CN102723054A - Online calibration system and online calibration method for ununiformity of LED (light-emitting diode) display screen - Google Patents

Abstract

The invention discloses an online calibration system and an online calibration method for ununiformity of an LED (light-emitting diode) display screen. The online calibration system and the online calibration method mainly solve problems that an existing test method is only used for testing brightness uniformity, tested brightness uniformity index cannot reflect an overall performance of an LED display screen, and ununiformity calibration cannot be carried out according to tested data. The online calibration method includes steps of acquiring test images displayed by the LED display screen by a camera, and dividing LED lamp points by a method based on background prediction after noise is eliminated; repositioning luminous zones of the lamp points and component pixels of the luminous zones; and then mixing the lamp points to form colors of the pixels, and fitting acquired chromaticity coordinates of the lamp points to form a Gaussian curve after gross errors are eliminated, and solving a mean value of the chromaticity coordinates; substituting the mean value of the chromaticity coordinates into a uniformity evaluation formula to evaluate the uniformity; and finally generating a calibration matrix and performing online real-time online ununiformity calibration point by point. The online calibration method for the ununiformity of the LED display screen is high in precision and efficiency and good in stability, is simple, convenient and feasible, and is capable of being used for testing a uniformity display performance and ununiformity calibration for the LED display screen.

Description

LED display heterogeneity on-line correction system and bearing calibration

Technical field

The present invention relates to field of measuring technique, relate to assessment and correction, before can be used for LED display and dispatching from the factory and the performance test and the Nonuniformity Correction of erecting stage the LED display show uniformity.

Background technology

LED has short, advantage such as display brightness is high, the linearity is good of response time; Therefore from being born, promptly be applied to the demonstration field, more be used to make large display screen such as outdoor large-size screen monitors, the live display screen of stadiums contest and indoor big screen display screen etc. in recent years.

Yet in development, production and the application of LED display; Reason owing to manufacturing process and device aging; Problems such as the difference of driving voltage and each LED lamp point degree of aging be inconsistent between inconsistent, the display module of Lamp cup curvature internal driving inconsistent, the display module encapsulation during inconsistent, the LED encapsulation of doping content cause the inhomogeneous of LED display colourity and brightness demonstration easily when existing led chip to make.This LED display shows that heterogeneity mainly is presented as color lump pockety and " mosaic " phenomenon, cause the distortion of displaying contents, and the human visual system shows that than other flaw is more responsive for this type display distortion.Therefore in actual displayed was used, the brightness uniformity of LED display and uniformity of chromaticity became LED screen display quality and the most key factor of user's viewing experience of influencing.So be necessary in LED display dispatches from the factory and uses, to carry out the assessment of colourity and brightness uniformity, promptly should carry out corresponding homogeneity correction when heterogeneity acquires a certain degree.

The general display performance that uses index evaluation LED displays such as showing high-high brightness, primary colours predominant wavelength error, high-contrast, dynamic range, lamp dot center spacing and display resolution.Existing standard criterion file about the test of LED display index parameter; Comprise " LED display general specification " and " LED display method of testing ", stipulated about using nitometer, colorimeter to measure test and acceptance index such as the white chromaticity coordinates of LED display, primary colours predominant wavelength error, pixel light intensity homogeneity, display module brightness uniformity and module brightness uniformity.Wherein, white the look that white chromaticity coordinates and primary colours predominant wavelength error have only described that the LED display overall situation shows and the order of accuarcy of three primary colours are for uniformity of chromaticity standard accordingly in addition not; Some lamp points, display module, the module brightness that pixel light intensity homogeneity, display module homogeneity and module brightness uniformity are employed in space distribution rule is evaluated at the brightness uniformity of describing whole screen on the display unit basis of three kinds of yardstick scales respectively.Aspect LED display manufacturer, the more subjective judgements that rely on the naked eyes range estimation of actual LED display uniformity test perhaps only adopt nitometer, colorimeter to gather on the screen the overall uniformity index that or 16 dot informations are judged LED display at 9 more.

These homogeneitys judge that common problem is to attempt the information with local; It is the homogeneity of random test minority lamp point or the region description overall situation; The data of test can not act on the homogeneity correction in later stage; And measuring process needs manual work to choose several regions to measure, can not comprehensively and objectively describe the LED display show uniformity.

Chinese Academy of Sciences's Changchun ray machine Song Xinli etc. delivered LED display brightness uniformity appraisal procedure based on grey level histogram in 2008; Choose the captured image pixel of the camera center most bright value of forming LED lamp point and carry out computing; Obtain the Gauss curve fitting curve based on grey level histogram, its variance is as the index of assessment LED display brightness uniformity.This method fast and stable; But owing to only choose the center most bright value participation computing that LED lamp point is formed pixel; The demonstration situation that the index that obtains can not reflect LED lamp point truly of testing; Test result can not be participated in the pointwise correction in later stage, and the influence of the easy camera subject acquisition parameters setting of test result, such as exposure, light sensitivity.The same year; The Changchun ray machine people such as Wang Yuqing delivered a kind of method for objectively evaluating of LED display uniformity of chromaticity; Utilize the homogeneity of structural similarity size assessment LED display between image block; This evaluation method realizes simple, but index and national standard definition are incompatible, and there is the shortcoming that can not act on the later stage correction too in measurement result.

In sum; The existing inhomogeneity method of LED display of measuring respectively has advantage; But mostly all only lay particular emphasis on the test of brightness uniformity, and test data only derives from the part of LED display, cause method of testing can't characterize inhomogeneous " color lump " that causes of LED display colourity; The index that test obtains can not reflect the homogeneity of the LED display overall situation, more can not act on the Nonuniformity Correction in later stage.

Summary of the invention

The objective of the invention is to propose a kind of LED display heterogeneity on-line correction system and bearing calibration; Can not test uniformity of chromaticity to solve existing homogeneity testing method; Measured uniformity index can not reflect LED display overall situation performance, and test index and data can not be carried out the problem of later stage Nonuniformity Correction.

Be to realize above-mentioned purpose, on-line correction of the present invention system comprises playing controller, equipment LED display to be assessed and camera, it is characterized in that: also comprise assessment unit and correcting unit,

Described assessment unit comprises:

The assessment data transport module is used to receive the LED display displaying contents that camera is taken;

The denoising module is used to read the data of assessment data transport module and carries out denoising;

The lamp point is cut apart module, and the image of the method for utilizing background forecast after to denoising carries out cutting apart of lamp point and background area;

The colour mixture module, utilize the colour mixture principle lamp point of cutting apart the back image is formed that pixel is carried out the colour mixture processing and with RGB RGB system coordinate conversion to L ^*u ^*v ^*The colour system coordinate obtains LED display displaying contents L ^*u ^*v ^*Lightness index L under the colour system coordinate ^*, horizontal ordinate chromaticity index U ^*With ordinate chromaticity index V ^*, i.e. lightness dimension L ^*With chromaticity coordinate U ^*, V ^*

The gaussian curve approximation module is to lightness dimension L ^*With chromaticity coordinate U ^*, V ^*Carry out gaussian curve approximation respectively; The lightness dimension that obtains according to curve fitting and the average of chromaticity coordinate; Calculate the index value of brightness uniformity and uniformity of chromaticity; If the uniformity of chromaticity index value surpass 0.03 or the brightness uniformity index value surpass 5%, judge that then this LED display show uniformity is not up to standard, and this information not up to standard passed to the correction matrix generation module;

The correction matrix generation module, the colour system coordinate average and the lamp point data that pass over according to the gaussian curve approximation module generate the correction matrix under the RGB system space RGB, and it is transferred to correcting unit;

Described correcting unit comprises:

The correction data transport module is used to receive from the image to be displayed of playing controller and the correction matrix of assessment unit, and the image to be displayed after will proofreading and correct is sent to LED display and shows;

Correction module is used for correction matrix that the correction matrix generation module is generated and the image to be displayed of importing from playing controller and multiplies each other by point, realizes the heterogeneity pointwise correction of real-time online.

Described assessment data transport module has two external interfaces at least, adopts the CameraLink interface on USB interface or network interface or the special image capture card to realize; Described correction data transport module has three external interfaces at least, adopts USB interface or network interface to realize.

Described correction data transport module is connected with playing controller, LED display and assessment data transport module respectively, and correction module links to each other with the correction data transport module, constitutes corrective system; Timing, correction module are according to from the correction matrix of assessment data transport module, and image to be displayed is proofreaied and correct on pointwise ground, and the data after will proofreading and correct send to LED display through the correction data transport module and show; During the homogeneity assessment, correction module is not proofreaied and correct data to be displayed, directly will output to LED display from the image to be displayed of playing controller and show.

Described camera adopts the focal plane array image-forming equipment after parameter calibration and anti-Gamma correction; Its focal plane array is classified CCD device or cmos device as, and the taken image of this camera is transferred to the assessment data transport module through hardware interface such as CameraLink or USB or network interface.

For realizing above-mentioned purpose, on-line correction method of the present invention comprises the steps:

1) the camera focal plane arrays (FPA) is carried out parameter calibration and anti-Gamma correction, i.e. γ ^-1Proofread and correct;

2) select a width of cloth test pattern f;

Described test pattern f, different according to homogeneity assessment and Nonuniformity Correction purpose have following three kinds of selection modes:

If the user need carry out Nonuniformity Correction, the white image of selecting to have high-high brightness is as test pattern f, and the gray-scale value of the red channel of this test pattern f, green channel, blue channel is 255 at this moment;

If the brightness uniformity of user's needs assessment LED display when showing the certain luminance image, select a width of cloth RGB triple channel gray-scale value to coexist the gray level image of 0 to 255 a certain value as test pattern f;

The uniformity of chromaticity when if user's needs assessment LED display shows the particular color image, the image of then selecting to have this color is as test pattern f;

3) calculate effective shooting distance:

3a) according to LED display width W and height H, confirm effective shortest photographing distance L _MinFor:

L _min＝argmax{W/tan15°,H/tan15°}，

Wherein, to represent to ask for the maximal value in the argument list be dependent variable output to argmax{};

3b) according to pixel line number M and columns N, the wide w of breadth and the high h of camera focal plane arrays (FPA), LED display lamp point line number K and columns L, LED display width W and height H, and camera lens focal length d are confirmed effectively the longest shooting distance L _MaxFor:

L _max＝argmin{4WLd/Nw,4HKd/Mh}，

Wherein, to represent to ask for the minimum value in the argument list be dependent variable output to argmin{};

The shortest photographing distance L that basis is tried to achieve when 3c) taking _MinThe longest shooting distance L _Max, the distance between assurance camera and the LED display is at L _MinWith L _MaxBetween;

4) the preheating LED display sets up camera, sets the light sensitivity and the camera lens focal length d of camera shutter speed, aperture F number size, focal plane arrays (FPA), makes the shooting distance scope [L in the step 3) _Min, L _Max] fall in the viewing areas of LED display instructions regulation, guarantee that the camera lens optical axis direction is consistent with the optical axis direction or the LED display surface normal direction of LED display lamp point;

5) the control LED display shows test pattern f, and utilizes camera that test pattern f is taken L time, obtains image sets f _l, l ∈ 0,1,2 ..., L}, L>=2;

6) with image sets f _lIn each width of cloth image addition make even all, remove the dark noise and LED display demonstration jittering noise of focal plane arrays (FPA), obtain the denoising image

7) utilize the method for background forecast, the lamp point zone and the background area of denoising image

cut apart:

7a) denoising image

is carried out the cap conversion of white top; To suppress the background area; Strengthen LED lamp point zone; Acquisition lamp point enhancing image D (x, y):

$D(x,y)=\stackrel{\‾}{f}(x,y)-(\stackrel{\‾}{f}\⊗B_{\mathrm{oi}})(x,y),$

Wherein: (x, y) location of pixels in the presentation video;

B _OiBe by Δ B and B _bThe mathematical morphology structural element of regulation, Δ B is the toroidal structural element, B _bBe the disc-shaped structure element;

is a kind of morphology operations of the present invention's definition;

expression is asked for the image that only comprises background from denoising image

, and

is the morphology dilation operation, represent to adopt structural element Δ B that image

is carried out expansive working;

Be the morphological erosion computing,

Structural element B is adopted in expression _bTo image

Carry out erosion operation;

7b) adopt big Tianjin method that lamp point is strengthened image D (x; Y) carry out binary conversion treatment; The two-value template that obtains; Two-value template and denoising image

are multiplied each other by point, only comprised the lamp dot image of LED lamp point content

7c) utilize gray scale paddy peak value algorithm to determine the light-emitting zone and the composition pixel thereof of each LED lamp point in the lamp dot image

;

Described gray scale paddy peak value algorithm; Comprise horizontal convolution projection and vertical convolution project, for being defined as of lamp dot image

its horizontal convolution project of M * N size:

$H_{\left[x\right]}=\frac{1}{N}\underset{x=1}{\overset{N}{\mathrm{\Σ}}}\stackrel{\‾}{F}(x,y),$

Being defined as of vertical convolution project:

$V_{\left[y\right]}=\frac{1}{M}\underset{y=1}{\overset{M}{\mathrm{\Σ}}}\stackrel{\‾}{F}(x,y),$

Utilize horizontal convolution projection and vertical convolution projection process lamp dot image

Obtain the lamp dot image Horizontal convolution projection H _[x]With vertical convolution projection V _[y]The result; Row and column split image with place, valley center in the projection is the rectangular area; With the center of LED lamp point in the location positioning rectangular area of place, peak value center row and column intersection in the projection, utilize the connected domain method to determine the light-emitting zone and the composition pixel of each LED lamp point;

8) theoretical based on blend of colors, in the mixed lamp dot image of rgb space

Middle lamp point light-emitting zone is formed color of pixel, and each grey scale pixel value that is about in the LED lamp point light-emitting zone is made even all by the addition of RGB triple channel, obtains lamp point RGB gray matrix [R, G, B], and it is converted to L ^*u ^*v ^*Color space obtains the L of all lamp points ^*u ^*v ^*Volume coordinate value matrix [L ^*, U ^*, V ^*], lightness dimension value matrix L wherein ^*With the chromaticity coordinate value matrix U ^*, V ^*Dimension is K * L;

9) the coordinate figure matrix fitting is become Gaussian curve, and carries out the uniformity index assessment:

9a) get the coordinate figure matrix L earlier ^*, U ^*, V ^*Carry out the frequency distribution measuring respectively, distinguish the fitted gaussian curve again after the gross error value during the rejecting frequency distributes, obtain L ^*u ^*v ^*The lightness dimension mean data that lamp point shows under the colour system space

With the chromaticity coordinate mean data

9b) according to L ^*u ^*v ^*Colour system coordinate mean data is calculated uniformity index:

With the L that is obtained ^*u ^*v ^*Colour system coordinate average

Substitution brightness uniformity assessment formula and uniformity of chromaticity assessment formula can obtain the brightness uniformity index of LED display

With the uniformity of chromaticity index

${\mathrm{\ΔME}}_l^{*}=\frac{1}{K\×L}\underset{i=1}{\overset{K}{\mathrm{\Σ}}}\underset{j=1}{\overset{L}{\mathrm{\Σ}}}\sqrt{{(L^{*}(i,j)-\stackrel{\‾}{l^{*}}/\stackrel{\‾}{l^{*}})}^2}\×100\%$

Wherein, (i, j) (i, j) individual element, i.e. (i, j) the individual lamp point in the denotation coordination value matrix;

9c) according to the brightness uniformity index

and chroma uniformity index

evaluate the uniformity of the LED display shows:

When the brightness uniformity index value of being tried to achieve and uniformity of chromaticity index value satisfy

respectively less than 5%,

less than 0.03 the time, judge that then the show uniformity of this LED display is up to standard;

When brightness uniformity index value

surpasses 5%; Or uniformity of chromaticity index value

was above 5% o'clock; The show uniformity of then judging this LED display is not up to standard, carries out next step;

10) according to L ^*u ^*v ^*Colour system coordinate mean data and lamp point RGB gray matrix generate correction matrix: with the coordinate mean data

Be converted to rgb space, and red channel average, green channel average and the blue channel average of rgb space is extended for the equal value matrix of congruence that K * L ties up respectively, obtain the equal value matrix μ of red channel _r, green channel equal value matrix μ _gEqual value matrix μ with blue channel _b, and, obtain the lamp point red channel correction matrix R under the rgb space through following formula _r, green channel correction matrix R _gWith blue channel correction matrix R _b:

R _r=μ _r./R

R _g=μ _g./G

R _b=μ _b./B，

Wherein " ./" be a some divided-by symbol, R _r, R _g, R _bBe the matrix of K * L dimension;

11) correction matrix that utilizes step 10) to obtain carries out pointwise correction to data to be displayed:

With above-mentioned correction matrix R _r, R _g, R _bRespectively with the red channel gray level image F of image to be displayed _r, green channel gray level image F _gWith blue channel gray level image F _bCarry out pointwise correction by following formula and handle, obtain proofreading and correct the red channel image G of back image to be displayed G _r, green channel image G _gWith blue channel image G _b:

G _r=R _r.×F _r

G _g=R _g.×F _g

G _b=R _b×.，

Wherein, ". * " be the dot product symbol, two matrixes that expression will be participated in computing multiply each other by element;

12) the red channel image G of the image to be displayed G after will proofreading and correct _r, green channel image G _gWith blue channel image G _bOutput to LED display and show, accomplish LED display heterogeneity pointwise correction.

The present invention compared with prior art has following advantage:

1. compare with the method that tradition is tested the LED display brightness uniformity; The present invention handles the uniformity of chromaticity and the brightness uniformity index that can accurately record LED display simultaneously through colour mixture and colour system space conversion; The gained index can not only accurately be estimated the non-homogeneous luminance distortion that causes of LED display display brightness, also can accurately estimate the order of severity of inhomogeneous " color lump " phenomenon that causes of LED display colourity;

2. compare with the homogeneity testing method that tradition uses nitometer or light power meter to gather its overall performance of LED display local luminance information evaluation, the present invention is owing to adopt through γ ^-1The camera of proofreading and correct is taken the LED display content displayed, with brightness of entire image data computation and uniformity of chromaticity index, makes test index can more truly, more fully characterize the overall brightness and the uniformity of chromaticity display performance of LED display;

3. the present invention is because through adding correcting unit between the playing control unit of former display system and LED display; Other parts are carried out under the technical modification need not; Realize the heterogeneity real-time online pointwise correction of third party, effectively saved the correction cost LED display.

Description of drawings

Fig. 1 is an on-line correction system architecture diagram of the present invention;

Fig. 2 is the morphological structure element distribution plan that the present invention is adopted in the lamp point is cut apart;

Fig. 3 is an on-line correction method flow diagram of the present invention.

Embodiment

Be the display quality that guarantees LED display, the sensory experience that improves the beholder; The present invention concentrates on test and the assessment to LED display display performance index brightness uniformity and uniformity of chromaticity index, uniformity index display screen not up to standard is carried out the heterogeneity real-time online proofread and correct.For narrating conveniently, below will combine accompanying drawing, the composition of LED display heterogeneity on-line correction of the present invention system and the detailed step of on-line correction method are intactly described.

With reference to Fig. 1, LED display heterogeneity on-line correction of the present invention system comprises playing controller 1, equipment LED display to be assessed 2, camera 3, assessment unit 4 and correcting unit 5, wherein:

Said camera 3; The focal plane array image-forming equipment of employing after parameter calibration and anti-Gamma correction; Its focal plane array is classified CCD device or cmos device as; Be used to take the test pattern that LED display shows, hardware interface such as CameraLink or USB or the network interface of the data of collection through camera is transferred to assessment unit 4;

Described assessment unit 4; Be used for calculating, assessing brightness uniformity and uniformity of chromaticity index; And the generation correction matrix, it is cut apart module 43, colour mixture module 44, gaussian curve approximation module 45 and correction matrix generation module 46 by assessment data transport module 41, denoising module 42, lamp point and connects and composes successively;

This assessment data transport module 41; Have and be no less than two external data interfaces; These interfaces adopt the CameraLink interface on USB interface or network interface or the special image capture card to realize; One of them interface receives LED display 2 content displayed that camera 3 is taken; The image of being gathered is sent to denoising module 42 and carries out denoising, and second interface is transferred to the Nonuniformity Correction that correcting unit 5 carries out image to be displayed with the correction matrix that correction matrix generation module 46 generates, and the residue interface is used for the expedited data transfer rate and control signal is provided;

This denoising module 42; Link to each other with assessment data transport module 41; Read the image data set that assessment data transport module 41 receives, all images is carried out addition make even all, eliminate the influence that LED display shows jittering noise and camera 3 focal plane arrays (FPA) dark noises;

This lamp point is cut apart module 43, links to each other with denoising module 42, is used to receive the denoising image; Go forward side by side portable lighter point and background segment; When cutting apart, at first suppress the background in the denoising image and strengthen lamp point zone, this lamp point of binaryzation strengthens image then; The lamp dot image that is only comprised lamp point light-emitting zone is located the light-emitting zone of each LED lamp point at last and is formed pixel;

This colour mixture module 44 is cut apart module 43 with the lamp point and is linked to each other, be used for that the lamp point light-emitting zone of lamp dot image is formed pixel and carry out colour mixture and handle, and with the lamp point RGB RGB system coordinate conversion after the colour mixture to L ^*u ^*v ^*The colour system coordinate obtains LED display displaying contents L ^*u ^*v ^*Lightness index L under the colour system coordinate ^*, horizontal ordinate chromaticity index U ^*With ordinate chromaticity index V ^*, i.e. lightness dimension L ^*With chromaticity coordinate U ^*, V ^*

This gaussian curve approximation module 45 links to each other with colour mixture module 44, is used to reject lightness dimension L ^*With chromaticity coordinate U ^*, V ^*In gross error; The frequency of calculating these three coordinate figures distributes, and then fits to respectively with them and to meet the Gaussian curve that lamp point shows the heterogeneity distribution, and the lightness dimension that obtains according to curve fitting and the average of chromaticity coordinate; Calculate the index value of brightness uniformity and uniformity of chromaticity; If the uniformity of chromaticity index value surpasses 0.03, judge that then the uniformity of chromaticity index of this LED display 2 is not up to standard, if the brightness uniformity index value surpasses 5%; Judge that then these LED display 2 brightness uniformity indexs are not up to standard, and this information not up to standard is passed to correction matrix generation module 46;

This correction matrix generation module 46; Link to each other with gaussian curve approximation module 45; The colour system coordinate average and the lamp point data that are used for passing over according to gaussian curve approximation module 45 generate the correction matrix under the RGB system space RGB, and it is transferred to correcting unit 5;

Described correcting unit 5; Interconnected by correction data transport module 51 and correction module 52 and to form, this correction data transport module 51 has and is no less than three external data interfaces; These interfaces adopt USB interface or network interface to realize; One of them interface is connected with playing controller 1, receives the image to be displayed from playing controller 1, and second interface is connected with assessment data transport module 41; Reception is from the correction matrix of assessment unit 4, and the image to be displayed after the 3rd interface will be proofreaied and correct is sent to LED display 2 and shows; This correction module 52 links to each other with correction data transport module 51, is used for correction matrix that correction matrix generation module 46 is generated and the image to be displayed of importing from playing controller 1 and multiplies each other by point, realizes the Nonuniformity Correction of real-time online.

During test, preheating LED display at first, and configure camera shutter speed, aperture F number, focal plane arrays (FPA) light sensitivity and camera lens focal length; Again camera frame is located in the viewing areas of LED display instructions regulation; Guarantee that the camera lens optical axis direction is consistent with the optical axis direction or the LED display surface normal direction of LED display lamp point in the test process; Regulate distance between camera and the LED display 2 through setting different camera lens focal length d, the pixel quantity that makes the projection on camera 3 focal plane arrays (FPA)s of LED lamp point more than 4 * 4 and LED lamp point optical field distribution and size shape do not change; After treating that the LED display demonstration is stable, utilize playing controller 1 control LED display 2 to show test patterns, each width of cloth test pattern that utilizes 3 pairs of LED displays of camera to show repeats shooting apace; Resulting image is transferred to assessment unit 4, and suppresses noise by denoising module 42; The denoising image is cut apart module 43 through the lamp point and is cut apart and locate LED lamp point; Then through colour mixture module 44 in the RGB colorimetric system space reciprocally incorporated lamp point color, and after changing colour mixture the colour system coordinate to L ^*u ^*v ^*The space; Utilize the colour system coordinate of gaussian curve approximation module 45 match gained to be Gaussian curve this moment again; And according to the colour system coordinate mean value computation uniformity of chromaticity that obtains and the numerical value of brightness uniformity index; If the uniformity of chromaticity index value surpasses 0.03; The uniformity of chromaticity index of then judging this LED display 2 is not up to standard; If the brightness uniformity index value surpasses 5%, judge that then these LED display 2 brightness uniformity indexs are not up to standard, and this information not up to standard is passed to the required correction matrix of correction matrix generation module 46 generations; Resulting correction matrix is transmitted to the correction data transport module 51 of correcting unit 5; Carry out the dot product operation through 52 pairs of correction matrixs of correction module and image to be displayed; To realize pointwise correction; To send LED display 2 to through correction data transport module 51 through the image to be displayed of overcorrect at last and show, accomplish demonstration heterogeneity pointwise correction process.After whole assessment, trimming process finished, corrective system can be held original annexation and stay in the LED display system, continues to bear online pointwise correction work; Also can only keep original annexation to stay in the LED display system correcting unit, continue to bear online pointwise correction work.

With reference to Fig. 3, LED display heterogeneity on-line correction method of the present invention, implementation step comprises as follows:

Step 1: the camera focal plane arrays (FPA) is carried out parameter calibration and anti-Gamma correction, i.e. γ ^-1Proofread and correct.

Because system of the present invention is used to test the uniformity index of weighing the demonstration of LED display lamp point; Wherein relate to test to brightness and colourity; Therefore at first need demarcate, promptly its redgreenblue light intensity response curve carried out γ the focal plane arrays (FPA) of camera such as the γ parameter of CCD chip and CMOS chip ^-1Proofread and correct, to obtain real LED display display brightness, chrominance information, to obtain real LED lamp point display message.

1a) demarcate camera γ parameter

Build camera γ parameter calibration system, use focal plane arrays (FPA) to be calibrated to take the redgreenblue incident light under the different brightness, obtain camera red channel measured value L _r, green channel measured value L _gWith blue channel measured value L _b, and use light power meter or nitometer to obtain the actual value C of red incident light brightness _r, green incident light brightness actual value C _gActual value C with blue incident light brightness _b, draw out the sensitometric curve of this camera redgreenblue passage in view of the above respectively, i.e. light intensity response curve, or substitution following formula can be tried to achieve focal plane arrays (FPA) red channel γ _rValue, green channel γ _gValue, blue channel γ _bValue:

$L_r={\mathrm{KC}}_r^{{\mathrm{\γ}}_r}+N_{\mathrm{\σ}}$

$L_g={\mathrm{KC}}_g^{{\mathrm{\γ}}_g}+N_{\mathrm{\σ}}$

$L_b={\mathrm{KC}}_b^{{\mathrm{\γ}}_b}+N_{\mathrm{\σ}},$

Wherein: K is the constant that the number with luminous intensity unit produces when transferring the gradation of image value to; N _σBe the estimation of white Gaussian noise that focal plane arrays (FPA) is produced in photoelectric conversion process, this noise is only relevant with the standard deviation sigma of noise gray-scale value.

1b) utilize the parameter of demarcating to carry out γ ^-1Proofread and correct

According to step 1a) the middle γ that demarcates gained _r, γ _g, γ _bNumerical value is to the estimation N of white Gaussian noise _σAnd focal plane arrays (FPA) is to the measured value L of actual scene _r, L _g, L _b, carry out γ ^-1Proofreading and correct, promptly is the following focal plane arrays (FPA) redgreenblue of the RGB triple channel gray-scale value difference substitution light intensity response inverse function C that stereoscopic camera is taken the gained image _r, C _g, C _b, ask for the corresponding real scene RGB triple channel gray-scale value C of this image _r, C _g, C _b:

$C_r={[(L_r-N_{\mathrm{\σ}})/K]}^{{{\mathrm{\γ}}_r}^{-1}}$

$C_g={[(L_g-N_{\mathrm{\σ}})/K]}^{{{\mathrm{\γ}}_g}^{-1}}$

$C_b={[(L_b-N_{\mathrm{\σ}})/K]}^{{{\mathrm{\γ}}_b}^{-1}}.$

Step 2: select a width of cloth test pattern f.

Different according to homogeneity assessment and Nonuniformity Correction purpose, test pattern f has following three kinds of selection modes:

If the user need carry out Nonuniformity Correction, the white image of selecting to have high-high brightness is as test pattern f, and the gray-scale value of the red channel of this test pattern f, green channel, blue channel is 255 at this moment;

If the brightness uniformity of user's needs assessment LED display when showing the certain luminance image, select a width of cloth RGB triple channel gray-scale value to coexist the gray level image of 0 to 255 a certain value as test pattern f;

The uniformity of chromaticity when if user's needs assessment LED display shows the particular color image, the image of then selecting to have this color is as test pattern f.

Step 3: calculate effective shooting distance.

3a) calculate effective shortest photographing distance

Because the LED lamp o'clock can not produce severe deformities at 30 degree visual angles with interior optical field distribution; In other words significant change can not take place in colourity; The human visual system can not think that colourity and monochrome information have serious distortion, therefore according to LED display width W and height H, can be known by the image optics principle; The shortest effective shooting distance is by the smaller value decision in the two of LED display width and height, i.e. effective shortest photographing distance L _MinFor:

L _min＝argmax{W/tan15°,H/tan15°}，

Wherein, to represent to ask for the maximal value in the argument list be dependent variable output to argmax{};

3b) calculate effectively the longest shooting distance

Be the precision of assurance test result and the accuracy of gaussian curve approximation; Need to guarantee that each LED lamp point has at least 4 * 4 focal plane arrays (FPA) pixels to represent; Therefore according to pixel line number M and columns N, the wide w of breadth and the high h of camera focal plane arrays (FPA), LED display lamp point line number K and columns L, LED display width W and height H; And camera lens focal length d; Can know that by geometric optical theory the longest effective shooting distance is by the smaller value decision in the two of LED display width and height, promptly effectively the longest shooting distance L _MaxFor:

L _max＝argmin{4WLd/Nw,4HKd/Mh}，

Wherein, to represent to ask for the minimum value in the argument list be dependent variable output to argmin{};

3c) according to the shooting bee-line L that tries to achieve _MinWith longest distance L _Max, the distance when confirming to take between camera and the LED display is at L _MinWith L _MaxBetween.

Step 4: build evaluating system, the initialization system parameter.

The preheating LED display;

Connect playing controller and LED display, camera and assessment unit, assessment unit and correcting unit;

Set the camera focal plane arrays (FPA) light sensitivity that is fit to prevailing circumstances brightness; All identical for guaranteeing each time shutter of taking simultaneously; All identical aperture F number and shutter speed in the time of should setting each take; Wherein aperture F number can be made as 8 or 9 or 11, and shutter speed can be made as 1/2000 second or 1/2500 second or 1/3200 second or 1/4000 second;

Make the shooting distance scope [L in the step 3) through setting camera lens focal length d _Min, L _Max] fall in the viewing areas of LED display instructions regulation, guarantee that simultaneously the camera lens optical axis direction is consistent with the optical axis direction or the LED display surface normal direction of LED display lamp point.

Step 5: take test pattern.

The control LED display shows test pattern f, and utilizes the camera that has configured parameter in the step 4 that test pattern f is repeated to take L time apace, obtains image sets f _l, l ∈ { 0,1,2;, L}, L>=2; Should guarantee in the shooting process that camera and LED display relative distance are constant, the result that all images shooting results all should be demarcated according to camera parameter in the step 1, by Color Channel respectively the inverse function of substitution focal plane arrays (FPA) RGB light intensity response carry out γ ^-1Proofread and correct, obtain the RGB triple channel gray-scale value of the corresponding real scene of this bearing member, image, the image f after will proofreading and correct again _lPassing to subsequent processing steps handles.

Step 6: Data Receiving and squelch.

For eliminating the noise that platform disturbance, camera focal plane arrays (FPA) dark current, electronics thermal motion and the flicker of LED display curtain cause, the image sets f that step 5 is obtained _lIn each width of cloth image addition make even all, remove the dark noise and LED display demonstration jittering noise of focal plane arrays (FPA), obtain the denoising image:

$\stackrel{\‾}{f}=\frac{1}{L}\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{f}_l.$

Step 7: utilize the method for background forecast, cut apart lamp point zone and the background area of denoising image

.

Comprise LED lamp point that is matrix distribution and the black background that is made up of morion in the LED display screen area, lamp point shape and optical field distribution thereof are generally circle, are projected on the camera focal plane also for circular.Therefore for obtaining the demonstration situation of LED display lamp point, eliminate the influence of background area, can utilize this type characteristics of LED display, utilize dividing method, separate the lamp point zone and the background area of LED display based on background forecast:

7a) background in the inhibition denoising image , enhancing lamp point light-emitting zone

This step adopts structural element as shown in Figure 2 to suppress background, strengthens the lamp point.Among Fig. 2, B _oAnd B _iRepresent two shape same disk shape structural elements, wherein B _oBe the bigger disc structure element of radius, B _iBe the less disc structure element of radius, Δ B=B _o– B _iBe big disk structural element B _oThan roundel structural element B _iAdditional circular ring structure element, B _bWhat represent is the disc structure element of a size in a certain scope, circular ring structure element Δ B and disc structure element B _bBe needed two structural elements.In addition, the radius of S (B) representative structure element, S (Δ B) represents the radius of circular ring structure element Δ B, and promptly circle ring center is to the distance of the peripheral profile of annulus, hence one can see that S (Δ B)=S (B _o), and S (B _o) greater than S (B _i).S (B in the practical application _b) can simply be taken as greater than interior ring radius S (B _i), less than the lamp dot spacing from D _iRadius distance value, for obtaining more accurate result, then B _bRadius should meet the regulation of following formula:

S(B _b)∈[3S(R _led)+S(B _o)-2S(B _i)/2,D _i-S(B _o)]，

Wherein, R _LedBe the projection of LED lamp point on image, S (R _Led) be its radius with pixel count tolerance.

Definition B _OiBe Δ B and B _bThe mathematical morphology structural element of joint provisions is with above-mentioned Δ B and B _bThe denoising image that step 6 is obtained Do the cap conversion of white top, with inhibition background, enhancing LED lamp point light-emitting zone, acquisition lamp point enhancing image D (x, y):

$D(x,y)=\stackrel{\‾}{f}(x,y)-(\stackrel{\‾}{f}\⊗B_{\mathrm{oi}})(x,y),$

Wherein: (x, y) locations of pixels in the presentation video;

is a kind of morphology operations of the present invention's definition;

expression is asked for the image that only comprises background from denoising image

, and

Be the morphology dilation operation,

Expression adopts structural element Δ B to the denoising image

Carry out expansive working, make that (u representes that v) the structural element center is at the denoising image

In the position, then $(\stackrel{\‾}{f}\⊕\mathrm{\Δ\; B})(x,y)=\underset{u,v}{\mathrm{Max}}\{f(x-u,y-v)+\mathrm{\Δ\; B}(u,v)\};$

Be the morphological erosion computing, Structural element B is adopted in expression _bTo image

Carry out erosion operation, wherein

7b) extract the lamp dot image that only comprises LED lamp point

This step adopts the big Tianjin method based on grey level histogram, through asking for the extraction that makes the maximized threshold value of inter-class variance realize the lamp dot image:

7b1) according to step 7a) the lamp point that obtains strengthens image D, is the Probability p (k) that the pixel of k occurs in image according to the computes gray-scale value:

$p\left(k\right)=\frac{1}{\mathrm{MN}}\underset{D(x,y)=k}{\mathrm{\Σ}}=1,$

Wherein, and D (x, y) (resolution that lamp point strengthens image D is M * N for x, the y) gray-scale value of position, and GTG is m, therefore 0≤x≤M, 0≤y≤N, D (x, y) ∈ [0, m-1] among the image D for lamp point strengthens;

Be t 7b2) with the gray-scale value threshold setting, background area and the target area collection of pixels of definition after cutting apart be respectively D (x, y)≤t} and { (x y)＞t}, obtains the pixel count proportion omegab that the background area pixels number accounts for entire image to D ₀(t) and lamp point area pixel number account for the pixel count proportion omegab of entire image ₁(t):

${\mathrm{\ω}}_0\left(t\right)=\underset{0\≤k\≤t}{\mathrm{\Σ}}p\left(k\right),$

${\mathrm{\&omega;}}_1\left(t\right)=\underset{t<k\&le;m-1}{\mathrm{\&Sigma;}}p\left(k\right);$

7b3) by step 7b1) and 7b2), be calculated as follows the brightness average μ that obtains the background area ₀(t) and the brightness average μ in lamp point zone ₁(t):

${\mathrm{\&mu;}}_0\left(t\right)=\underset{0\&le;k\&le;t}{\mathrm{\&Sigma;}}\mathrm{kp}\left(k\right)/{\mathrm{\&omega;}}_0\left(t\right),$

${\mathrm{\&mu;}}_1\left(t\right)=\underset{t<k\&le;m-1}{\mathrm{\&Sigma;}}\mathrm{kp}\left(k\right)/{\mathrm{\&omega;}}_1\left(t\right);$

7b4) according to step 7b3) in the brightness average μ of the background area that obtains ₀(t) and the brightness average μ in lamp point zone ₁(t), calculate the brightness grand mean μ that lamp point strengthens all pixels of image D:

μ＝ω ₀(t)μ ₀(t)+ω ₁(t)μ ₁(t)；

The data of 7b5) trying to achieve according to above-mentioned steps obtain the needed image optimum threshold value of big Tianjin threshold segmentation method g:

$g=\underset{0\&le;t\&le;m-1}{\max }\{{\mathrm{\&omega;}}_0\left(t\right){[{\mathrm{\&mu;}}_0\left(t\right)-\mathrm{\&mu;}]}^2+{\mathrm{\&omega;}}_1\left(t\right){[{\mathrm{\&mu;}}_1\left(t\right)-\mathrm{\&mu;}]}^2\};$

(x y) puts 0, with (x y) puts 1, obtains being used for extracting the two-value template of image lamp point content greater than the grey scale pixel value D of threshold value g among the lamp point enhancing image D 7b6) lamp point to be strengthened the grey scale pixel value D that is less than or equal to threshold value g among the image D;

7b7) with 7b6) in the two-value template and the denoising image

that obtain do point multiplication operation, can only be comprised the lamp dot image

of LED lamp point content

7c) utilize gray scale paddy peak value algorithm, determine lamp point light-emitting zone and form pixel

Because the bearing member, image that the camera focal plane arrays (FPA) collects is three-dimensional; Horizontal ordinate x in the image and ordinate y recording pixel locus; The 3rd z is grey scale pixel value, is similar to the height in the topomap, and wherein high gray-scale value and low gray-scale value are similar to mountain peak and the low ebb in the topomap; Therefore gray scale paddy peak value algorithm capable of using confirms that the LED lamp puts the drop shadow spread of luminous light field in bearing member, image, the i.e. light-emitting zone of LED lamp point.Simultaneously owing to each LED lamp point light-emitting zone in the lamp dot image of cutting apart through big Tianjin method

does not interconnect; Closure according to connected region; The gray scale peak value of each LED lamp point light-emitting zone is the maximal value of photosensitive unit; The lamp dot image

of the central point of this LED lamp point just, so gray scale paddy peak value algorithm process M capable of using * N size

Described gray scale paddy peak value algorithm comprises horizontal convolution projection and vertical convolution project, calculates by its definition, can obtain the lamp dot image

Horizontal convolution projection H _[x]With vertical convolution projection V _[y]The result:

$H_{\left[x\right]}=\frac{1}{N}\underset{x=1}{\overset{N}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{F}(x,y),$

$V_{\left[y\right]}=\frac{1}{M}\underset{y=1}{\overset{M}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{F}(x,y),$

Row and column split image with place, valley center in the projection is the rectangular area; Be decided to be (i with the capable j row of i in a projection point that place, peak value center row and column intersects; J) center of individual LED lamp point; Wherein 0≤i≤K-1,0≤j≤L-1 utilize the connected domain determination methods can determine the light-emitting zone and the composition pixel thereof of each LED lamp point then.

Step 8: colour mixture and colour system coordinate conversion are handled.

8a) the composition pixel colour mixture of lamp point light-emitting zone

Owing to arrive the picture that retina becomes after the long propagation of RGB three beams light process that each lamp point that human eye is seen is a LED lamp point to be sent when watching LED display; This three beams light is not accomplished color mixture in communication process; Therefore the LED lamp dot image color mixture of utilizing focal plane arrays (FPA) to obtain is inhomogeneous, be not LED lamp point the color and the brightness that really will show.The LED display displaying contents that the present invention proposes to utilize the colour mixture principle that focal plane arrays (FPA) is obtained carries out the colour mixture processing; According to step 7b) resulting lamp dot image F and step 7c) the LED lamp point light-emitting zone that obtains and form pixel location information; The gray-scale value of forming pixel in each LED lamp point light-emitting zone is made even all by the addition of RGB triple channel, obtain the lamp point RGB gray matrix of LED display under the RGB colorimetric system coordinate, be designated as [R; G; B], wherein R, G, B are the two-dimensional matrix of K * L, the three-channel demonstration situation of expression all lamp point RGBs of LED display;

8b) conversion colour system coordinate

To be converted to CIE L as follows at the lamp point RGB gray matrix [R, G, B] under the RGB colorimetric system coordinate ^*u ^*v ^*The space:

8b1) press following formula and change colour system coordinate to the XYZ space under the rgb space:

[X,Y,Z]=[T][R,G,B]，

Wherein: [X, Y, Z] is the colour system coordinate of LED lamp point demonstration situation under XYZ space;

$\left[T\right]={\left[\begin{array}{ccc}3.240479& -1.53715& -0.498535\\ -0.969256& 1.875992& 0.041556\\ 0.055648& -0.204043& 1.057311\end{array}\right]}^{-1};$

8b2) press colour system coordinate under the following formula conversion XYZ space to L ^*u ^*v ^*The space obtains lamp point L ^*u ^*v ^*Colour system coordinate L under the space ^*, u ^*And v ^*, the L of all lamp points ^*u ^*v ^*Spatial value is formed the coordinate figure matrix L of K * L dimension ^*, U ^*, V ^*:

$\left\{\begin{array}{c}{L}^{*}=116{(y/Y_0)}^{1/3}-16\\ u^{*}=13L^{*}(u^{\&prime;}-{u^{\&prime;}}_0)\\ v^{*}=13L^{*}(v^{\&prime;}-{v^{\&prime;}}_0)\end{array}\right.$

Wherein: x, y, z are the elements in [X, Y, Z], the colour system coordinate of indication lamp point display color under XYZ space;

X ₀, Y ₀, Z ₀White look reference coordinate for XYZ space under the D65 lighting condition;

U', v' are the intermediate variable of chromaticity coordinate, and its computing formula is:

$\left\{\begin{array}{c}{u}^{\&prime;}=4x/(x+15y+3z)\\ v^{\&prime;}=y/(x+15y+3z)\end{array}\right.;$

U' ₀, v' ₀Be L under the D65 lighting condition ^*u ^*v ^*White the look reference coordinate in space, its computing formula is:

$\left\{\begin{array}{c}{u^{\&prime;}}_0=4X_0/(X_0+15Y_0+{3Z}_0)\\ {v^{\&prime;}}_0=9Y_0/(X_0+15Y_0+{3Z}_0)\end{array}\right..$

Step 9: the coordinate figure matrix fitting is become Gaussian curve, and carry out the uniformity index assessment.

9a) gaussian curve approximation

Get the L of all lamp points of LED display respectively ^*, U ^*, V ^*The coordinate figure matrix carries out the frequency distribution measuring, is about to L ^*, U ^*, V ^*The three-dimensional value matrix is got its maximal value respectively and minimum value is subtracted each other, and span is carried out five equilibrium according to LED lamp point resolution K * L, calculates L in each five equilibrium ^*, U ^*, V ^*The quantity of coordinate figure, this is the frequency of three-dimensional value, after the gross error value during the rejecting frequency distributes, by L ^*, U ^*, V ^*The coordinate figure matrix is the fitted gaussian curve respectively, can obtain the lightness dimension average of lamp point display color

The chromaticity coordinate average

With

9b) calculate uniformity index with the L that is obtained according to colour system coordinate mean data ^*u ^*v ^*Colour system coordinate average

Substitution brightness uniformity assessment formula and uniformity of chromaticity are assessed formula, can obtain the brightness uniformity index value of LED display

With the uniformity of chromaticity index value

${\mathrm{\&Delta;ME}}_l^{*}=\frac{1}{K\&times;L}\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{L}{\mathrm{\&Sigma;}}}\sqrt{{(L^{*}(i,j)-\stackrel{\&OverBar;}{l^{*}}/\stackrel{\&OverBar;}{l^{*}})}^2}\&times;100\%$

Wherein, (i, j) (i, j) individual element, i.e. (i, j) the individual lamp point in the denotation coordination value matrix;

9c) the show uniformity of assessment LED display

When the brightness uniformity index value of being tried to achieve and uniformity of chromaticity index value satisfy

respectively less than 5%,

less than 0.03 the time, judge that then the brightness uniformity of this LED display and uniformity of chromaticity are up to standard;

When brightness uniformity index value surpasses 5%; The brightness uniformity of then judging LED display is not up to standard; When uniformity of chromaticity index value

surpasses 0.03; The uniformity of chromaticity of then judging this LED display is not up to standard, carries out next step and generates correction matrix;

If the homogeneity also during needs assessment test LED display display image only need be got back to step 3 control LED display and shown this test pattern f, and carry out subsequent step once more successively and get final product.

Step 10: generate correction matrix.

With the coordinate mean data

Be converted to rgb space, and red channel average, green channel average and the blue channel average of rgb space is extended for the equal value matrix of congruence that K * L ties up respectively, obtain the equal value matrix μ of red channel _r, green channel equal value matrix μ _gEqual value matrix μ with blue channel _b, and, obtain the lamp point red channel correction matrix R under the rgb space through following formula _r, green channel correction matrix R _gWith blue channel correction matrix R _b:

R _r=μ _r./R

R _g=μ _g./G

R _b=μ _b./B，

Wherein " ./" be a some divided-by symbol, R _r, R _g, R _bBe the matrix of K * L dimension.

Step 11: the correction matrix that utilizes step 10) to obtain carries out pointwise correction to data to be displayed.

With above-mentioned correction matrix R _r, R _g, R _bRespectively with the red channel gray level image F of image to be displayed _r, green channel gray level image F _gWith blue channel gray level image F _bCarry out pointwise correction and handle, obtain proofreading and correct the red channel image G of back image to be displayed G _r, green channel image G _gWith blue channel image G _b:

G _r=R _r.×F _r

G _g=R _g.×F _g

G _b=R _b×.，

Wherein, ". * " be the dot product symbol, two matrixes that expression will be participated in computing multiply each other by element.

Step 12: the red channel image G of the image to be displayed G after will proofreading and correct _r, green channel image G _gWith blue channel image G _bOutput to LED display and show, promptly accomplished the online pointwise correction of LED display heterogeneity.

More than describing only is an instantiation of the present invention; Do not constitute any restriction of the present invention; Obviously to those skilled in the art, after having understood content of the present invention and principle, all maybe be under the situation that does not deviate from the principle of the invention, structure; Carry out various corrections and change on form and the details, but these are based on the correction of inventive concept with change still within claim protection domain of the present invention.

Claims (7)

1. a LED display heterogeneity on-line correction system comprises playing controller (1), equipment LED display to be assessed (2) and camera (3), it is characterized in that: also comprise assessment unit (4) and correcting unit (5),

Described assessment unit (4) comprising:

Assessment data transport module (41) is used to receive LED display (2) displaying contents that camera (3) is taken;

Denoising module (42) is used to read the data of assessment data transport module (41) and carries out denoising;

The lamp point is cut apart module (43), utilizes the method for background forecast that the denoising image is carried out cutting apart of lamp point and background area;

Colour mixture module (44), utilize the colour mixture principle lamp point of cutting apart the back image is formed that pixel is carried out the colour mixture processing and with RGB RGB system coordinate conversion to L ^*u ^*v ^*The colour system coordinate obtains LED display displaying contents L ^*u ^*v ^*Lightness index L under the colour system coordinate ^*, horizontal ordinate chromaticity index U ^*With ordinate chromaticity index V ^*, i.e. lightness dimension L ^*With chromaticity coordinate U ^*, V ^*

Gaussian curve approximation module (45) is to lightness dimension L ^*With chromaticity coordinate U ^*, V ^*Carry out gaussian curve approximation respectively; The lightness dimension that obtains according to curve fitting and the average of chromaticity coordinate; Calculate the index value of brightness uniformity and uniformity of chromaticity; If the uniformity of chromaticity index value surpass 0.03 or the brightness uniformity index value surpass 5%, judge that then this LED display (2) show uniformity is not up to standard, and this information not up to standard passed to correction matrix generation module (46);

Correction matrix generation module (46), colour system coordinate average that passes over according to gaussian curve approximation module (45) and lamp point colourity and lightness dimension generate L ^*u ^*v ^*Correction matrix under the colour system coordinate is converted to this correction matrix under the RGB of RGB system space, and the correction matrix of the rgb space that obtains is transferred to correcting unit (5).

Described correcting unit (5) comprising:

Correction data transport module (51) is used for receiving from the image to be displayed of playing controller (1) and the correction matrix of assessment unit (4), and the image to be displayed after will proofreading and correct is sent to LED display (2) and shows;

Correction module (52) is used for correction matrix that correction matrix generation module (46) is generated and the image to be displayed of importing from playing controller (1) and multiplies each other by point, realizes the heterogeneity pointwise correction of real-time online.

2. LED display heterogeneity on-line correction according to claim 1 system; It is characterized in that: described assessment data transport module (41) has two external interfaces at least, adopts the CameraLink interface on USB interface or network interface or the special image capture card to realize; Described correction data transport module (51) has three external interfaces at least, adopts USB interface or network interface to realize.

3. LED display heterogeneity on-line correction according to claim 1 system; It is characterized in that: said correction data transport module (51) is connected with playing controller (1), LED display (2) and assessment data transport module (41) respectively; Correction module (52) links to each other with correction data transport module (51), constitutes corrective system; Timing, correction module (52) are according to from the correction matrix of assessment data transport module (41), and image to be displayed is proofreaied and correct on pointwise ground, and the data after will proofreading and correct send to LED display (2) through correction data transport module (51) and show; During the homogeneity assessment, correction module (52) is not proofreaied and correct data to be displayed, directly will output to LED display (2) from the image to be displayed of playing controller (1) and show.

4. LED display heterogeneity on-line correction according to claim 1 system; It is characterized in that: said camera (3) adopts the focal plane array image-forming equipment after parameter calibration and anti-Gamma correction; Its focal plane array is classified CCD device or cmos device as, and the taken image of this camera is transferred to assessment data transport module (41) through hardware interface such as CameraLink or USB or network interface.

5. a LED display heterogeneity on-line correction method comprises the steps:

1) the camera focal plane arrays (FPA) is carried out parameter calibration and anti-Gamma correction, i.e. γ ^-1Proofread and correct;

2) select a width of cloth test pattern f;

3) calculate effective shooting distance:

3a) according to LED display width W and height H, confirm effective shortest photographing distance L _MinFor:

L _min＝argmax{W/tan15°,H/tan15°}，

Wherein, to represent to ask for the maximal value in the argument list be dependent variable output to argmax{};

3b) according to pixel line number M and columns N, the wide w of breadth and the high h of camera focal plane arrays (FPA), LED display lamp point line number K and columns L, LED display width W and height H, and camera lens focal length d are confirmed effectively the longest shooting distance L _MaxFor:

L _max＝argmin{4WLd/Nw,4HKd/Mh}，

Wherein, to represent to ask for the minimum value in the argument list be dependent variable output to argmin{};

The shortest photographing distance L that basis is tried to achieve when 3c) taking _MinThe longest shooting distance L _Max, the distance between assurance camera and the LED display is at L _MinWith L _MaxBetween;

4) the preheating LED display sets up camera, sets the light sensitivity and the camera lens focal length d of camera shutter speed, aperture F number size, focal plane arrays (FPA), makes the shooting distance scope [L in the step 3) _Min, L _Max] fall in the viewing areas of LED display instructions regulation, guarantee that the camera lens optical axis direction is consistent with the optical axis direction or the LED display surface normal direction of LED display lamp point;

5) the control LED display shows test pattern f, and utilizes camera that test pattern f is taken L time, obtains image sets f _l, l ∈ 0,1,2 ..., L}, L>=2;

6) with image sets f _lIn each width of cloth image addition make even all, remove the dark noise and LED display demonstration jittering noise of focal plane arrays (FPA), obtain the denoising image

7) utilize the method for background forecast, the lamp point zone and the background area of denoising image

cut apart:

7a) denoising image

is carried out the cap conversion of white top; To suppress the background area; Strengthen LED lamp point zone; Acquisition lamp point enhancing image D (x, y):

$D(x,y)=\stackrel{\&OverBar;}{f}(x,y)-(\stackrel{\&OverBar;}{f}\&CircleTimes;B_{\mathrm{oi}})(x,y),$

Wherein: (x, y) location of pixels in the presentation video;

B _OiBe by Δ B and B _bThe mathematical morphology structural element of regulation, Δ B is the toroidal structural element, B _bBe the disc-shaped structure element;

is a kind of morphology operations of the present invention's definition;

expression is asked for the image that only comprises background from denoising image , and

is the morphology dilation operation,

represent to adopt structural element Δ B that image is carried out expansive working;

Be the morphological erosion computing,

Structural element B is adopted in expression _bTo image

Carry out erosion operation;

7b) adopt big Tianjin method that lamp point is strengthened image D (x; Y) carry out binary conversion treatment; The two-value template that obtains; Two-value template and denoising image

are multiplied each other by point, only comprised the lamp dot image

of LED lamp point content

7c) utilize gray scale paddy peak value algorithm to determine the light-emitting zone and the composition pixel thereof of each LED lamp point in the lamp dot image

;

8) theoretical based on blend of colors, in the mixed lamp dot image of rgb space

Middle lamp point light-emitting zone is formed color of pixel, and each grey scale pixel value that is about in the LED lamp point light-emitting zone is made even all by the addition of RGB triple channel, obtains lamp point RGB gray matrix [R, G, B], and it is converted to L ^*u ^*v ^*The space obtains the L of all lamp points ^*u ^*v ^*Volume coordinate value matrix [L ^*, U ^*, V ^*], lightness dimension value matrix L wherein ^*With the chromaticity coordinate value matrix U ^*, V ^*Dimension is K * L;

9) the coordinate figure matrix fitting is become Gaussian curve, and carries out the uniformity index assessment:

9a) get the coordinate figure matrix L earlier ^*, U ^*, V ^*Carry out the frequency distribution measuring respectively, distinguish the fitted gaussian curve again after the gross error value during the rejecting frequency distributes, obtain L ^*u ^*v ^*The lightness dimension mean data that lamp point shows under the colour system space With the chromaticity coordinate mean data

9b) according to L ^*u ^*v ^*Colour system coordinate mean data is calculated uniformity index:

With the L that is obtained ^*u ^*v ^*Colour system coordinate average

Substitution brightness uniformity assessment formula and uniformity of chromaticity assessment formula can obtain LED display brightness uniformity index With the uniformity of chromaticity index

${\mathrm{\&Delta;ME}}_l^{*}=\frac{1}{K\&times;L}\underset{i=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{L}{\mathrm{\&Sigma;}}}\sqrt{{(L^{*}(i,j)-\stackrel{\&OverBar;}{l^{*}}/\stackrel{\&OverBar;}{l^{*}})}^2}\&times;100\%$

Wherein, (i, j) (i, j) individual element, i.e. (i, j) the individual lamp point in the denotation coordination value matrix;

9c) according to the brightness uniformity index

and color uniformity index

evaluate the uniformity of the LED display shows:

When the brightness uniformity index value of being tried to achieve and uniformity of chromaticity index value satisfy

respectively less than 5%,

less than 0.03 the time, judge that then the show uniformity of this LED display is up to standard;

When brightness uniformity index value

surpasses 5%; Or uniformity of chromaticity index value

was above 5% o'clock; The show uniformity of then judging this LED display is not up to standard, carries out next step;

10) according to L ^*u ^*v ^*Colour system coordinate mean data and lamp point RGB gray matrix generate correction matrix:

With the coordinate mean data Be converted to rgb space, and red channel average, green channel average and the blue channel average of rgb space is extended for the equal value matrix of congruence that K * L ties up respectively, obtain the equal value matrix μ of red channel _r, green channel equal value matrix μ _gEqual value matrix μ with blue channel _b, and, obtain the lamp point red channel correction matrix R under the rgb space through following formula _r, green channel correction matrix R _gWith blue channel correction matrix R _b:

R _r=μ _r./R

R _g=μ _g./G

R _b=μ _b./B，

Wherein " ./" be a some divided-by symbol, R _r, R _g, R _bBe the matrix of K * L dimension;

11) correction matrix that utilizes step 10) to obtain carries out pointwise correction to data to be displayed:

With above-mentioned correction matrix R _r, R _g, R _bRespectively with the red channel gray level image F of image to be displayed _r, green channel gray level image F _gWith blue channel gray level image F _bCarry out pointwise correction by following formula and handle, obtain proofreading and correct the red channel image G of back image to be displayed G _r, green channel image G _gWith blue channel image G _b:

G _r=R _r.×F _r

G _g=R _g.×F _g

G _b=R _b×.，

Wherein, ". * " be the dot product symbol, two matrixes that expression will be participated in computing multiply each other by element;

12) the red channel image G of the image to be displayed G after will proofreading and correct _r, green channel image G _gWith blue channel image G _bExport to LED display and show, accomplish LED display heterogeneity pointwise correction.

6. LED display heterogeneity on-line correction method according to claim 5 is characterized in that: the test pattern f said step 2), and different according to homogeneity assessment and Nonuniformity Correction purpose have following three kinds of selection modes:

If the user need carry out Nonuniformity Correction, the white image of selecting to have high-high brightness is as test pattern f, and the gray-scale value of the red channel of this test pattern f, green channel, blue channel is 255 at this moment;

If the brightness uniformity of user's needs assessment LED display when showing the certain luminance image, select a width of cloth RGB triple channel gray-scale value to coexist the gray level image of 0 to 255 a certain value as test pattern f;

The uniformity of chromaticity when if user's needs assessment LED display shows the particular color image, the image of then selecting to have this color is as test pattern f.

7. LED display heterogeneity on-line correction method according to claim 5; It is characterized in that: gray scale paddy peak value algorithm said step 7c); Comprise horizontal convolution projection and vertical convolution project, for being defined as of lamp dot image

its horizontal convolution project of K * L size:

$H_{\left[x\right]}=\frac{1}{N}\underset{x=1}{\overset{N}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{F}(x,y),$

Being defined as of vertical convolution project:

$V_{\left[y\right]}=\frac{1}{M}\underset{y=1}{\overset{M}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{F}(x,y),$

Utilize horizontal convolution projection and vertical convolution projection process lamp dot image

Obtain the lamp dot image

Horizontal convolution projection H _[x]With vertical convolution projection V _[y]The result; Row and column split image with place, valley center in the projection is the rectangular area; With the center of LED lamp point in the location positioning rectangular area of place, peak value center row and column intersection in the projection, utilize the connected domain method to determine the light-emitting zone and the composition pixel of each LED lamp point.

Images