic diagram of a general model of a multi-step cascade of Gamma characteristics;

Figure 5 is a schematic diagram of a plurality of Gamma links in Case A;

Figure 6 is a schematic diagram of a plurality of Gamma links in case B;

Figure 7 is a schematic diagram of correcting the Gamma characteristics of a single Gamma link;

Figure 8 is a schematic diagram of correcting Gamma characteristics of a plurality of Gamma links; 9(a) is a schematic diagram showing the gamma characteristic of the first type of link in the embodiment of the present invention, and FIG. 9(b) is a schematic diagram showing the gamma characteristic of the second type of link in the embodiment of the present invention;

Figure 10 is the area on the coordinate plane [0, l] x [0, l];

Figure 11 is a schematic diagram showing the insertion point of the correction link in the second embodiment of the present invention; Figure 12 is a schematic diagram of the front-back correction using two sub-links in the second embodiment of the present invention. Mode for carrying out the invention

In order to make the objects, technical solutions and advantages of the present invention more comprehensible, the present invention will be further described in detail below.

Studies have shown that the more precise gamma characteristic equivalent model is a linear combination of a power function and a constant function, or a combination of a linear function and a power function. In different environments, the parameters in the actual measurement function relationship are measured. A specific functional relationship can be obtained more accurately.

First Embodiment: Method for determining gamma characteristic equivalent model and its parameters

In this embodiment, the following two types of single-link equivalent models are selected as a general model of the Gamma characteristic as an example for detailed description:

The first type of Gamma characteristic equivalent model:

L _o =pL _in ^a +(lp) 0<p≤l, a>l (5) where L _in is the input signal, L. _Ut is the output signal, and α and p are parameters. The domain of the function shown in equation (5) (the range of argument values) is the interval [0, 1], and the range of values (the range of values of the function values) is the interval [(Ι-ρ), Ι], and the curve characteristics are as follows. Figure 9 (a). It can be seen from equation (5) that for the first type of Gamma characteristic equivalent model, if p = l and α = 1, there is L. _Ut =L _in . Therefore, in the actual measurement, if the parameters of the first type of equivalent model are p=l and α=1, or ρ is sufficiently close to 1 and α is sufficiently close to 1, the gamma characteristic of the link can be considered to be neglected. Slightly, no correction is required. A typical example of the first type of Gamma feature is a CRT display. The second type of Gamma characteristic equivalent model:

i

L _mt -(qL _in +{lq)Y q≥l, ≥l (6) where L _in is the input signal, L. _Ut is the output signal, and β and q are parameters. The domain of the function shown by equation (6) is the interval [ll/q, l], the value range is the interval [0, 1], and the curve characteristics are shown in Figure 9(b). As can be seen from equation (6), for the second type of equivalent model, if q = l and β = 1, there is L. _Ut =L _in . Similarly, in the actual measurement, if q = 1 and β = 1, or q is sufficiently close to 1 and β is sufficiently close to 1, it can be considered that the gamma characteristic of the link can be ignored without correction. A typical example of the second type of gamma feature is a video camera.

In addition, if q = l / p and α = β, then the first and second equivalent models are inverse functions of each other, so they can compensate each other to obtain linear characteristics. That is: If the Gamma link has the first type of Gamma characteristics, then the correction link has the second type of Gamma characteristics; if the Gamma link has the second type of Gamma characteristics, then the correction link has the first type of Gamma characteristics.

After cascading multiple Gamma links, the integrated Gamma characteristics, from the mathematical one: no longer have a single-link first-class or second-class equivalent model. However, it is found in the research that the mathematical model of the multi-link integrated Gamma characteristic has the following characteristics:

1. The function image is limited to the region [0, 1] χ [0, 1] of the coordinate plane as shown in Fig. 10 (where "X" represents the Cartesian Product of the two sets, or is called the direct product. ) Inside;

2. The function is monotonically increasing;

3. The function curve may be convex (the curve is bent to the left, or outward from the origin of the coordinate) or the lower convex (the curve is bent to the lower right, or inward from the origin of the coordinate:);

4. The function curve intersects with the Lin axis or intersects the Lout axis;

5. The function curve passes through the (1,1) point. Therefore, the combination of the characteristics of 3, 4 forms four composite cases: Case 1, and! ^Axis intersects, is convex;

Situation 2, and! ^Axis intersects, lower convex;

Situation 3, and! ^. ^Axis intersects, is convex;

Situation 4, and! The ^^ axes intersect and are convex.

According to the above qualitative analysis, the integrated Gamma characteristic equivalent model of multiple Gamma link cascades can have the following two types, and the functional relations are respectively formula (7) and formula (8):

The first type of Gamma characteristic equivalent model (corresponding to cases 3 and 4):

L _0Ut =pL _tn ^a +(\-p) 0<p≤\,a>0 (7) The second type of Gamma characteristic equivalent model (corresponding to cases 1 and 2):

L _0Ut =( _qL _in + (lq)yq≥i, >0 (8) It should be pointed out that, from the perspective of function form, the first-class equivalent model and the second-class equivalent model of multi-link integrated Gamma characteristics are respectively It is the same as the first-class equivalent model and the second-class equivalent model of the single-link Gamma characteristic. However, compared with the single-link Gamma characteristic equivalent model, the multi-link first-class integrated Gamma characteristic equivalent model is based on qualitative The analysis result and the empirical value of the actual measurement, the range of the index α becomes α>0, and in the multi-link second-class integrated Gamma characteristic equivalent model, the range of the index β becomes β >0.

After selecting the Gamma characteristic equivalent model, it is necessary to determine the parameters in the specific application environment. The parameters are directly related to whether the functional relationship between the input and output signals in the final Gamma characteristic equivalent model is accurate. Among them, for the first type of Gamma characteristic equivalent model, the parameters p and oc need to be determined; for the second type of Gamma characteristic equivalent model, the parameters q and 0 need to be determined.

The method of determining the single link and Gamma characteristics equivalent model parameter comprises the following steps: Step 11, the input luminance signal L _in, selects N sample points on the interval [0,1]: L _in (0), L _in (l), L _in (2) to L _in (i) to L _in (N-2), L _in (Nl), , where N is an integer greater than or equal to 1. The N sampling points may be evenly spaced or may be non-uniform, preferably using N sampling points that are evenly spaced.

In step 12, the N sample values of the luminance signal are respectively input into the link, and the corresponding N actual output luminance signal values are measured: L ^p . _Ut (0), L ^p . _Ut (l), L ^p . _Ut (2)...! /. _Ut (i)...! . _Ut (N-2), L ^P out(Nl)„

Step 13. Select one type of equivalent model from the above two types of equivalent models as the equivalent model to be tested. Here, the first type equivalent model is selected as an example.

Step 14. For the first type of equivalent model, select a set of initial parameters, and construct a fitted objective function, set the threshold T of the objective function and the maximum number of iterations. '

The objective function and the actually detected output luminance signal are related to the difference between the theoretical output luminance signals determined by the Gamma characteristic equivalent model, and the smaller the difference is, the closer the effect of the model is to the actual situation.

There are many construction methods for the objective function. The most common ones are the following formula (9), formula (10), formula (11) or formula (12): (9) (10)

F _Tl {p,a) = ^N fU _ml {i)-p _in {i) ~{\-p)\ (11)

/'=0

W-l 丄

Ρ _τ , β) = ∑ l L ^p . _Ut ()- ( , ,() + ( ) I (1.2)

N-l

Wherein, formula (9) and formula (10) are equivalent to =∑(L ^p _out ( - _t ()) ² , and the formula ί+=0

(11) and formula (12) correspond to = 1^()-1^(/)|. Formulas (9) and (11) In the first type of equivalent model, equations (10) and (12) correspond to the second type of equivalent model. The following formula (9) formula (10) is taken as an example.

In step 15, the mathematical optimization method is used to find the most suitable parameter value. First, for the objective function = §(L ^p _out (0-L _/fl ( ^a -(l-^)) ² ) as shown in equation (9), some mathematical optimization techniques are used, such as hill climbing method, 0.618 method (Hua Luogeng) The preferred method), the steepest descent method or the conjugate gradient method are used to find the minimum value. This process is actually an iterative process. During this process, the parameters p and a are constantly adjusted, and the function value F is decreasing continuously. When the function value drops to less than After the threshold T is given, it is considered that the minimum point has been found. The corresponding parameters p and α at this time are taken as the true parameters of the application environment model. It should be noted that the values of the parameters ρ and α are: 0 <ρ≤1, α≥1. If for F _n ( _A « =∑(L ^p _oat (i) - _P L, ) - (1 - p)f after M iterations, 'the function cannot be lowered to the threshold Below T, the model is considered incorrect, and the second type 丄^ , the equivalent model should be selected, so repeat the above steps for =∑( _t i)-(q _in (i) ₊ (i- _q )yy

14 and step 15, the corresponding model parameters q and β are obtained. It should be noted that the values of the parameters ρ and α are: q≥l, β≥1.

In order to further improve the accuracy of the parameters, it is possible to iterate several times after the objective function value F falls below the threshold Τ. The objective function value F may be continuously decreasing, or may be rising or rising directly after falling. Regardless of the change of the objective function value F, it is preferable to select the parameter corresponding to the minimum value as a result, which will be to some extent. Further improve the accuracy of the parameters.

It can be seen that the determination of the model type and the determination of the parameters are performed simultaneously. In practical applications, the types of the equivalent models are not limited to the above two types. By the above method, the most suitable one can be found in all relevant equivalent models. The above method can also be used to determine the multi-link integrated Gamma characteristic equivalent model parameters, the specific steps are as follows:

Step 21, the input luminance signal L _in, selects N sample points on the interval _{[0,1]: L in (0), L} in (l), L in (2) ... L in (i). ..L _in (N-2), L _in (Nl), where N is an integer greater than or equal to 1. The N sampling points may be evenly spaced or may be non-uniform, preferably using N sampling points that are evenly spaced.

In step 22, the N sample values of the luminance signal are respectively input into the link, and the corresponding N actual output luminance signal values . ' L ^{p are} measured. _Ut (0), L ^p . _Ut (l), L ^p . _Ut (2)...L ^p . _Ut (i)...L ^p . _{ut (N-2), L} P 0Ut (Nl).

Step 23: Select one type of equivalent model from the above two types of equivalent models as the equivalent model to be tested, and take the first type equivalent model as an example.

Step 24: For the first type of equivalent model, select a set of initial parameters, and construct a fitted objective function, set the threshold T of the objective function and the maximum number of iterations.

The objective function and the actually detected output luminance signal are related to the difference between the theoretical output luminance signals determined by the Gamma characteristic equivalent model, and the smaller the difference is, the closer the effect of the model is to the actual situation. There are many construction methods for the objective function, such as formula (9), formula (10), formula (11), or formula (12). The following equation (9) and formula (10) are still used as examples.

In step 25, the mathematical optimization method is used to find the most suitable parameter value. First, for the first class model, the objective function F _Tl (p, a) = (L ^p _out (/) - _P L _in (0 - (1 - , using some mathematical optimization techniques, such as hill climbing, 0.618, speed) The descending method or the conjugate gradient method is used to obtain the minimum value. This process is actually an iterative process. During this process, the parameters ρ and α are continuously adjusted, and the function value F is decreasing continuously. When the function value falls below a given threshold After that, I think I have already found At the smallest point. Taking the corresponding parameters p and α at this time as the real parameters of the application environment model, it should be noted that the difference between the single-link measurement and the multi-link parameters ρ and α are: 0 < ρ≤ 1, 0. Similarly, if F _n Q^) = ∑(^, ₍ ( -^( ° ~ { -P)f after M iterations, can not make the function fall below the threshold T, then the model is considered wrong, should choose The second type of equivalent model, then repeat the above steps 24 and 25 for (^) = ∑(L ^p _out ( - ( L, ( + d - ^) ² ), and obtain the corresponding model parameters q and β, which should be noted. Yes, the difference from the single-link measurement is that the multi-link parameters ρ and _{α have} the following values: q≥l, β≥0.

In the same way as determining the single-link parameter, in order to further improve the accuracy of the parameter, after the objective function value F falls below the threshold ,, it is iterated several times, and the parameter corresponding to the minimum value is selected as the result, which will be To some extent, the accuracy of the parameters is further improved.

As with the method of determining the single-link parameters, the determination of the model type and the determination of the parameters are performed simultaneously. In practical applications, the types of the equivalent models are not limited to the above two types. By the above methods, the most relevant models can be found in all relevant models. The right one. Second Embodiment: A method of correcting the Gamma characteristic of any link.

Based on the method for determining the gamma characteristic equivalent model and its parameters according to the first embodiment, the functional relationship of the gamma characteristic equivalent model can be obtained, and the inverse function of the function is used to construct the correction model, which can be used for single or multiple links. The gamma characteristics are corrected. The following is a detailed description of the specific application in the multimedia information system.

Assume that there is N _t (N _t is an integer greater than or equal to 1) Gamma in the multimedia information system. Link. A second embodiment of the invention comprises the following steps:

Step 31: Determine a correction link insertion point, and divide the N _t links into N _a links located before the insertion point of the correction link and N _p links located after the insertion point of the correction link, where N _a > 0, N _p > 0, and N _a + N _p = N _t .

As shown in FIG. 11, the multimedia information system is usually cascaded by multiple Gamma links. For different situations, such as Case A and Case B, the participation in the cascade is ever-changing. In general, link 1 is the camera/camera, and link N _t (the last one) is the display. In theory, a correction link (circuit implementation or software implementation) can be inserted between any two links. Even the correction link can be inserted at the front or the end, but the actual situation may not be the case, for example, in the display frame to store the Gamma link and The correction link cannot be inserted between the LUT Gamma links. Therefore, there may usually be P correction link insertion points. In this embodiment, as long as one of the P correction point insertion points is selected, and the correction step is inserted at the point, all the gamma corrections can be realized.

For Case A, an example of the present invention is to add a correction link between the camera and the display frame.

For case B, an example of the invention is:

1) For the local video/image, add a correction link between the camera and the display frame memory;

2) For the far-end video/image, a correction link is added between the decoder and the display frame store;

3) For self-loop video/images, add a correction link between the camera and the encoder, or add a correction link between the decoder and the display frame.

Starting from the insertion point of this correction link, the number of previous links is N _a (link 1 to link N _a ), and the number of subsequent links is N _p (links N _a +1 to N _t ). There are two special cases: When 1^ or 1^ is equal to zero, it corresponds to inserting the correction link at the front or the end, which is equivalent to treating the system as a multi-link integrated gamma characteristic for correction; when N _a or N _{When p} is equal to 1, it is equivalent to separately correcting the camera/camera or display. The cascaded integrated Gamma characteristic of the Na links of the link 1 to the link Na is G _a (.), and the cascaded integrated Gamma characteristic of the N _p links of the link N _a +1 to the link N _t is G _p (.). With the method of the first embodiment, the model of the correction link can be obtained conveniently and accurately. It should be noted that the method of this embodiment is not limited to the first type and the second type equivalent model described in the first embodiment. For other equivalent models, the sub-block decomposition method of the present invention is applied if the mathematical form of G _a (.) and G _p (.) and the inverse form of the inverse function (closed form) are obtained.

In addition, the present invention may also adopt other forms of models, such as a model in the form of a data table, which is applicable to G _a (.), G _p (.) itself has no functional form (for example, implemented by a look-up table method, The inverse function obviously has no parsing form. For the model itself to exist in the form of a data table, then its inverse model is the inverse of the data table. For example, a table has two columns and multiple rows, the left column (input column) is the sampled value of the input signal, that is, the signal value to be corrected, and the right column (output column) is the corresponding output signal value, that is, the corrected signal. The value, the number of rows depends on the number of sampling points, and the more the number of rows, the more accurate the correction result. The inverse table is the new data table obtained by adjusting the left and right columns. In addition, sometimes in order to improve the accuracy of the table, you need to do some processing after swapping the left and right columns of the original table. For example, the left column of the original table is generally evenly spaced. For example, the interval between two adjacent entries is 0.01, but the right column is generally not evenly spaced, so after the swap, the left column of the inverse table is not evenly spaced. Checking the table may be inconvenient. Therefore, it is necessary to adjust more entries by inserting more points. On the one hand, the entries are made denser, and on the other hand, the left columns are evenly spaced, thereby improving the accuracy of the table.

Step 32: Determine, according to the method described in the first embodiment, _a first equivalent model G _a (.) of an equivalent pre-N _a link Gamma characteristic and a first inverse model GM corresponding thereto, and determine an equivalent N _p The second equivalent model G _p (.) of the Gamma characteristic and the second inverse model G _pc (.) corresponding _thereto obtain the syndrome section G _ac (.) and the syndrome section G _pc (.).

For the calibrator link G _ae (.), if G _a (.) belongs to the integrated first-class equivalent model: L _oul = Pa +(1-P _a ) 0< _Pa ≤ \,a _a >0 (13) Then the G _ac (.) model uses its inverse function:

1 1 1

L _oui = (—L _in + (1—— ) _Pa ≥ l, a _a >0 (14)

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

丄

w=(^A + (i- )^ .≥ι,Α>ο (15) Then the G _ac (.) model uses its inverse function:

+ --) 9 _α ≥ β _α >0 (16) For the calibrator link G _pc (.), if G _p (.) belongs to the integrated first-class equivalent model:

4," = p _p L +(lp _p ) 0< _Pp ≤\, a _p >0 (17) Then the G _pc (.) model uses its inverse function:

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

丄

L _out = (q _p L _in + (1 - q _p ) ^ >1, >5 _ρ >0 (19) Then the G _pc (.) model uses its inverse function: „,=丄+(1 -丄) q _p ≥\, _p >0 (20) q _P q _P

Step 33, cascading two syndrome sections to form a correction link, using the correction link to determine _a correction signal of the last output signal of the N _a links, and inputting the correction signal into the N _p links, thereby implementing N _t Gamma correction for each link.

As shown in Fig. 12, the correction link G _c (.) is obtained by sub-link G _ac (.) and sub-link G _pc (.) in the order of G _ac (.) first and G _pc (.).

After the G _ae (.) and G _pe (.) models are established, the cascading of the two is implemented by various methods. This embodiment provides the following methods as examples, but the present invention is not limited to these types. method. 1) Direct calculation method: According to the definition of function compound, the parameters of G _e (.) are calculated according to the parameters of G _ac (.) and G _pe (.). Since the indices α and β are not necessarily integers or reciprocals of integers, the functional form after compounding generally involves a generalized Newtonian binomial expansion of non-integer exponents.

(Generalized Netwon's binomial series expansion ), containing an infinite number of multiples, computationally complex, and causing computational errors. For the convenience of calculation, the first few items can be intercepted for calculation. The output corrected signal is then calculated in real time based on this composite model and the input luminance signal.

In short, the direct calculation method is to use the composite function G _e (.) of G _ae (.) and G _pc (.) to calculate the correction signal of the last output signal of the N _a links.

2) Two-step calculation method: Firstly, the result of the correction of G _ac (.) is calculated, that is, the input luminance signal is first corrected by G _ac (.), and the calibration result is secondarily corrected as the input of G _pc (.). The secondary corrected luminance signal of the output of G _pc (.) is used as the final correction result.

In other words, the two-step calculation method is to use G _ae (.) to calculate the primary correction signal of the last output signal of the Na links, and then use G _pe (.) to calculate the secondary correction signal of the primary correction signal, which will be used twice. A correction signal is used as the correction signal.

3) Look up the table method: According to 1) or 2), calculate enough points for the input brightness signal value interval, and record the correction result as a look-up table. Then, when the correction is made, the calibration result is obtained by looking up the table for the input signal value that needs to be corrected. The more the number of entries, the denser the sample collection, the more accurate the lookup effect.

For video data with a large amount of data, the calculation of real-time calculation according to the direct calculation method and the two-step calculation method is large, and the table lookup method is the most practical method. The structure of the table generally includes two columns and multiple rows, wherein the signal value to be corrected is the left column, and the correction result is correspondingly listed in the right column, and the number of rows depends on the sampled value. The method of looking up the table is to find in the left column of the table according to the signal value to be corrected, and if found, directly use the corresponding right column value as the result of the table lookup; if not, use the linear interpolation calculation to set the correction The signal value is a, which is located between b, c (c>a>b) and two adjacent left list items, and the right list items corresponding to b and c are respectively e, then finally The result of the lookup table is _f = _{e +} ^d.

C-b c-b

In short, the look-up table method is to calculate a correction value corresponding to a plurality of sample values in the value interval of the last output signal of the N _a link according to the direct calculation method or the two-step calculation method, and The correspondence is stored in a data table, and then the correction value of any value to be corrected is determined by querying the data table.

The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modifications, equivalents, improvements, etc., which are included in the spirit and scope of the present invention, should be included in the present invention. Within the scope of protection.

Claims

Claim

A method for determining a gamma Gamma characteristic equivalent model and a parameter thereof, wherein the Gamma characteristic equivalent model is used for a Gamma characteristic in an equivalent link, characterized in that the method comprises the following steps:

2. The method according to claim 1, wherein the N sample values in step S1 are evenly spaced or unevenly spaced.

The method according to claim 1, wherein in the step S5, after the adopting the equivalent model to be tested is the final equivalent model, the method further comprises: presetting Within the number of cycles, the mathematical optimization method is used to adjust the parameters and calculate the objective function value to obtain a parameter corresponding to the minimum objective function value.

4. The method of claim 1, wherein the objective function is

^ =∑ ( ( - ()) ² or ^ HIL ^p _out ( - (1.

The method according to claim 1, wherein the Gamma characteristic is a single-link Gamma characteristic or a multi-link cascade integrated Gamma characteristic.

6. The method of claim 1 wherein the set of alternative equivalent models comprises:

The functional relationship of the Gamma characteristic equivalent model is L _oul = pL _! ^a +(l~p) , where the domain of the function is the interval [0, 1] and the value range is the interval [(Ι-ρ), Ι], the inverse function of the function is Κ丄+ (1—丄) ; and/or

The functional relationship of the Gamma characteristic equivalent model of PP is L _out

Where the domain of the function is the interval [ll/q, l], and the range of values is the interval [0, 1]; then the inverse function of the function is U丄丄);

q q

Where L _in is the input signal value, L. _ut is the output signal values, p and q and [alpha] and β are parameters; and when N _a = l, or _{N p = l, 0 <ρ≤1} , α 1, q≥l, β≥1; if N _a When >l or Np>l, 0<p≤l, o^O, q≥l, β≥0.

The method according to claim 1 or 3, wherein the mathematical optimization method is a hill climbing method, a 0.618 method, a steepest descent method or a conjugate gradient method.

8. The method according to claim 1, wherein one of the sample values of the input signal in step S1 is located in the interval [0, 1].

9. A Gamma characteristic correction method for correcting the gamma characteristic of N _t links, Where N _t is an integer greater than or equal to 1, and the method includes the following steps:

A. determining a correction link insertion point, dividing the N _t links into N _a links located before the insertion point of the correction link and N _p links located after the insertion point of the correction link, where >^ and N _p are An integer greater than or equal to 0, and N _a +N _p =N _t ;

B. determining an equivalent model equivalent to the Gamma characteristic of the link and an inverse model corresponding thereto, and determining an equivalent model equivalent to the Gamma characteristic of the N _p links and an inverse model corresponding thereto;

10. The method as claimed in claim 9, characterized in that the number N _t of the Gamma characteristic aspects multimedia information system links the N _t Gamma characteristics; the signal is a luminance signal.

11. The method according to claim 10, wherein in step A it is determined that the correction link insertion point is between the imaging device and the display frame store, or between the camera device and the encoder or between the decoder and the display frame store. .

12. The method according to claim 9, wherein the equivalent model is in the form of a data table, and the inverse model adopts an inverse table form corresponding to the data table.

13. The method according to claim 9, wherein the equivalent model takes the form of a function, and the inverse model adopts an inverse function form corresponding to a function of the equivalent model.

14. The method according to claim 13, wherein the step of determining an equivalent model in step B comprises:

B1. Input N sample values L _in (i) of the input signal into the link, and detect N actual output signal values L ^p generated by the link. _Ut (i), where N is an integer greater than or equal to 1, i is an integer greater than or equal to 0 and less than or equal to N1;

B5. determining whether the target function value is equal to or smaller than the threshold, and if yes, adopting the equivalent model to be tested as a final equivalent model, and using a parameter corresponding to the smallest objective function value as the equivalent model The parameters, the result of this process; otherwise, step B6;

The method according to claim 14, wherein the N sample values in step B1 are evenly spaced or unevenly spaced.

The method according to claim 14, wherein in the step B5, after the using the equivalent model to be tested as the final equivalent model, the method further comprises: utilizing within a preset number of cycles The mathematical optimization method adjusts the parameter and calculates the objective function value to obtain a parameter corresponding to the minimum objective function value.

The method according to claim 14, wherein the step of determining the inverse model corresponding to the equivalent model in step B comprises:

The parameters of the equivalent model are brought into the corresponding inverse function to obtain the function of the inverse model.

18. The method according to claim 14 or 17, wherein in step C, a direct calculation method, a two-step calculation method or a table look-up method is used to construct the correction link model. 19. The method according to claim 14 or 17, wherein in step C, a look-up table method is used to construct a correction link model, and in the look-up table method:

When the input to be corrected value is in the data table used in the lookup table method, the corresponding correction value is directly obtained by looking up the table;

When the input value to be corrected is not in the data table, a linear interpolation method is used according to other values to be corrected in the data table to calculate a correction value corresponding to the input value to be corrected.

20. The method of claim 14, wherein the objective function is

N-1 N-1

F =∑ ( () - (0) ² or =∑ 1 L ^p _out (0 - (0 I.

21. The method of claim 14 wherein said set of alternative equivalent models comprises:

The functional relationship of the Gamma characteristic equivalent model is L _out = _P L _in ^a + (1 - , where the domain of the function is the interval [0, 1], and the range is the interval [(lp), l], The inverse function of the function is K丄^+(1—丄^ ; and/or

The functional relationship of the Gamma characteristic equivalent model of PP is L _out = (qL _in + , where the domain of the function is the interval [ll/q, l], and the range is the interval [0, 1]; then the function The inverse function relation is = +(1-丄);

q q

Where L _in is the input signal value, L. _Ut is the output signal value, p and a, and q and β are parameters respectively; and when N _a = l or N _p = l, 0 < _p ≤ l, o ^ l, q ≥ l, β >1; _{When a} >1 or ^>1, 0<p≤l, c^O, q≥l, β≥0.

The method according to claim 14 or 16, wherein the mathematical optimization method is a hill climbing method, a 0.618 method, a steepest descent method or a conjugate gradient method.

The method according to claim 14, wherein the input in step B1 is The N samples of the signal are located in the interval [0, 1].