A method for globally evaluating mura defects of a display panel based on a minimum perceptible difference

Abstract

The application discloses a display panel Mura defect global evaluation method based on a minimum perceptible difference, overcomes the defects of manual detection of panel Mura defects, makes the final Mura defect quantitative evaluation not affected by artificial subjective identification factors, significantly improves the efficiency and accuracy of display panel quality detection, and especially compared with previous evaluation methods, the method can evaluate the display template Mura defect by considering the images captured by the camera from multiple angles, can simultaneously quantitatively evaluate the brightness and chroma Mura defects, and adopts an image mask processing method to avoid the missed detection of the panel edge Mura defect, the method is not only for detecting and evaluating the most obvious Mura defect, but also comprehensively evaluates according to the different weights of all observable defect intensities from the global, so that the accuracy and reliability of the display panel Mura defect evaluation are improved.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, specifically relating to a global evaluation method for display panel Mura defects based on minimum perceptible difference. Background Technology

[0002] Currently, the market share of electronic devices such as mobile phones and computers is experiencing explosive growth. Among them, electronic displays such as liquid crystal displays (LCDs), plasma displays (PDPs), and organic light-emitting diodes (OLEDs) are becoming increasingly popular among consumers as high-quality displays. This has also placed increasingly higher demands on the production capacity and quality of display panels. At present, most manufacturers rely mainly on quality inspectors to visually assess the quality of display panels. This has problems such as the significant impact of human subjectivity on the test results, the lack of fixed standards for judgment, and the positive correlation between error and working time. Therefore, there is an urgent need to develop a complete display panel quality inspection system and improve the accuracy and efficiency of the inspection.

[0003] Generally, a major factor reducing the image quality of a panel display is non-uniformity, also known as "mura." Mura is a common visual defect in displays such as LCD and OLED, typically manifesting as low contrast, uneven brightness areas, and blurred edges, causing visual discomfort to users. Mura defects are also among the most complex and difficult to detect and assess, and include various types such as dot-like mura, linear mura, and blocky mura.

[0004] Currently, most electronic display manufacturers use limited samples for visual inspection. However, this method inevitably introduces factors such as human subjectivity, resulting in low accuracy and inefficient inspection. With technological advancements, researchers have begun using machine vision to replace the human eye, employing high-precision cameras combined with efficient detection algorithms to evaluate panels. High-precision cameras capture images of the display screen under specific conditions, and then appropriate algorithms are used to detect and evaluate Mura defects in the images, providing quantitative indicators of Mura defects in the display panel.

[0005] However, accurately quantifying and assessing Mura defects during inspection has always been a recognized challenge in the industry. Therefore, a method for high-precision assessment of Mura defects is urgently needed. While various measurement methods for luminance Mura or chromaticity Mura have been reported at academic conferences, these methods are still in the laboratory research and testing phase, and no reports have clearly provided a complete automated quantitative assessment scheme for Mura that can be used in industrial production.

[0006] I. Existing methods for assessing Mura defects in display panels generally only evaluate brightness uniformity and cannot take into account chromatic Mura defects.

[0007] Second, existing evaluation methods generally require complex image algorithms to detect and identify the shape and location of the Mura, and then evaluate different Mura types. These methods mainly target the local defect characteristics of the screen and ignore relatively weak defects, so they often cannot evaluate the global uniformity of the display panel as a whole.

[0008] Third, existing methods generally only evaluate images taken from the angle where the camera is directly facing the screen, and cannot take into account the Mura defects of the screen observed from other angles. Some display panel defects are not as easy to observe from the front view as they are from the side view.

[0009] Fourth, existing methods generally do not perform special processing on the screen edge area, making it difficult to avoid the problem of missing detection of edge Mura defects;

[0010] Fifth, existing methods are often only validated on simulated defect images or real-world data with a limited number of Mura types, making it difficult to guarantee that the methods are applicable to all types of Mura defects. Summary of the Invention

[0011] The technical problem to be solved by the present invention is to overcome the existing defects and provide a global evaluation method for display panel Mura defects based on minimum perceptible difference, so as to solve the problems mentioned in the background art.

[0012] To achieve the above objectives, the present invention provides the following technical solution: a global evaluation method for display panel Mura defects based on minimum perceptible difference, comprising the following steps:

[0013] Step 1: The system first acquires complete images of the display panel from multiple angles using a high-definition camera, locates the screen area for cropping, scaling, and noise reduction filtering, then converts the pre-processed RGB color gamut image into a CIE-XYZ color gamut image, and further transforms the CIE-XYZ color gamut image into a color space (wk, rg, by).

[0014] Step 2: Calculate the test contrast map and the reference contrast map, and generate a differential contrast map from the two. For Mura defects present in the image, an edge mask is introduced to process the image.

[0015] Step 3: Perform Fourier transform on the three components of color space (wk, rg, by) to obtain frequency components, calculate the contrast sensitivity CSF function describing the human visual system, apply the single-channel contrast sensitivity function (CSF) of each of the three frequency components wk / rg / by to filter, and perform inverse Fourier transform to the corresponding spatial domain.

[0016] Step 4: Convert the inverse Fourier transform spatial domain (wk,rg,by) image to CIE-XYZ space, and further transform it to CIE-Lab space to calculate the luminance component L and chrominance component C in the CIE-Lab color gamut.

[0017] Step 5: Calculate the minimum perceptible difference (JND) map for the luminance component L and the chrominance component C respectively. Calculate the global maximum JND value of the luminance and chrominance components and the ratio of the Mura defect area to the total screen area according to the formula.

[0018] Step 6: Calculate the evaluation indices LEV and CEV for the luminance and chrominance components based on the global maximum JND value and the area ratio of the Mura defect region. Summarize the evaluation results of the multi-angle images and finally comprehensively evaluate the luminance and chrominance evaluation indices to obtain the final quantitative evaluation value MEV for the Mura defect of the display panel.

[0019] Preferably, in step one:

[0020] When shooting the display panel from multiple angles, the camera can generally take three angles: angle 1, where the camera is directly facing the center point of the screen and the lens axis coincides with the central axis of the screen; angle 2, where the camera is directly facing the center point of the screen, on the left side of the screen and at an angle of θ1 to the central axis of the screen; and angle 3, where the camera is directly facing the center point of the screen, on the right side of the screen and at an angle of θ2 to the central axis of the screen.

[0021] The specific conversion formula for converting an RGB color gamut image to a CIE-XYZ color gamut image is as follows:

[0022]

[0023]

[0024]

[0025] The original RGB values ​​of an image are typically read from 0 to 255. This invention first normalizes the RGB values ​​to reduce their range to 0-1. Noise denoising can be achieved by combining spatial domain Gaussian filtering and median filtering. The median filter performs convolutions of [5,5] kernel size on the three channels and expands the image by copying edge region values ​​to avoid the case where the values ​​at the four corners of the filtered image are 0. Then, the XYZ values ​​are calculated using the above formula and multiplied by 100 to reduce the range of XYZ values ​​to [0-100] for subsequent calculations.

[0026] Transform a CIE-XYZ color gamut image to a color space using the following transformation matrix:

[0027] wk = 0.279X + 0.72Y - 0.107Z

[0028] rg = -0.449X + 0.29Y - 0.077Z.

[0029] by=0.086X-0.59Y+0.501Z

[0030] Preferably, in step two:

[0031] The differential contrast map DC(x,y) is derived from the test contrast map C. Test (x,y) and reference contrast plot C Refence (x,y) is generated as follows:

[0032]

[0033] Where C Test (x,y) represents the original image converted to the color space (wk,rg,by), C Refence (x, y) is the background image after the original image has been filtered out of all mura defects by a Gaussian filter with a sufficiently low cutoff frequency, as shown in the following formula.

[0034]

[0035] In the above formula, x and y represent the row and column of the image;

[0036] r — visual resolution, unit: px / deg;

[0037] s — Gaussian scale, unit: deg, generally taken as 2 deg;

[0038] If mura defects exist at the edges of an image, traditional methods are prone to missing them. Therefore, an edge mask is needed to process the image. The specific algorithm is as follows:

[0039]

[0040] In the formula, M and N represent the image height and width, respectively, in pixels.

[0041] b gain — Boundary gain, which can generally be set to 1;

[0042] b scale — Boundary scale, unit: deg, generally 0.5 deg is acceptable;

[0043] By performing a dot product between each pixel of the differential contrast image DC(x,y) and the aforementioned edge mask function B(x,y), the accuracy of the algorithm for edge mura defect detection can be improved.

[0044] Preferably, in step three:

[0045] The formula for calculating the contrast sensitivity function (CSF) of the human visual system is shown below:

[0046]

[0047] in

[0048] ω=(u 2 +v 2 ) 1 / 2

[0049]

[0050] (u,v) is a spatial frequency coordinate, with the unit being each degree of period. This angle is the angle between the human eye and a point on the screen and the vertical line of the screen.

[0051] When processing the wk component, the values ​​in the contrast sensitivity function CSF can be taken as follows:

[0052]

[0053] γ(ω)=0.003100ω 2 -0.10680ω+1.396

[0054] m p =5.5σ p =1 / 40 ε=1

[0055] m c =5.5σ c =1 / 12 b ias =55

[0056] When processing the rg component, the values ​​in the contrast sensitivity function CSF can be taken as follows:

[0057]

[0058] γ(ω)=0.001531ω 2 -0.06149ω+1.140

[0059] m p =1.5σ p =1 / 70 ε=0

[0060] When processing the by component, the values ​​in the contrast sensitivity function CSF can be taken as follows:

[0061]

[0062] γ(ω)=0.001919ω 2 -0.06427ω+1.090

[0063] m p =1.5σ p =1 / 45 ε=1 / 4

[0064] m c =7.5σ c =1 / 20 b ias =0

[0065] In the above formula Let wk, rg, and by be the average values ​​of the three components of the color space;

[0066] Based on the above formula, the transfer function of the human visual system CSF model for the three components wk, rg, and by in the color space can be calculated. It's important to note that the spatial frequency ω in the above formula is in periods per degree, meaning it's given from the observer's perspective. However, the unit of the spectrum obtained through Fourier transform is periods per pixel. Therefore, the units are inconsistent, and the following formula is needed to convert the units when calculating the spatial frequency ω of the CSF:

[0067]

[0068] Where H R To measure the number of pixels horizontally on the screen in an image, V D W represents the vertical distance for visual observation. L Width of the display screen (mm);

[0069] In addition, in calculations involving calculations When there are three components, since the values ​​may be negative, it is necessary to take the modulus of the mean of the three components wk, rg, and by to ensure that the mean of the three components is always positive. This way, the data calculated by the CSF model can be consistent with the experimental results.

[0070] Then, the transfer function calculated using the CSF model is applied to perform frequency domain filtering on the three components (WK, RG, BY). The frequency domain filtering is completed by performing matrix multiplication between the spectrum of the three components and their corresponding CSF functions. It should be noted that the CSF function values ​​corresponding to the three components (WK, RG, BY) are relatively large. To prevent anomalies in subsequent calculations, this invention first synchronously normalizes the CSF function values ​​of the three components before frequency domain filtering. Specifically, the maximum value among the three CSF function values ​​is found, and each component is divided by this maximum value to ensure that the CSF function values ​​of the three components are always within [0,1], facilitating subsequent calculations. Finally, the frequency domain graphs of the three components (WK, RG, BY) after CSF function filtering are transformed into spatial domain components (wk, rg, by) using an inverse Fourier transform.

[0071] Preferably, in step four:

[0072] The transformation matrix for converting the color space (wk,rg,by) to the CIE-XYZ space is as follows:

[0073] X=0.626555wk-1.867178rg-0.153156by

[0074] Y=1.369855wk+0.934756rg+0.436229by

[0075] Z=1.505651wk+1.421324rg+2.536021by

[0076] Then, the image is transformed from the CIE-XYZ space and further transformed to the CIE-Lab space to obtain three components L,a,b;

[0077]

[0078] Where f(x) is the following piecewise function

[0079]

[0080] In the above formula, X n ,Y n Z n These are the most frequent values ​​of the X, Y, and Z components in the CIE-XYZ space, which can generally be obtained by statistically analyzing the rounded values ​​of each component.

[0081] The chromaticity component C can be calculated using the following formula based on a and b:

[0082] Preferably, in step five:

[0083] After obtaining the luminance component L and the chromaticity component C, the minimum perceptible difference (JND) maps for the luminance and chromaticity components are calculated separately according to Weber's theorem.

[0084]

[0085] Where Δ represents the luminance component L and the chromaticity component C, If it is a Gaussian window function, then β is the Minkowski distance coefficient, with a typical value of 2.4;

[0086] Based on the local JND diagram above, the global JND values ​​of the luminance and chrominance components can be further derived using the Minkowski function. This process can be called pooling, and its expression is:

[0087]

[0088] As γ→∞, the global maximum JND value for both the luminance and chrominance components can be obtained, i.e.

[0089] JND Δ_max =max[JND(x,y)]Δ∈{L,C};

[0090] Next, the ratio of the Mura defect area to the total screen area in the luminance and chrominance JND maps is calculated. The calculation method is as follows:

[0091]

[0092]

[0093] Where p(x,y) and q(x,y) are respectively

[0094]

[0095]

[0096] Preferably, in step six:

[0097] The luminance evaluation index LEV is LEV = JND. L_area1 +0.28JND L_max +6.47JND L_area2

[0098] Among them, JND L_area1 JND is the ratio of the Mura defect area with a brightness JND value greater than or equal to 1 to the total screen area. L_max JND is the global maximum value of the luminance JND map. L_area2 This represents the ratio of Mura defect areas with a brightness JND value greater than or equal to 2 to the total screen area.

[0099] The colorimetric evaluation index CEV is CEV = JND. C_area1 +0.16JND C_max +3.19JND C_area2

[0100] Among them, JND C_area1 JND is the ratio of Mura defect areas with a chromaticity JND value greater than or equal to 1 to the total screen area. C_max JND is the global maximum value of the chroma JND map. C_area2 This represents the ratio of Mura defect areas with a chromaticity JND value greater than or equal to 2 to the total screen area.

[0101]

[0102]

[0103] Where N is the total number of display panel images captured from multiple angles, LEV i ,i∈{1,2,3,…,N} represents the brightness evaluation index under multiple angles, CEV i ,i∈{1,2,3,…,N} are the colorimetric evaluation indexes under multiple angles, γ is the ratio between the two items, which can generally be taken as 0.6 here, and p is the Minkowski coefficient, which can generally be taken as 2;

[0104] The final Mura defect quantification value (MEV) for the display panel is calculated by comprehensively evaluating the brightness and chromaticity indicators using the following formula: MEV = 2.353 LEV. total +3.573CEV total ;

[0105] At this point, the quality of the display panel can be quantitatively evaluated based on the calculated Mura defect quantitative evaluation value MNEV.

[0106] Compared with existing technologies, this invention provides a global evaluation method for display panel Mura defects based on minimum perceptible difference, which has the following advantages:

[0107] 1. The system architecture of the present invention can not only evaluate brightness uniformity, but also take into account chromaticity Mura defects, compared with the existing display panel Mura defect evaluation method. Its algorithm has better scalability.

[0108] 2. The method proposed in this invention does not require the prior detection and identification of the shape and location of the mura before quantitative evaluation, thus enabling a more comprehensive and global defect assessment of the panel.

[0109] 3. Compared with existing methods, the method proposed in this invention takes into account the evaluation of screen images captured by the camera from multiple angles and fuses the defect evaluation results from multiple angles, resulting in better robustness of the algorithm.

[0110] 4. Compared with existing methods, the method proposed in this invention takes into account the problem that Mura defects in the screen edge area are easy to miss and proposes a method to process the image using an edge mask, which has higher algorithm intelligence;

[0111] 5. The system proposed in this invention overcomes the shortcomings of existing methods that are only verified on simulated defect images or actual data with a limited number of Mura types. It has been verified using 20 types of simulated Mura and collected actual Mura data, ensuring that the method is applicable to various types of Mura defects.

[0112] This invention overcomes the drawbacks of manual inspection of panel Mura defects, making the assessment of Mura defects unaffected by subjective human judgment. It can be applied not only to the quantitative assessment of brightness and chromaticity Mura defects in display panels, but also to the comprehensive global assessment considering the situation of cameras shooting from different angles on the screen. Furthermore, it takes into account the impact of missed detection of Mura defects in the screen edge area on the assessment, significantly improving the accuracy of obtaining display panel Mura defects. This technology can benefit more panel manufacturers and promote the establishment of relevant standards for the quantitative assessment of display panel Mura defects. Attached Figure Description

[0113] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention, but do not constitute a limitation thereof. In the drawings:

[0114] Figure 1 This is a schematic diagram of the Mura defect in the display panel proposed in this invention;

[0115] Figure 2 This is a schematic diagram of the multi-angle display panel image acquisition proposed in this invention;

[0116] Figure 3 This is a schematic diagram of the global evaluation method for Mura defects in display panels proposed in this invention;

[0117] Figure 4 This is a schematic diagram comparing the correlation coefficients between the subjective evaluation of defects in the actual testing method proposed in this invention and the algorithm evaluation results.

[0118] Figure 5 This diagram illustrates the correlation coefficients between the subjective evaluation and algorithmic assessment results of the method proposed in this invention for 20 major Mura defects.

[0119] Figure 6 This is a schematic diagram of the global evaluation system for Mura defects in display panels proposed in this invention. Detailed Implementation

[0120] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0121] Please see Figure 1-6 This invention provides a technical solution: a global evaluation method for display panel Mura defects based on minimum perceptible difference, comprising the following steps:

[0122] Step 1: The system first acquires complete images of the display panel from multiple angles using a high-definition camera, locates the screen area for cropping, scaling, and noise reduction filtering, then converts the pre-processed RGB color gamut image into a CIE-XYZ color gamut image, and further transforms the CIE-XYZ color gamut image into a color space (wk, rg, by).

[0123] Step 2: Calculate the test contrast map and the reference contrast map, and generate a differential contrast map from the two. For Mura defects present in the image, an edge mask is introduced to process the image.

[0124] Step 3: Perform Fourier transform on the three components of color space (wk, rg, by) to obtain frequency components, calculate the contrast sensitivity CSF function describing the human visual system, apply the single-channel contrast sensitivity function (CSF) of each of the three frequency components wk / rg / by to filter, and perform inverse Fourier transform to the corresponding spatial domain.

[0125] Step 4: Convert the inverse Fourier transform spatial domain (wk,rg,by) image to CIE-XYZ space, and further transform it to CIE-Lab space to calculate the luminance component L and chrominance component C in the CIE-Lab color gamut.

[0126] Step 5: Calculate the minimum perceptible difference (JND) map for the luminance component L and the chrominance component C respectively. Calculate the global maximum JND value of the luminance and chrominance components and the ratio of the Mura defect area to the total screen area according to the formula.

[0127] Step 6: Calculate the evaluation indices LEV and CEV for the luminance and chrominance components based on the global maximum JND value and the area ratio of the Mura defect region. Summarize the evaluation results of the multi-angle images and finally comprehensively evaluate the luminance and chrominance evaluation indices to obtain the final quantitative evaluation value MEV for the Mura defect of the display panel.

[0128] In this invention, preferably, in step one:

[0129] When shooting the display panel from multiple angles, the camera can generally take three angles: angle 1, where the camera is directly facing the center point of the screen and the lens axis coincides with the central axis of the screen; angle 2, where the camera is directly facing the center point of the screen, on the left side of the screen and at an angle of θ1 to the central axis of the screen; and angle 3, where the camera is directly facing the center point of the screen, on the right side of the screen and at an angle of θ2 to the central axis of the screen.

[0130] The specific conversion formula for converting an RGB color gamut image to a CIE-XYZ color gamut image is as follows:

[0131]

[0132]

[0133]

[0134] The original RGB values ​​of an image are typically read from 0 to 255. This invention first normalizes the RGB values ​​to reduce their range to 0-1. Noise denoising can be achieved by combining spatial domain Gaussian filtering and median filtering. The median filter performs convolutions of [5,5] kernel size on the three channels and expands the image by copying edge region values ​​to avoid the case where the values ​​at the four corners of the filtered image are 0. Then, the XYZ values ​​are calculated using the above formula and multiplied by 100 to reduce the range of XYZ values ​​to [0-100] for subsequent calculations.

[0135] Transform a CIE-XYZ color gamut image to a color space using the following transformation matrix:

[0136] wk = 0.279X + 0.72Y - 0.107Z

[0137] rg = -0.449X + 0.29Y - 0.077Z.

[0138] by=0.086X-0.59Y+0.501Z

[0139] In this invention, preferably, in step two:

[0140] The differential contrast map DC(x,y) is derived from the test contrast map C. Test(x,y) and reference contrast plot C Refence (x,y) is generated as follows:

[0141]

[0142] Where C Test (x,y) represents the original image converted to the color space (wk,rg,by), C Refence (x, y) is the background image after the original image has been filtered out of all mura defects by a Gaussian filter with a sufficiently low cutoff frequency, as shown in the following formula.

[0143]

[0144] In the above formula, x and y represent the row and column of the image;

[0145] r — visual resolution, unit: px / deg;

[0146] s — Gaussian scale, unit: deg, generally taken as 2 deg;

[0147] If mura defects exist at the edges of an image, traditional methods are prone to missing them. Therefore, an edge mask is needed to process the image. The specific algorithm is as follows:

[0148]

[0149] In the formula, M and N represent the image height and width, respectively, in pixels.

[0150] b gain — Boundary gain, which can generally be set to 1;

[0151] b scale — Boundary scale, unit: deg, generally 0.5 deg is acceptable;

[0152] By performing a dot product between each pixel of the differential contrast image DC(x,y) and the aforementioned edge mask function B(x,y), the accuracy of the algorithm for edge mura defect detection can be improved.

[0153] In this invention, preferably, in step three:

[0154] The formula for calculating the contrast sensitivity function (CSF) of the human visual system is shown below:

[0155]

[0156] in

[0157] ω=(u 2 +v 2 ) 1 / 2

[0158]

[0159] (u,v) is a spatial frequency coordinate, with the unit being each degree of period. This angle is the angle between the human eye and a point on the screen and the vertical line of the screen.

[0160] When processing the wk component, the values ​​in the contrast sensitivity function CSF can be taken as follows:

[0161]

[0162] γ(ω)=0.003100ω 2 -0.10680ω+1.396

[0163] m p =5.5σ p =1 / 40 ε=1

[0164] m c =5.5σ c =1 / 12 b ias =55

[0165] When processing the rg component, the values ​​in the contrast sensitivity function CSF can be taken as follows:

[0166]

[0167] γ(ω)=0.001531ω 2 -0.06149ω+1.140

[0168] m p =1.5σ p =1 / 70 ε=0

[0169] When processing the by component, the values ​​in the contrast sensitivity function CSF can be taken as follows:

[0170]

[0171] γ(ω)=0.001919ω 2 -0.06427ω+1.090

[0172] m p =1.5σ p =1 / 45 ε=1 / 4

[0173] m c =7.5σ c =1 / 20 b ias =0

[0174] In the above formula Let wk, rg, and by be the average values ​​of the three components of the color space;

[0175] Based on the above formula, the transfer function of the human visual system CSF model for the three components wk, rg, and by in the color space can be calculated. It's important to note that the spatial frequency ω in the above formula is in periods per degree, meaning it's given from the observer's perspective. However, the unit of the spectrum obtained through Fourier transform is periods per pixel. Therefore, the units are inconsistent, and the following formula is needed to convert the units when calculating the spatial frequency ω of the CSF:

[0176]

[0177] Where H R To measure the number of pixels horizontally on the screen in an image, V D W represents the vertical distance for visual observation. L Width of the display screen (mm);

[0178] In addition, in calculations involving calculations When there are three components, since the values ​​may be negative, it is necessary to take the modulus of the mean of the three components wk, rg, and by to ensure that the mean of the three components is always positive. This way, the data calculated by the CSF model can be consistent with the experimental results.

[0179] Then, the transfer function calculated using the CSF model is applied to perform frequency domain filtering on the three components (WK, RG, BY). The frequency domain filtering is completed by performing matrix multiplication between the spectrum of the three components and their corresponding CSF functions. It should be noted that the CSF function values ​​corresponding to the three components (WK, RG, BY) are relatively large. To prevent anomalies in subsequent calculations, this invention first synchronously normalizes the CSF function values ​​of the three components before frequency domain filtering. Specifically, the maximum value among the three CSF function values ​​is found, and each component is divided by this maximum value to ensure that the CSF function values ​​of the three components are always within [0,1], facilitating subsequent calculations. Finally, the frequency domain graphs of the three components (WK, RG, BY) after CSF function filtering are transformed into spatial domain components (wk, rg, by) using an inverse Fourier transform.

[0180] In this invention, preferably, in step four:

[0181] The transformation matrix for converting the color space (wk,rg,by) to the CIE-XYZ space is as follows:

[0182] X=0.626555wk-1.867178rg-0.153156by

[0183] Y=1.369855wk+0.934756rg+0.436229by

[0184] Z=1.505651wk+1.421324rg+2.536021by

[0185] Then, the image is transformed from the CIE-XYZ space and further transformed to the CIE-Lab space to obtain three components L,a,b;

[0186] L=116f(Y / Y n )-16

[0187] a = 500[f(X / X)] n )-f(Y / Y n )]

[0188] b = 200[f(Y / Y)] n )-f(Z / Z n )]

[0189] Where f(x) is the following piecewise function

[0190]

[0191] In the above formula, X n ,Y n Z n These are the most frequent values ​​of the X, Y, and Z components in the CIE-XYZ space, which can generally be obtained by statistically analyzing the rounded values ​​of each component.

[0192] The chromaticity component C can be calculated using the following formula based on a and b:

[0193] In this invention, preferably, in step five:

[0194] After obtaining the luminance component L and the chromaticity component C, the minimum perceptible difference (JND) maps for the luminance and chromaticity components are calculated separately according to Weber's theorem.

[0195]

[0196] Where Δ represents the luminance component L and the chromaticity component C, If it is a Gaussian window function, then β is the Minkowski distance coefficient, with a typical value of 2.4;

[0197] Based on the local JND diagram above, the global JND values ​​of the luminance and chrominance components can be further derived using the Minkowski function. This process can be called pooling, and its expression is:

[0198]

[0199] As γ→∞, the global maximum JND value for both the luminance and chrominance components can be obtained, i.e.

[0200] JND Δ_max =max[JND(x,y)]Δ∈{L,C};

[0201] Next, the ratio of the Mura defect area to the total screen area in the luminance and chrominance JND maps is calculated. The calculation method is as follows:

[0202]

[0203]

[0204] Where p(x,y) and q(x,y) are respectively

[0205]

[0206]

[0207] In this invention, preferably, in step six:

[0208] The luminance evaluation index LEV is LEV = JND. L_area1 +0.28JND L_max +6.47JND L_area2

[0209] Among them, JND L_area1 JND is the ratio of the Mura defect area with a brightness JND value greater than or equal to 1 to the total screen area. L_max JND is the global maximum value of the luminance JND map. L_area2 This represents the ratio of Mura defect areas with a brightness JND value greater than or equal to 2 to the total screen area.

[0210] The colorimetric evaluation index CEV is CEV = JND. C_area1 +0.16JND C_max +3.19JND C_area2

[0211] Among them, JND C_area1 JND is the ratio of Mura defect areas with a chromaticity JND value greater than or equal to 1 to the total screen area. C_max JND is the global maximum value of the chroma JND map. C_area2 This represents the ratio of Mura defect areas with a chromaticity JND value greater than or equal to 2 to the total screen area.

[0212]

[0213]

[0214] Where N is the total number of display panel images captured from multiple angles, LEV i ,i∈{1,2,3,…,N} represents the brightness evaluation index under multiple angles, CEV i ,i∈{1,2,3,…,N} are the colorimetric evaluation indexes under multiple angles, γ is the ratio between the two items, which can generally be taken as 0.6 here, and p is the Minkowski coefficient, which can generally be taken as 2;

[0215] The final Mura defect quantification value (MEV) for the display panel is calculated by comprehensively evaluating the brightness and chromaticity indicators using the following formula: MEV = 2.353 LEV. total +3.573CEV total ;

[0216] At this point, the quality of the display panel can be quantitatively evaluated based on the calculated Mura defect quantitative evaluation value MNEV.

[0217] This invention has implemented the corresponding method and system, and verified them through simulated Mura images of 20 different Mura types and Mura images acquired from actual display panels. The closer the correlation coefficient is to 1, the higher the consistency between the results of this patented method and subjective judgment. Figure 4 As shown, the overall average correlation coefficient between the evaluation results of this patent's evaluation method on the actual acquired display panel image and the subjective visual evaluation results is 0.934;

[0218] Figure 5 The correlation coefficients between the subjective judgments and algorithmic evaluations of the method in this patented patent for 20 major Mura defects show that the correlation coefficients are greater than 0.92 for 16 of the 20 Mura types, proving the feasibility of this method.

[0219] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for globally evaluating Mura defects of a display panel based on a just noticeable difference, characterized in that, The method comprises the following steps: Step one, the system first acquires complete images of the display panel from multiple angles through a high-definition camera, locates the screen area for cropping, scaling, and denoising filtering, then converts the preprocessed RGB color gamut image into a CIE-XYZ color gamut image, and further transforms the CIE-XYZ color gamut image into an opponent color space (wk, rg, by); Step two, the test contrast image and the reference contrast image are calculated, and a difference contrast image is generated from the two images, and an edge mask is introduced to process the image for the Mura defects existing in the image part; Step three, the three components of the opponent color space (wk, rg, by) are respectively subjected to Fourier transform into frequency components, a contrast sensitivity function (CSF) describing the human visual system is calculated, the single-channel contrast sensitivity function (CSF) of each component is applied to filter the three frequency components of wk / rg / by, and inverse Fourier transform is performed to the corresponding spatial domain; Step four, transform the inverse Fourier transformed spatial domain (wk, rg, by) image to CIE-XYZ space and further to CIE-Lab space, calculate the luminance component under CIE-Lab color gamut and chrominance components ; Step 5: Calculate the luminance components separately. and chromaticity components The minimum perceptible difference (JND) map is used to calculate the global maximum JND values ​​of the luminance and chrominance components and the ratio of the Mura defect area to the total screen area using the formula; in step five: The luminance component and the chrominance component After that, the minimum perceptible difference map (JND) of the luminance and chrominance components are calculated respectively according to Weber's theorem ； wherein represents a luminance component and a chrominance component , is a Gaussian window function, then is a Minkowski distance coefficient, typically with a value of 2.4; According to the above local JND image, the global JND values of the luminance component and the chrominance component can be further induced using the Minkowski function, which can be called pooling, so the expression is: ； When the global maximum JND value for both luminance and chrominance components is obtained, i.e. ； Next, the ratio of the Mura defect area of the luminance and chrominance JND images to the total screen area is calculated, and the calculation method is as follows ； wherein and are respectively ； Step six, calculating the evaluation index of the luminance component and the chrominance component according to the global maximum JND value and the area proportion of the Mura defect region , and the evaluation results of the multi-angle shooting images are summarized, and finally the final display panel Mura defect quantitative evaluation value is obtained by comprehensively judging the luminance and chrominance evaluation indexes , — rows and columns of the image; In the step six: Brightness evaluation index To ; wherein, is the ratio of the area of the Mura defect region with the luminance JND map value greater than or equal to 1 to the total screen area region, is the global maximum value of the luminance JND map, is the ratio of the area of the Mura defect region with the luminance JND map value greater than or equal to 2 to the total screen area region; Colorimetric assessment index To ; wherein, is the ratio of the area of the Mura defect regions with chroma JND map values greater than or equal to 1 to the total screen area region, is the global maximum of the chroma JND map, is the ratio of the area of the Mura defect regions with chroma JND map values greater than or equal to 2 to the total screen area region; ； wherein is the total number of display panel images taken at multiple angles, is the luminance evaluation index at multiple angles, is the chrominance evaluation index at multiple angles, is the ratio between the two, taken as 0.6 here, is the Minkowski coefficient, taken as 2; - image height and width in pix; The final display panel Mura defect quantitative evaluation value is obtained by comprehensively calculating the brightness and chrominance evaluation indexes according to the following formula To ; At this point, the Mura defect quantification evaluation value calculated The degree of good or bad of the display panel is quantitatively evaluated. 2.The method of claim 1, wherein the method further comprises: determining a minimum perceptible difference (MPD) of the display panel; and determining a Mura defect of the display panel based on the MPD. In the step one: Camera multi-angle shooting display panel takes 3 angles, the camera is opposite to the center point of the screen, the lens axis coincides with the screen central axis, the view angle 1; the camera is opposite to the center point of the screen, the view angle 2 is in the left side of the screen and the screen central axis forms an angle; the camera is opposite to the center point of the screen, the view angle 3 is in the right side of the screen and the screen central axis forms an angle. Camera multi-angle shooting display panel takes 3 angles, the camera is opposite to the center point of the screen, the lens axis coincides with the screen central axis, the view angle 1; the camera is opposite to the center point of the screen, the view angle 2 is in the left side of the screen and the screen central axis forms an angle; the camera is opposite to the center point of the screen, the view angle 3 is in the right side of the screen and the screen central axis forms an angle. Camera multi-angle shooting display panel takes 3 angles, the camera is The specific conversion formula for converting the RGB color gamut image into the CIE-XYZ color gamut image is as follows: ； Wherein, the range of the original RGB value read to the image is usually 0-255, first, the value of RGB is normalized to change the range to 0-1, the denoising filtering can be completed by using the joint Gaussian filter and median filter in the spatial domain, the median filter is performed on the three channels with a [5, 5] convolution kernel size, and the image is expanded by copying the edge region value to avoid the situation that the value of the four corners of the image after filtering is 0; then the XYZ value is calculated by using the above formula, and then multiplied by 100 to change the range of the XYZ value to [0-100] for subsequent calculation; The CIE-XYZ color gamut image is transformed into the opponent color space according to the following conversion matrix: 。 3. The global evaluation method for display panel Mura defects based on minimum perceptible difference according to claim 1, characterized in that: In the step two: difference contrast map from the test contrast map and the reference contrast map is generated, specifically as follows: ； wherein is the background image of the original image converted to the color space (wk, rg, by), is the background image of the original image after filtering out all mura defects by a Gaussian filter with a sufficiently low cut-off frequency, as follows: ； - visual resolution in px / deg; - Gaussian scale, in deg, taken as 2 deg; If there is a Mura defect in the edge part of the image, the traditional method is easy to miss detection, therefore, an edge mask is used to process the image, and the specific algorithm is as follows: ； In the formula, - a boundary gain, taking the value 1 ; - Border Proportion, unit: deg, take 0.5 deg; Each pixel of the differential contrast map is multiplied by the edge mask function described above, which improves the accuracy of the algorithm for detecting edge mura defects.

4. The method of claim 1, wherein the method is a method for globally evaluating Mura defects of a display panel based on a just noticeable difference. In the step three: The formula for calculating the contrast sensitivity function (CSF) of the human visual system is as follows: ； Wherein: ； is the spatial frequency coordinate in cycles per degree, which is the angle between the line from the eye to the point on the screen and the normal to the screen; When processing the wk component, the values in the contrast sensitivity function (CSF) can be taken as: ； When processing the rg component, the values in the contrast sensitivity function (CSF) can be taken as: ； When processing the by component, the values in the contrast sensitivity function (CSF) can be taken as: ； In the above formulae For the color space The mean value of each of the three components; According to the above equation, the color space CSF model of the human visual system for three components; spatial frequency involved in the above equation The unit is cycle / degree, that is, the equation is given from the perspective of the observer's angle, while the unit of the spectrum obtained by Fourier transform is cycle / pixel, so the units are inconsistent, and the following equation needs to be used to convert the unit when calculating the spatial frequency of the CSF: The unit is cycle / degree, that is, the equation is given from the perspective of the observer's angle, while the unit of the spectrum obtained by Fourier transform is cycle / pixel, so the units are inconsistent, and the following equation needs to be used to convert the unit when calculating the spatial frequency of the CSF: ； wherein is the number of pixels in the image at the screen level, is the visual vertical distance, is the width of the display screen (mm); In addition, in calculations involving calculations When there are three components, since the number may be negative, it is necessary to adjust... The mean of the three components is moduloed to ensure that the mean of the three components is always positive, so that the data calculated by the CSF model can be consistent with the experimental results. Then, the transfer function calculated by the CSF model is applied to the frequency domain filtering of the three components (WK, RG, BY), and the spectral graph of the three components (WK, RG, BY) is multiplied by the corresponding CSF function to complete the frequency domain filtering; it should be noted that the CSF function values of the three components (WK, RG, BY) are large to prevent abnormal calculation in the subsequent calculation, and the CSF function values of the three components are normalized before the frequency domain filtering, that is, the maximum value in the CSF function values of the three components is found, and the maximum value is divided by the maximum value respectively, so that the CSF function values of the three components are always within [0, 1], so as to facilitate the subsequent calculation; finally, the three components (WK, RG, BY) in the frequency domain filtered by the CSF function are inversely Fourier transformed into the spatial domain components (wk, rg, by).

5. The method of claim 1, wherein the method is a method for globally evaluating Mura defects of a display panel based on a just noticeable difference. In the step four: The conversion matrix for converting the color space (wk, rg, by) into the CIE-XYZ space is as follows: ； The image is then transformed from CIE-XYZ space and further to CIE-Lab space, resulting in three components ; ； wherein is the following piecewise function ； In the above formula respectively, in the CIE-XYZ space The value of the highest frequency of the component, and the value of the integer of each component is counted. By The chrominance components can be calculated according to the formula i.e. .

Images