WO2019242294A1 - 基于独立成分自适应选择的Mura侦测方法 - Google Patents

基于独立成分自适应选择的Mura侦测方法 Download PDF

Info

Publication number
WO2019242294A1
WO2019242294A1 PCT/CN2019/070008 CN2019070008W WO2019242294A1 WO 2019242294 A1 WO2019242294 A1 WO 2019242294A1 CN 2019070008 W CN2019070008 W CN 2019070008W WO 2019242294 A1 WO2019242294 A1 WO 2019242294A1
Authority
WO
WIPO (PCT)
Prior art keywords
independent component
pixels
value
background range
brightness
Prior art date
Application number
PCT/CN2019/070008
Other languages
English (en)
French (fr)
Inventor
史超超
Original Assignee
深圳市华星光电半导体显示技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市华星光电半导体显示技术有限公司 filed Critical 深圳市华星光电半导体显示技术有限公司
Publication of WO2019242294A1 publication Critical patent/WO2019242294A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/0008—Industrial image inspection checking presence/absence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/90—Dynamic range modification of images or parts thereof
    • G06T5/94—Dynamic range modification of images or parts thereof based on local image properties, e.g. for local contrast enhancement
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30121—CRT, LCD or plasma display

Definitions

  • the present invention relates to the field of display technology, and in particular, to a Mura detection method based on independent component adaptive selection.
  • liquid crystal displays Liquid Crystal Display, LCD
  • organic light emitting diode display devices Organic Light Emitting Display, OLED
  • other flat display devices due to high picture quality, power saving, thin body and wide application range
  • LCD Liquid Crystal Display
  • OLED Organic Light Emitting Display
  • Mura refers to the phenomenon of various traces caused by uneven brightness of the display panel.
  • Various traces may be horizontal stripes or forty-five-degree angle stripes, may be very straight cut squares, may appear in a corner, or there may be no marks at all, usually we will appear this kind of various
  • the display area of the trace is called a Mura area.
  • the presence of Mura will not affect the use of the display panel, but it will reduce the viewing comfort of the user, so Mura restricts the development of LCD displays and OLED displays. Because the background of the displayed image is complex, the Mura area has low contrast compared to the background and there are no obvious boundaries. It is difficult to quantify the Mura area.
  • the method of Independent Component Analysis includes: combining multiple input images into a mixing matrix and performing the mixing matrix. ICA transform to obtain multiple independent components, and select the independent component closest to the original image as the target independent component to perform inverse ICA transformation to obtain a defect (Mura) enhanced image.
  • the process of selecting the independent component closest to the original image requires The human eye judges and chooses, it is not convenient to actively and adaptively select the closest independent component, which cannot meet the needs of production automation.
  • the purpose of the present invention is to provide a Mura detection method based on independent component adaptive selection, which can adaptively select target independent components, improve the Mura detection method, and meet the needs of production automation.
  • the present invention provides a method for Mura detection based on independent component adaptive selection, including the following steps:
  • Step S1 Convert N input images into one-dimensional vectors and combine them into a mixing matrix, where N is an integer greater than 1.
  • Step S2 ICA transform the mixed matrix to obtain N independent components
  • Step S3 Select one of the N input images as a comparison image, and calculate the SSIM value between each independent component and the comparison image;
  • Step S4 Set the background range of each independent component and count the number of brightness extreme points in each independent component, where the brightness extreme points are the number of pixels in the independent component whose grayscale values are outside their background range;
  • Step S5 Calculate the comparison value of each independent component according to the SSIM value of each independent component and the number of brightness extreme points of each independent component, and select the independent component with the largest comparison value as the target independent component;
  • Step S6 Perform inverse ICA transformation on the target independent components to obtain a defect enhanced image, and select a defect threshold to perform defect segmentation on the defect enhanced image.
  • the independent component to be calculated is defined as image X and the comparison image is image Y.
  • the formula for calculating the SSIM value between the independent component and the comparison image in step S3 is:
  • ⁇ x is the average value of the grayscale values of all pixels in the independent component
  • ⁇ y is the average value of the grayscale values of all pixels in the comparison image
  • ⁇ x is the variance of the grayscale values of all pixels in the independent component
  • ⁇ y is the covariance of the grayscale values of all pixels in the compared image and the grayscale values of all pixels in the independent component.
  • C1, C2, and C3 are the brightness constant and contrast, respectively.
  • Constants and structural constants, L, C and S are the brightness comparison value, contrast comparison value and structure comparison value, respectively
  • SSIM is the SSIM value between the independent component and the comparison image;
  • the step S4 specifically includes:
  • the number of brightness extreme points in the independent component is counted according to the background range of each row of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to be ⁇ '- ⁇ 'to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average and variance of the grayscale values of the column of pixels in the independent component;
  • the number of luminance extreme points in the independent component is counted according to the background range of each column of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to be ⁇ '- ⁇ 'to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average and variance of the grayscale values of the column of pixels in the independent component;
  • the step S4 specifically includes:
  • the number of brightness extreme points in the independent component is counted according to the background range of each row of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to ( ⁇ '+ g min ) / 2 to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average of the grayscale values of the pixels in the column in the independent component.
  • Value and variance, g min is the minimum value of the grayscale value of all pixels in the independent component;
  • the number of luminance extreme points in the independent component is counted according to the background range of each column of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to ( ⁇ '+ g min ) / 2 to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average of the grayscale values of the pixels in the column in the independent component. Values and variances;
  • step S1 the three input images are respectively converted into one-dimensional vectors, and the three input images are respectively the image with the highest brightness, the image with 50% brightness, and the image with the lowest brightness. .
  • w is equal to 0 and 0.1 in step S5
  • w is equal to 1
  • w is equal to 3
  • the present invention provides a Mura detection method based on independent component adaptive selection, including the following steps:
  • Step S1 Convert N input images into one-dimensional vectors and combine them into a mixing matrix, where N is an integer greater than 1.
  • Step S2 ICA transform the mixed matrix to obtain N independent components
  • Step S3 Select one of the N input images as a comparison image, and calculate the SSIM value between each independent component and the comparison image;
  • Step S4 Set the background range of each independent component and count the number of brightness extreme points in each independent component, where the brightness extreme points are the number of pixels in the independent component whose grayscale values are outside their background range;
  • Step S5 Calculate the comparison value of each independent component according to the SSIM value of each independent component and the number of brightness extreme points of each independent component, and select the independent component with the largest comparison value as the target independent component;
  • Step S6 Perform inverse ICA transformation on the target independent components to obtain a defect enhanced image, and select a defect threshold to perform defect segmentation on the defect enhanced image.
  • the independent component to be calculated is defined as image X and the comparison image is image Y.
  • the formula for calculating the SSIM value between the independent component and the comparison image in step S3 is:
  • ⁇ x is the average value of the grayscale values of all pixels in the independent component
  • ⁇ y is the average value of the grayscale values of all pixels in the comparison image
  • ⁇ x is the variance of the grayscale values of all pixels in the independent component
  • ⁇ y is the covariance of the grayscale values of all pixels in the compared image and the grayscale values of all pixels in the independent component.
  • C1, C2, and C3 are the brightness constant and contrast, respectively.
  • Constants and structure constants, L, C, and S are brightness comparison value, contrast comparison value, and structure comparison value, respectively.
  • SSIM is the SSIM value between the independent component and the comparison image.
  • the step S4 specifically includes:
  • the number of brightness extreme points in the independent component is counted according to the background range of each row of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to be ⁇ '- ⁇ 'to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average and variance of the grayscale values of the column of pixels in the independent component;
  • the number of luminance extreme points in the independent component is counted according to the background range of each column of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to be ⁇ '- ⁇ 'to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average and variance of the grayscale values of the column of pixels in the independent component;
  • the step S4 specifically includes:
  • the number of brightness extreme points in the independent component is counted according to the background range of each row of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to ( ⁇ '+ g min ) / 2 to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average of the grayscale values of the pixels in the column in the independent component.
  • Value and variance, g min is the minimum value of the grayscale value of all pixels in the independent component;
  • the number of luminance extreme points in the independent component is counted according to the background range of each column of pixels in the independent component.
  • the step S4 specifically includes:
  • the background range of each column of pixels in the independent component is set to ( ⁇ '+ g min ) / 2 to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average of the grayscale values of the pixels in the column in the independent component. Values and variances;
  • step S1 the three input images are respectively converted into one-dimensional vectors, and the three input images are respectively the image with the highest brightness, the image with 50% brightness, and the image with the lowest brightness. .
  • w is equal to 0 and 0.1 in step S5
  • w is equal to 1
  • w is equal to 3
  • the present invention provides a method for Mura detection based on independent component adaptive selection, including the following steps: combining N input images into a hybrid matrix respectively; performing ICA transformation on the hybrid matrix to obtain N independent Component; one of the N input images is selected as a comparison image, and the SSIM value between each independent component and the comparison image is calculated separately; the background range of each independent component is set and the brightness in each independent component is counted The number of extreme points; the comparison value of each independent component is calculated based on the SSIM value of each independent component and the number of brightness extreme points of each independent component, and the independent component with the largest comparison value is selected as the target independent component; the ICA inverse of the target independent component is performed Transform to obtain a defect enhanced image, and select a defect threshold to perform defect segmentation on the defect enhanced image. Calculate the SSIM value of each independent component and count the number of extreme brightness points in each independent component, and select the target independent component accordingly. Can replace the human eye's adaptive selection of target independent components and improve Mura
  • FIG. 1 is a flowchart of a Mura detection method based on independent component adaptive selection according to the present invention
  • FIG. 2 is a schematic diagram of steps S1 and S2 of a first embodiment of a Mura detection method based on independent component adaptive selection according to the present invention
  • FIG. 3 is a schematic diagram of steps S1 and S2 of a second embodiment of the Mura detection method based on independent component adaptive selection of the present invention.
  • the present invention provides a Mura detection method based on independent component adaptive selection, including the following steps:
  • Step S1 N input images are respectively converted into one-dimensional vectors and combined into a mixing matrix, where N is an integer greater than 1.
  • N in the step S1 is equal to 3, that is, the step S1 converts three input images into one-dimensional vectors, respectively, and forms a mixing matrix, where the 3
  • the input images are the highest brightness image, the lowest brightness image, and the 50% brightness image.
  • the images A, B, and C shown in FIG. 2 are respectively, the images A, B, and C are 2 rows and 2 columns, and the images A, B, and C are converted into a one-dimensional vector A of 1 row and 4 columns, respectively. ', B', and C ', and mixing the one-dimensional vectors A', B ', and C' to obtain a two-row, four-column mixing matrix D.
  • N in the step S1 is equal to 2, that is, the step S1 converts the two input images into one-dimensional vectors and forms a mixture matrix, where the 2
  • the input images are the highest brightness image and the lowest brightness image, as shown in FIG. 3.
  • the highest brightness image and the lowest brightness image in step S1 are the images A1 and B1 shown in FIG. 3.
  • the images A1 and B1 are both 2 rows and 2 columns, and the images A1 and B1 are respectively converted into one-dimensional and four-dimensional one-dimensional vectors A1 'and B1', and the one-dimensional vectors A1 'and B1' are mixed A two-row, four-column mixed matrix D1 is obtained.
  • Step S2 Perform ICA transformation on the mixed matrix to obtain N independent components.
  • the process of ICA transformation includes iteratively processing the mixed matrix to maximize the difference of each row in the mixed matrix, and then restore the mixed matrix to independent components in the same format as the input image. For example, when the input image is 2 rows and 2 columns , The independent components are also 2 rows and 2 columns.
  • the ICA transformation process includes first performing an iterative process on the mixing matrix D to maximize the difference in each row in the mixing matrix to obtain the mixing matrix D ′, Then the mixing matrix is restored to three independent components E, F and G in the same format as the input image.
  • One of the E, F and G is the target independent component to be identified, and the other two independent components are non-target independent components.
  • the non-target independent component may be a noise component or a moiré component.
  • the ICA transformation process includes first performing an iterative process on the mixing matrix D1 so as to maximize the difference of each row in the mixing matrix to obtain the mixing matrix D1 ', Then the mixture matrix is restored to two independent components E1 and F1 in the same format as the input image.
  • One of the E1 and F1 is the target independent component to be identified, and the other is a non-target independent component.
  • the non-target independent component can be It is a noise component or a moiré component.
  • step S3 one of the N input images is selected as a comparison image, and an SSIM value between each independent component and the comparison image is calculated.
  • the structural similarity (Similarity index) value ranges from 0 to 1.
  • the larger the SSIM value the smaller the image distortion, that is, the more it meets the requirements of the target independent component.
  • the calculation measures the similarity of the image from three aspects of brightness, contrast, and structure, and defines the independent component to be calculated as image X and the comparison image as image Y.
  • the specific calculation formula is:
  • ⁇ x is the average value of the grayscale values of all pixels in the independent component
  • ⁇ y is the average value of the grayscale values of all pixels in the comparison image
  • ⁇ x is the variance of the grayscale values of all pixels in the independent component
  • ⁇ y is the covariance of the grayscale values of all pixels in the compared image and the grayscale values of all pixels in the independent component.
  • C1, C2, and C3 are the brightness constant and contrast, respectively.
  • Constants and structural constants, L, C, and S are the brightness comparison value, contrast comparison value, and structure comparison value, respectively
  • SSIM is the SSIM value between the independent component and the comparison image.
  • the comparison image may be selected from N input images as required.
  • step S3 the image A with the highest brightness is selected as a comparison image to calculate the SSIM value.
  • step S3 specifically includes: calculating an independent component E and the image A.
  • the SSIM value between the independent component F and the image A is calculated, and the SSIM value between the independent component G and the image A is calculated.
  • the image B with the lowest brightness may also be selected as the comparative image for the calculation of the SSIM value.
  • step S3 corresponds to calculating the independent component E and the image B.
  • the SSIM value between the independent component F and the image B is calculated, and the SSIM value between the independent component G and the image B is calculated.
  • step S3 corresponds to the calculation between the independent component E and the image C.
  • SSIM value calculating the SSIM value between the independent component F and the image C, and calculating the SSIM value between the independent component G and the image C.
  • the step S3 in the second embodiment of the present invention, the image A1 with the highest brightness is selected as the comparison image to perform the calculation of the SSIM value. Then, the step S3 specifically includes: calculating the difference between the independent component E1 and the image A1. SSIM value, calculating the SSIM value between the independent component F1 and the image A1. When necessary, in the first embodiment of the present invention, the image B1 with the lowest brightness may also be selected as the comparative image for calculation of the SSIM value. At this time, step S3 corresponds to calculating the independent component E1 and the image B1. The SSIM value between the independent components F1 and the image B1.
  • Step S4 Set the background range of each independent component and count the number of brightness extreme points in each independent component.
  • the brightness extreme point is the number of pixels in the independent component whose grayscale value is outside its background range.
  • the brightness extreme point is actually a defect area in the independent component, that is, the Mura area.
  • the setting of the background range of each independent component and counting the number of brightness extreme points in each independent component can have a variety of different method.
  • the step S4 specifically includes: separately setting the background range of each row of pixels in the independent component to be ⁇ - ⁇ to ⁇ + ⁇ , where ⁇ and ⁇ are respectively the The average and variance of the grayscale values of the row pixels; the number of brightness extreme points in the independent component is counted according to the background range of each row of pixels in the independent component.
  • the step S4 specifically includes: setting the background range of each column of pixels in the independent component to be ⁇ '- ⁇ 'to ⁇ ' + ⁇ ', wherein ⁇ ' and ⁇ 'are respectively The average value and variance of the grayscale values of the pixels in the column in the independent component; the number of luminance extreme points in the independent component is counted according to the background range of the pixels in each column in the independent component.
  • the step S4 specifically includes: setting the background range of each row of pixels in the independent component to be ⁇ - ⁇ to ⁇ + ⁇ , respectively.
  • ⁇ and ⁇ are the average and variance of the grayscale values of the pixels in the row of the independent component respectively; the number of brightness extreme points in the independent component is counted according to the background range of each row of pixels in the independent component; the independent components are set in the independent component respectively
  • the background range of the pixels in each column is ⁇ '- ⁇ 'to ⁇ ' + ⁇ ', where ⁇ ' and ⁇ 'are the average and variance of the grayscale values of the pixels in the column in the independent component respectively; according to each column in the independent component
  • the background range of pixels again counts the number of extreme brightness points in the independent component; compares the number of extreme brightness points based on the statistics of the background range of each row of pixels in the independent component with the statistics of the background range of each column of pixels in the independent component The obtained
  • the background range of each row of pixels in the independent component can be set to ( ⁇ + g min ) / 2 to ⁇ + ⁇ , where ⁇ and ⁇ are the pixels of the row in the independent component.
  • the average and variance of the grayscale values, g min is the minimum value of the grayscale values of all pixels in the independent component
  • the background range of each column of pixels is set to ( ⁇ '+ g min ) / 2 to ⁇ ' + ⁇ ',
  • ⁇ ' and ⁇ ' are the average and variance of the grayscale values of the pixels in the column in the independent component
  • g min is the minimum value of the grayscale values of all pixels in the independent component, to better distinguish low Gray scale Mura.
  • Step S5 Calculate the comparison value of each independent component according to the SSIM value of each independent component and the number of brightness extreme points of each independent component, and select the independent component with the largest comparison value as the target independent component;
  • step S5 when the SSIM value of the independent component is between 0 and 0.1, w is equal to 0, and when the SSIM value of the independent component is between 0.1 and 0.5, w is equal to 1, and when the SSIM value of the independent component is When greater than 0.5, w is equal to 3.
  • Step S6 Perform inverse ICA transformation on the target independent components to obtain a defect enhanced image, and select a defect threshold to perform defect segmentation on the defect enhanced image.
  • the process of selecting the defect threshold may be performed by referring to the method of setting a background region in step S4.
  • the present invention provides a method for Mura detection based on independent component adaptive selection, including the following steps: combining N input images into a mixture matrix respectively; performing ICA transformation on the mixture matrix to obtain N independent components ; Select one of the N input images as a comparison image, and calculate the SSIM value between each independent component and the comparison image; set the background range of each independent component and count the brightness of each independent component Number of value points; calculate the comparison value of each independent component according to the SSIM value of each independent component and the number of brightness extreme points of each independent component, and select the independent component with the largest comparison value as the target independent component; perform inverse ICA transformation on the target independent component To obtain a defect-enhanced image and select a defect threshold to perform defect segmentation on the defect-enhanced image.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

一种基于独立成分自适应选择的Mura侦测方法。该方法包括如下步骤：将N张输入图像分别组合成一个混合矩阵；获得N个独立成分；分别计算每一个独立成分与所述比较图像之间的SSIM值；设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量；计算各个独立成分的比较值，选择比较值最大的独立成分作为目标独立成分；对目标独立成分进行ICA逆变换。

Description

基于独立成分自适应选择的Mura侦测方法 技术领域
本发明涉及显示技术领域，尤其涉及一种基于独立成分自适应选择的Mura侦测方法。
背景技术
随着显示技术的发展，液晶显示器(Liquid Crystal Display，LCD)及有机发光二极管显示器件(Organic Light Emitting Display，OLED)等平面显示装置因具有高画质、省电、机身薄及应用范围广等优点，而被广泛的应用于手机、电视、个人数字助理、数字相机、笔记本电脑、台式计算机等各种消费性电子产品，成为显示装置中的主流。
技术问题
随着科技的发展及人们物质生活的需求，现金平面显示器的尺寸做得越来越大，显示分辨率也越来越高，对生产工艺的要求也越来越严苛。目前在显示面板生产过程中由于生产工艺等原因经常会产生Mura，所谓Mura是指因显示面板亮度不均匀造成各种痕迹的现象。通过在暗室中将显示面板切换到黑色画面以及其他低灰阶画面，然后从各种不同角度去看显示面板是否存在痕迹即可判断该显示面板是否存在Mura。各种痕迹可能是横向条纹或四十 五度角条纹，可能是切得很直的方块，可能是某个角落出现一块，也可能是没有规则可言的痕迹，通常我们将这种出现各种痕迹的显示区域称为Mura区域。
Mura的存在不会对显示面板的使用功能造成影响，但是会降低用户的观看舒适度，因此Mura制约了LCD显示器与OLED显示器的发展。由于显示图像背景复杂，Mura区域相对于背景具有低对比度且无明显边界，很难对Mura区域进行量化处理。现有技术中常常通过独立成分分析(Independent Component Correlation Algorithm，ICA)的方法来增强Mura区域与背景区域的对比度，一般独立成分分析的过程包括：将多张输入图像组成混合矩阵并对混合矩阵进行ICA变换，得到多个独立成分，并选定最接近原图的独立成分作为目标独立成分进行ICA逆变换，得到缺陷(Mura)增强的图像，其中选定最接近原图的独立成分的过程需要人眼判断选择，不便于主动自适应选择最接近的独立成分，不能满足生产自动化的需要。
技术解决方案
本发明的目的在于提供一种基于独立成分自适应选择的Mura侦测方法，能够自适应选择目标独立成分，改善Mura侦测方法，满足生产自动化的需要。
为实现上述目的，本发明提供了一种基于独立成分自 适应选择的Mura侦测方法，包括如下步骤：
步骤S1、将N张输入图像分别转换为一维向量，并组合成一个混合矩阵，N为大于1的整数；
步骤S2、对混合矩阵进行ICA变换，获得N个独立成分；
步骤S3、选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值；
步骤S4、设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量，所述亮度极值点为独立成分中灰阶值位于其背景范围外的像素的数量；
步骤S5、根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选择比较值最大的独立成分作为目标独立成分；
所述比较值的计算公式为：Q＝SSIM+w/M，其中Q为独立成分的比较值，SSIM为独立成分的SSIM值，w为与SSIM值相关的系数，M为独立成分的亮度极值点数量；
步骤S6、对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷分割。
根据本发明的Mura侦测方法，定义待计算的独立成分为图像X、比较图像为图像Y，所述步骤S3中计算独立成分与比较图像之间的SSIM值的公式为：
SSIM＝L×C×S；
其中，
Figure PCTCN2019070008-appb-000001
其中，μ x为该独立成分中所有像素的灰阶值的均值，μy为比较图像中所有像素的灰阶值的均值，σ x为该独立成分中所有像素的灰阶值的方差，σ y为比较图像中所有像素的灰阶值的方差，σ xy为比较图像中所有像素的灰阶值与独立成分中所有像素的灰阶值的协方差，C1、C2及C3分别为亮度常数、对比度常数及结构常数，L、C及S分别为亮度比较值、对比度比较值及结构比较值，SSIM为独立成分与比较图像之间的SSIM值；
所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
依据独立成分中每一列像素的背景范围的再次统计独立成分中亮度极值点数量；
比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与该依据独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
依据独立成分中每一列像素的背景范围的再次统计独立成分中亮度极值点数量；
比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与该依据独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的 较大值作为最终的亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S1中将3张输入图像分别转换为一维向量，所述3张输入图像分别为亮度最高的图像、亮度为50％的图像及亮度最低的图像。
根据本发明的Mura侦测方法，所述步骤S5中当独立成分的SSIM值位于0～0.1之间时，w等于0，当独立成分的SSIM值位于0.1～0.5之间时，w等于1，当独立成分的SSIM值大于0.5时，w等于3。
为实现上述目的，本发明提供了一种基于独立成分自适应选择的Mura侦测方法，包括如下步骤：
步骤S1、将N张输入图像分别转换为一维向量，并组合成一个混合矩阵，N为大于1的整数；
步骤S2、对混合矩阵进行ICA变换，获得N个独立成分；
步骤S3、选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值；
步骤S4、设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量，所述亮度极值点为独立成分中灰阶值位于其背景范围外的像素的数量；
步骤S5、根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选择比较值最大的独立成分作为目标独立成分；
所述比较值的计算公式为：Q＝SSIM+w/M，其中Q为独立成分的比较值，SSIM为独立成分的SSIM值，w为与SSIM值相关的系数，M为独立成分的亮度极值点数量；
步骤S6、对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷分割。
根据本发明的Mura侦测方法，定义待计算的独立成分为图像X、比较图像为图像Y，所述步骤S3中计算独立成分与比较图像之间的SSIM值的公式为：
SSIM＝L×C×S；
其中，
Figure PCTCN2019070008-appb-000002
其中，μ x为该独立成分中所有像素的灰阶值的均值，μy为比较图像中所有像素的灰阶值的均值，σ x为该独立成分中所有像素的灰阶值的方差，σ y为比较图像中所有像素的灰阶值的方差，σ xy为比较图像中所有像素的灰阶值与独立成分中所有像素的灰阶值的协方差，C1、C2及C3分别为亮度常数、对比度常数及结构常数，L、C及S分别为亮度比较值、对比度比较值及结构比较值，SSIM为独立成分与比较图像之间的SSIM值。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰 阶值的平均值和方差；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
依据独立成分中每一列像素的背景范围的再次统计独立成分中亮度极值点数量；
比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与该依据独立成分中每一列像素 的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S4中具体包括：
分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
依据独立成分中每一行像素的背景范围统计独立成分 中亮度极值点数量；
分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
依据独立成分中每一列像素的背景范围的再次统计独立成分中亮度极值点数量；
比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与该依据独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
根据本发明的Mura侦测方法，所述步骤S1中将3张输入图像分别转换为一维向量，所述3张输入图像分别为亮度最高的图像、亮度为50％的图像及亮度最低的图像。
根据本发明的Mura侦测方法，所述步骤S5中当独立成分的SSIM值位于0～0.1之间时，w等于0，当独立成分的SSIM值位于0.1～0.5之间时，w等于1，当独立成分的SSIM值大于0.5时，w等于3。
有益效果
本发明的有益效果：本发明提供一种基于独立成分自适应选择的Mura侦测方法，包括如下步骤：将N张输入图像分别组合成一个混合矩阵；对混合矩阵进行ICA变换， 获得N个独立成分；选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值；设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量；根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选择比较值最大的独立成分作为目标独立成分；对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷分割，通过分别计算各个独立成分的SSIM值及统计各个独立成分中亮度极值点数量，并据此选择目标独立成分，能够取代人眼自适应选择目标独立成分，改善Mura侦测方法，满足生产自动化的需要。
附图说明
为了能更进一步了解本发明的特征以及技术内容，请参阅以下有关本发明的详细说明与附图，然而附图仅提供参考与说明用，并非用来对本发明加以限制。
附图中，
图1为本发明的基于独立成分自适应选择的Mura侦测方法的流程图；
图2为本发明的基于独立成分自适应选择的Mura侦测方法的第一实施例的步骤S1和步骤S2的示意图；
图3为本发明的基于独立成分自适应选择的Mura侦测方法的第二实施例的步骤S1和步骤S2的示意图。
本发明的实施方式
为更进一步阐述本发明所采取的技术手段及其效果，以下结合本发明的优选实施例及其附图进行详细描述。
请参阅图1，本发明提供一种基于独立成分自适应选择的Mura侦测方法，包括如下步骤：
步骤S1、将N张输入图像分别转换为一维向量，并组合成一个混合矩阵，N为大于1的整数。
具体地，在本发明的第一实施例中，所述步骤S1中的N等于3，即所述步骤S1将3张输入图像分别转换为一维向量，并组成一个混合矩阵，其中所述3张输入图像分别为亮度最高的图像、亮度最低的图像及亮度为50％的图像，如图2所示，所述步骤S1所述亮度最高的图像、亮度最低的图像及亮度为50％的图像分别为图2所示的图像A、B及C，所述图像A、B及C均为2行2列，将所述图像A、B及C分别转换为1行4列的一维向量A’、B’及C’，并将所述一维向量A’、B’及C’进行混合得到2行4列混合矩阵D。
具体地，在本发明的第二实施例中，所述步骤S1中的N等于2，即所述步骤S1将2张输入图像分别转换为 一维向量，并组成一个混合矩阵，其中所述2张输入图像分别为亮度最高的图像、亮度最低的图像，如图3所示，所述步骤S1所述亮度最高的图像、及亮度最低的图像分别为图3所示的图像A1和B1，所述图像A1和B1均为2行2列，将所述图像A1和B1分别转换为1行4列的一维向量A1’和B1’，并将所述一维向量A1’和B1’进行混合得到2行4列混合矩阵D1。
当然，在本发明的其他实施例中也可以输入其他数量及亮度的图像，这并不是对本发明的限制。
步骤S2、对混合矩阵进行ICA变换，获得N个独立成分。
具体地，ICA变换的过程包括先对混合矩阵进行迭代处理使得混合矩阵中每行差异最大化，然后再将混合矩阵还原为与输入图像相同格式的独立成分，例如输入图像为2行2列时，则独立成分也为2行2列。
具体地，如图2所示，在本发明的第一实施例中，所述ICA变换的过程包括先对混合矩阵D进行迭代处理使得混合矩阵中每行差异最大化，得到混合矩阵D’，接着混合矩阵还原为与输入图像相同格式的三个独立成分E、F和G，所述E、F和G中的一个为待识别的目标独立成分，另外两个独立成分为非目标独立成分，所述非目标独立成分可以为噪声成分或摩尔纹成分。
具体地，如图3所示，在本发明的第二实施例中，所述ICA变换的过程包括先对混合矩阵D1进行迭代处理使得混合矩阵中每行差异最大化，得到混合矩阵D1’，接着混合矩阵还原为与输入图像相同格式的两个独立成分E1和F1，所述E1和F1中的一个为待识别的目标独立成分，另外一个为非目标独立成分，所述非目标独立成分可以为噪声成分或摩尔纹成分。
步骤S3、选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值。
具体地，所述结构相似性(Structural Similarity index，SSIM)值的取值范围为0～1，SSIM值越大，表示图像失真越小，即越符合目标独立成分的要求，所述SSIM值的计算分别从亮度、对比度及结构三个方面度量图像相似性，定义待计算的独立成分为图像X、比较图像为图像Y，具体计算公式为：
SSIM＝L×C×S；
其中，
Figure PCTCN2019070008-appb-000003
其中，μ x为该独立成分中所有像素的灰阶值的均值，μy为比较图像中所有像素的灰阶值的均值，σ x为该独立成分中所有像素的灰阶值的方差，σ y为比较图像中所有像素的灰阶值的方差，σ xy为比较图像中所有像素的灰阶值与 独立成分中所有像素的灰阶值的协方差，C1、C2及C3分别为亮度常数、对比度常数及结构常数，L、C及S分别为亮度比较值、对比度比较值及结构比较值，SSIM为独立成分与比较图像之间的SSIM值。
具体地，所述比较图像的可以根据需要进行从N张输入图像任选一张。
例如，在本发明的第一实施例中，所述步骤S3中选择亮度最高的图像A作为比较图像进行SSIM值的计算，则所述步骤S3具体包括：计算独立成分E与所述图像A之间的SSIM值，计算独立成分F与所述图像A之间的SSIM值，计算独立成分G与所述图像A之间的SSIM值。必要时，在本发明的第一实施例中，也可以选择亮度最低的图像B作为比较比较计较图像进行SSIM值的计算，此时所述步骤S3相应为计算独立成分E与所述图像B之间的SSIM值，计算独立成分F与所述图像B之间的SSIM值，计算独立成分G与所述图像B之间的SSIM值。当然在本发明的第一实施例中，还可以选择亮度最低的图像C作为比较比较计较图像进行SSIM值的计算，此时所述步骤S3相应为计算独立成分E与所述图像C之间的SSIM值，计算独立成分F与所述图像C之间的SSIM值，计算独立成分G与所述图像C之间的SSIM值。
在本发明的第二实施例中，所述步骤S3中选择亮度最高的图像A1作为比较图像进行SSIM值的计算，则所述步骤 S3具体包括：计算独立成分E1与所述图像A1之间的SSIM值，计算独立成分F1与所述图像A1之间的SSIM值。必要时，在本发明的第一实施例中，也可以选择亮度最低的图像B1作为比较比较计较图像进行SSIM值的计算，此时所述步骤S3相应为计算独立成分E1与所述图像B1之间的SSIM值，计算独立成分F1与所述图像B1之间的SSIM值。
步骤S4、设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量，所述亮度极值点为独立成分中灰阶值位于其背景范围外的像素的数量。
具体地，所述亮度极值点实际上就是独立成分中的缺陷区域，即Mura区域，所述设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量可以有多种不同的方法。
例如，在本发明的一实施例中，所述步骤S4具体包括：分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
在本发明的另一实施例中，所述步骤S4具体包括：分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；依据独立成分中每一列像素 的背景范围统计独立成分中亮度极值点数量。
在本发明的又一实施例中，为了进一步增强亮度极值点数量统计的准确性，所述步骤S4具体包括：分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；依据独立成分中每一列像素的背景范围再次统计独立成分中亮度极值点数量；比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与依据该独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
此外，在本发明的其他实施例中还可以设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值，设定每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值，以更好的区分低灰阶的Mura。
步骤S5、根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选择比较值最大的独立成分作为目标独立成分；
所述比较值的计算公式为：Q＝SSIM+w/M，其中Q为独立成分的比较值，SSIM为独立成分的SSIM值，w为与SSIM值相关的系数，M为独立成分的亮度极值点数量。
具体地，所述步骤S5中当独立成分的SSIM值位于0～0.1之间时，w等于0，当独立成分的SSIM值位于0.1～0.5之间时，w等于1，当独立成分的SSIM值大于0.5时，w等于3。
步骤S6、对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷分割。
具体地，选择缺陷阈值的过程可参考所述步骤S4中设定背景区域的方法进行。
综上所述，本发明提供一种基于独立成分自适应选择的Mura侦测方法，包括如下步骤：将N张输入图像分别组合成一个混合矩阵；对混合矩阵进行ICA变换，获得N个独立成分；选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值；设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量；根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选 择比较值最大的独立成分作为目标独立成分；对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷分割，通过分别计算各个独立成分的SSIM值及统计各个独立成分中亮度极值点数量，并据此选择目标独立成分，能够取代人眼自适应选择目标独立成分，改善Mura侦测方法，满足生产自动化的需要。
以上所述，对于本领域的普通技术人员来说，可以根据本发明的技术方案和技术构思作出其他各种相应的改变和变形，而所有这些改变和变形都应属于本发明权利要求的保护范围。

Claims (19)

  1. 一种基于独立成分自适应选择的Mura侦测方法，其包括如下步骤：
    步骤S1、将N张输入图像分别转换为一维向量，并组合成一个混合矩阵，N为大于1的整数；
    步骤S2、对混合矩阵进行ICA变换，获得N个独立成分；
    步骤S3、选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值；
    步骤S4、设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量，所述亮度极值点数量为独立成分中灰阶值位于其背景范围外的像素的数量；
    步骤S5、根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选择比较值最大的独立成分作为目标独立成分；
    所述比较值的计算公式为：Q＝SSIM+w/M，其中Q为独立成分的比较值，SSIM为独立成分的SSIM值，w为与SSIM值相关的系数，M为独立成分的亮度极值点数量；
    步骤S6、对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷 分割；
    其中定义待计算的独立成分为图像X，比较图像为图像Y，所述步骤S3中计算独立成分与比较图像之间的SSIM值的公式为：
    SSIM＝L×C×S；
    其中，
    Figure PCTCN2019070008-appb-100001
    其中，μ x为该独立成分中所有像素的灰阶值的均值，μ y为比较图像中所有像素的灰阶值的均值，σ x为该独立成分中所有像素的灰阶值的方差，σ y为比较图像中所有像素的灰阶值的方差，σ xy为比较图像中所有像素的灰阶值与独立成分中所有像素的灰阶值的协方差，C1、C2及C3分别为亮度常数、对比度常数及结构常数，L、C及S分别为亮度比较值、对比度比较值及结构比较值，SSIM为独立成分与比较图像之间的SSIM值。
  2. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
  3. 如权利要求1所述的基于独立成分自适应选择的 Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
    依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
  4. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
    分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
    依据独立成分中每一列像素的背景范围再次统计独立成分中亮度极值点数量；
    比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与依据该独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
  5. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
  6. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
    依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
  7. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，gmin为该独立成分中所有像素的灰阶值的最小值；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
    分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
    依据独立成分中每一列像素的背景范围再次统计独立成分中亮度极值点数量；
    比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与该依据独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
  8. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S1中将3张输入图像分别转换为一维向量，所述3张输入图像分别为亮度最高的图像、亮度为50％的图像及亮度最低的图像。
  9. 如权利要求1所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S5中当独立成分的SSIM值位于0～0.1之间时，w等于0，当独立成分的SSIM值位于0.1～0.5之间时，w等于1，当独立成分的SSIM值大于0.5时，w等于3。
  10. 一种基于独立成分自适应选择的Mura侦测方法，其中包括如下步骤：
    步骤S1、将N张输入图像分别转换为一维向量，并组合成一个混合矩阵，N为大于1的整数；
    步骤S2、对混合矩阵进行ICA变换，获得N个独立成分；
    步骤S3、选定N张输入图像中的一张输入图像作为比较图像，分别计算每一个独立成分与所述比较图像之间的SSIM值；
    步骤S4、设定各个独立成分的背景范围并统计各个独立成分中亮度极值点数量，所述亮度极值点数量为独立成分中灰阶值位于其背景范围外的像素的数量；
    步骤S5、根据各个独立成分的SSIM值及各个独立成分的亮度极值点数量计算各个独立成分的比较值，并选择比较值最大的独立成分作为目标独立成分；
    所述比较值的计算公式为：Q＝SSIM+w/M，其中Q为独立成分的比较值，SSIM为独立成分的SSIM值，w为与SSIM值相关的系数，M为独立成分的亮度极值点数量；
    步骤S6、对目标独立成分进行ICA逆变换，得到缺陷增强图像，并选择缺陷阈值对所述缺陷增强图像进行缺陷分割。
  11. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中定义待计算的独立成分为图像X，比较图像为图像Y，所述步骤S3中计算独立成分与比较 图像之间的SSIM值的公式为：
    SSIM＝L×C×S；
    其中，
    Figure PCTCN2019070008-appb-100002
    其中，μ x为该独立成分中所有像素的灰阶值的均值，μ y为比较图像中所有像素的灰阶值的均值，σ x为该独立成分中所有像素的灰阶值的方差，σ y为比较图像中所有像素的灰阶值的方差，σ xy为比较图像中所有像素的灰阶值与独立成分中所有像素的灰阶值的协方差，C1、C2及C3分别为亮度常数、对比度常数及结构常数，L、C及S分别为亮度比较值、对比度比较值及结构比较值，SSIM为独立成分与比较图像之间的SSIM值。
  12. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
  13. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶 值的平均值和方差；
    依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
  14. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为μ-σ至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
    分别设定独立成分中每一列像素的背景范围为μ’-σ’至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
    依据独立成分中每一列像素的背景范围再次统计独立成分中亮度极值点数量；
    比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与依据该独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
  15. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为(μ +g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量。
  16. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一列像素的背景范围为(μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差，g min为该独立成分中所有像素的灰阶值的最小值；
    依据独立成分中每一列像素的背景范围统计独立成分中亮度极值点数量。
  17. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S4中具体包括：
    分别设定独立成分中每一行像素的背景范围为(μ+g min)/2至μ+σ，其中μ和σ分别为该独立成分中该行像素的灰阶值的平均值和方差，gmin为该独立成分中所有像素的灰阶值的最小值；
    依据独立成分中每一行像素的背景范围统计独立成分中亮度极值点数量；
    分别设定独立成分中每一列像素的背景范围为 (μ’+g min)/2至μ’+σ’，其中μ’和σ’分别为该独立成分中该列像素的灰阶值的平均值和方差；
    依据独立成分中每一列像素的背景范围再次统计独立成分中亮度极值点数量；
    比较依据该独立成分中每一行像素的背景范围的统计得到的亮度极值点数量与该依据独立成分中每一列像素的背景范围的统计得到的亮度极值点数量，并以两者中的较大值作为最终的亮度极值点数量。
  18. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S1中将3张输入图像分别转换为一维向量，所述3张输入图像分别为亮度最高的图像、亮度为50％的图像及亮度最低的图像。
  19. 如权利要求10所述的基于独立成分自适应选择的Mura侦测方法，其中所述步骤S5中当独立成分的SSIM值位于0～0.1之间时，w等于0，当独立成分的SSIM值位于0.1～0.5之间时，w等于1，当独立成分的SSIM值大于0.5时，w等于3。
PCT/CN2019/070008 2018-06-19 2019-01-02 基于独立成分自适应选择的Mura侦测方法 WO2019242294A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810631082.8A CN109064441B (zh) 2018-06-19 2018-06-19 基于独立成分自适应选择的Mura侦测方法
CN201810631082.8 2018-06-19

Publications (1)

Publication Number Publication Date
WO2019242294A1 true WO2019242294A1 (zh) 2019-12-26

Family

ID=64820565

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/070008 WO2019242294A1 (zh) 2018-06-19 2019-01-02 基于独立成分自适应选择的Mura侦测方法

Country Status (2)

Country Link
CN (1) CN109064441B (zh)
WO (1) WO2019242294A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114112323A (zh) * 2021-11-08 2022-03-01 云谷（固安）科技有限公司 显示面板显示均匀性的检测方法及检测装置

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109064441B (zh) * 2018-06-19 2020-07-28 深圳市华星光电半导体显示技术有限公司 基于独立成分自适应选择的Mura侦测方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130265408A1 (en) * 2010-12-06 2013-10-10 Kohei Yamaguchi Charged particle beam apparatus
CN105913419A (zh) * 2016-04-07 2016-08-31 南京汇川图像视觉技术有限公司 基于ICA学习和多通道融合的TFT-LCD mura缺陷检测方法
CN107845087A (zh) * 2017-10-09 2018-03-27 深圳市华星光电半导体显示技术有限公司 液晶面板亮度不均匀缺陷的检测方法和系统
CN108053402A (zh) * 2018-01-04 2018-05-18 深圳信息职业技术学院 一种缺陷图像分割方法
CN109064441A (zh) * 2018-06-19 2018-12-21 深圳市华星光电半导体显示技术有限公司 基于独立成分自适应选择的Mura侦测方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101909525B (zh) * 2008-01-11 2013-06-05 株式会社岛津制作所 图像处理方法、装置以及断层摄影装置
CN103413288B (zh) * 2013-08-27 2015-10-21 南京大学 一种lcd总体检测缺陷方法
CN103792699A (zh) * 2013-09-09 2014-05-14 中华人民共和国四川出入境检验检疫局 基于B样条曲面拟合的TFT-LCD Mura缺陷机器视觉检测方法
CN107328791B (zh) * 2017-07-31 2020-06-30 京东方科技集团股份有限公司 一种缺陷检测方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130265408A1 (en) * 2010-12-06 2013-10-10 Kohei Yamaguchi Charged particle beam apparatus
CN105913419A (zh) * 2016-04-07 2016-08-31 南京汇川图像视觉技术有限公司 基于ICA学习和多通道融合的TFT-LCD mura缺陷检测方法
CN107845087A (zh) * 2017-10-09 2018-03-27 深圳市华星光电半导体显示技术有限公司 液晶面板亮度不均匀缺陷的检测方法和系统
CN108053402A (zh) * 2018-01-04 2018-05-18 深圳信息职业技术学院 一种缺陷图像分割方法
CN109064441A (zh) * 2018-06-19 2018-12-21 深圳市华星光电半导体显示技术有限公司 基于独立成分自适应选择的Mura侦测方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
BI XIN: "Study on the methods of machine vision inspection for the mura defect TFT-LCD process", CHINA DOCTORAL DISSERTATIONS FULL-TEXT DATABASE, 15 July 2010 (2010-07-15), ISBN: 1674-022X *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114112323A (zh) * 2021-11-08 2022-03-01 云谷（固安）科技有限公司 显示面板显示均匀性的检测方法及检测装置
CN114112323B (zh) * 2021-11-08 2024-03-22 云谷（固安）科技有限公司 显示面板显示均匀性的检测方法及检测装置

Also Published As

Publication number Publication date
CN109064441A (zh) 2018-12-21
CN109064441B (zh) 2020-07-28

Similar Documents

Publication Publication Date Title
JP6354101B2 (ja) 自己相似性によって推進するブレンドを用いた詳細部保存性を有する非局所平均画像ノイズ除去
WO2019137038A1 (zh) 注视点确定方法、对比度调节方法及对比度调节装置、虚拟现实设备及存储介质
CN107293265B (zh) 显示屏画面调整方法、显示终端及可读存储介质
US20230094077A1 (en) Backlight adjusting method of display device, backlight adjusting device and display device
WO2016045146A1 (zh) 用于显示器的图像色彩增强方法及装置
CN109243384B (zh) 显示设备及其驱动方法、驱动装置和计算机可读介质
WO2016197450A1 (zh) 液晶面板及其驱动方法
CN104167194A (zh) 液晶面板的灰阶值设定方法以及液晶显示器
US11074888B2 (en) Image data processing method and apparatus, image display method and apparatus, storage medium and display device
WO2019242294A1 (zh) 基于独立成分自适应选择的Mura侦测方法
CN110246470B (zh) 执行图像自适应色调映射的方法和采用该方法的显示装置
CN116825039B (zh) 背光亮度计算方法、显示设备及计算机可读存储介质
WO2019101005A1 (zh) 像素补偿方法、装置和终端设备
CN107290345B (zh) 一种基于aoi的显示面板缺陷分类方法及装置
CN107103865B (zh) 检测显示屏中显示区域的方法和装置
WO2016070455A1 (zh) 液晶面板及其像素单元设定方法
CN113990263B (zh) 显示屏的背光处理方法、装置、存储介质及电子设备
US9509902B1 (en) Methods and devices for correction of camera module sensitivity and flash color variation
JP2017530390A (ja) 画像の表示方法及びディスプレイシステム
Chang et al. SSIM-based quality-on-demand energy-saving schemes for OLED displays
TWI672669B (zh) 影像處理方法與裝置
CN114495812B (zh) 显示面板亮度补偿方法、装置、电子设备及可读存储介质
US9571744B2 (en) Video processing method and apparatus
US11978183B2 (en) Image processing device and image enhancing method thereof
WO2021184931A1 (zh) 适用于光学穿透式头戴显示器的颜色对比增强绘制方法、装置以及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19822031

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19822031

Country of ref document: EP

Kind code of ref document: A1