JP2004212311A - Method and apparatus for detecting unevenness defect - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To highly accurately and automatically detect unevenness defects of a screen regardless of the size of the defects, and to perform quantitative evaluation on the detected unevenness defects by providing a constitution in which a threshold is automatically determined on the basis of luminance statistical data in a detection image. <P>SOLUTION: The screen to be inspected 10 is imaged by a CCD camera 6. The difference between the image captured by a computer 7 an a background image 14 is acquired to create the inspection image (background difference image) 15. A plurality of stages of image smoothing processing is performed on the inspection image or its reduced image to create first and second smoothed images 16 and 17. Differential processing is performed between the inspection image or its reduced image and the first smoothed image 16 and the first smoothed image 16 and the second smoothed image 17. Luminance data in the images acquired by the differential processing is statistically computed. On the basis of the statistical data, the threshold value is determined to extract defect candidates, and evaluation values of the defect candidates are computed. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

これらの計算式により、欠陥候補として検出された白ムラ欠陥、黒ムラ欠陥を、その大きさ（面積）、個数（Ｂｌｏｂ番号）と共に、客観的なデータで定量的に評価することができる。従って、ムラ欠陥の検出精度が高いものとなる。
【００２８】
また、算出された評価値は不良判断のために用いるだけでなく、その値の大きさに基づいて製品をいくつかのグループに分類することで、ムラ欠陥のランク（等級）付けや製品の等級化が可能となる。例えば、液晶パネルをＲ（赤）、Ｇ（緑）、Ｂ（青）のカラー別に、プロジェクタのライトバルブとして使用する場合に、比視感度は緑のとき（波長λ＝５５５ｎｍ近辺のとき）が最も高いので、評価値の最も小さい方のランクのものをＧの製品、次に大きいランクのものをＲの製品、その次のものをＢの製品、さらにそれ以上の評価値のものを不良として、等級化することができる。
【００２９】
この実施形態は、以上のように構成されているので、液晶パネル等の表示デバイスの画面に存在するムラ欠陥を欠陥サイズの大小にかかわらず、高精度に自動的に検出することができ、かつ、それらの欠陥を個々に定量的に評価することができる。
また、輝度統計データに基づいてムラ欠陥を評価するものであるので、製品や部品の品質データを収集・分析することにより、品質管理に役立てることができ、更なる品質の向上を目指した手法を構築することも可能となる。
【００３０】
本発明は、前記のようなＴＦＴ素子を用いた液晶パネル（液晶ライトバルブ）に限られるものではなく、その他のダイオード素子を用いた液晶パネルやプラズマディスプレイ、有機ＥＬディスプレイ、ＤＭＤ（ダイレクト・ミラー・デバイス）などの表示体部品、ならびにそれらを使用した表示装置・製品の検査に利用することができるものであり、これらに使用した場合でも本発明の範囲から除外されるものでないことはいうまでもない。
【図面の簡単な説明】
【図１】本発明の実施の形態によるムラ欠陥検出装置の構成図。
【図２】ムラ欠陥の検出処理を示すフローチャート。
【図３】入力画像からムラ欠陥検出画像に至るまでの各画像の模式図。
【図４】画像サイズの縮小方法の説明図。
【符号の説明】
１ プロジェクタ、２ 液晶パネル、３ スクリーン、４ 画像、５ パターンジェネレータ、６ ＣＣＤカメラ、７ コンピュータ、８ 表示装置、１０ 検査対象画面、１１ 入力画像、１２ 表示エリア画像、１３ 原画像（表示エリア補正画像）、１４ 背景画像、１５ 検査画像（背景差分画像）、１６ 第１の平坦化画像、１７ 第２の平坦化画像、１８ 第１の検出画像、１９ 第２の検出画像、２０ ムラ欠陥、２０ａ サイズ大のムラ欠陥、２０ｂ サイズ小のムラ欠陥、３１ スクリーンの縁部分[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a method and a device for detecting a non-uniformity defect that automatically detects display non-uniformity in a screen in an inspection process of a liquid crystal panel or the like.
[0002]
[Prior art]
One of the defects that appear on a screen of a liquid crystal display device or the like is a so-called unevenness defect. An uneven defect is a state in which a certain area of a display screen has a difference in luminance from another area, and a state in which a bright area or a dark area is present over a relatively large area. Since such uneven defects deteriorate image quality, they are subject to a visual inspection of a display body. Conventionally, a visual inspection by a person has been usual, but recently, an automatic inspection has been performed. There have already been many proposals for an automatic inspection method for uneven defects, for example, there are Patent Documents 1 and 2.
Patent Literature 1 discloses a method in which an image captured by imaging is divided into a predetermined number of meshes, and a difference between an image obtained by integrating density values of pixels in each mesh and an image obtained by averaging the density added value is obtained. is there.
Further, Japanese Patent Application Laid-Open No. H11-157210 discloses a method of creating a binarized image by binarizing a captured image using a plurality of thresholds, and inspecting the binarized image for unevenness defects.
[0003]
[Patent Document 1]
JP-A-8-145904 (claims, paragraphs [0011] to [0024], FIGS. 1 to 8)
[Patent Document 2]
JP-A-6-147867 (claims, paragraphs [0011] to [0015], FIGS. 1 and 2)
[0004]
[Problems to be solved by the invention]
However, in the detection method of Patent Document 1, since the detectable defect size depends on the size of the mesh division, only the defect size corresponding to the mesh size can be detected. Therefore, there is a problem that detectable defect sizes are limited, and various sizes cannot be covered.
Further, in the detection method of Patent Document 2, it is necessary to set a plurality of thresholds, but how to determine the threshold corresponding to the defect size is a major problem. Therefore, a problem in defect detection accuracy arises depending on how the threshold value is determined.
[0005]
SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and has a configuration in which a threshold value is automatically determined based on luminance statistical data in a detected image, so that a screen unevenness defect can be reduced to a defect size. It is an object of the present invention to provide a method and apparatus for detecting a mura defect which automatically detects the mura defect with high accuracy regardless of the size of the mura defect and enables quantitative evaluation of the detected mura defect.
[0006]
[Means for Solving the Problems]
A method for detecting a non-uniformity defect according to the present invention includes the steps of:
From the image captured by the imaging, a step of taking a difference from a background image created in advance to create an inspection image,
Performing a plurality of stages of image flattening processing on the inspection image or a reduced image obtained by reducing the inspection image to create first and second flattened images;
Performing difference processing between the image of the inspection image or the reduced image of the inspection image and the first flattened image, and between the images of the first flattened image and the second flattened image. When,
Statistically calculating the luminance data of each pixel in the image obtained by the difference processing,
A step of determining a threshold based on the luminance statistical data and extracting a defect candidate;
It is characterized by having.
[0007]
In the present invention, in order to detect a non-uniformity defect, first, a screen to be inspected is imaged, and a difference from the background image is taken from the image to create an inspection image which is a background difference image. Then, the inspection image or the reduced image of the inspection image is subjected to image flattening processing in a plurality of stages, and between the image with the first flattened image obtained by the flattening processing, and the first and second images. A plurality of detected images are obtained by performing difference processing between the flattened images. When an uneven defect exists in the inspection image, an uneven defect having a relatively large size remains in the first detected image, and an uneven defect having a relatively small size remains in the second detected image. .
A threshold is determined to extract these uneven defects as defect candidates. The threshold value is determined based on statistical calculation of luminance data of each pixel in the detected image, and based on the luminance statistical data. Therefore, since the threshold value for extracting as a defect candidate is automatically determined based on the luminance statistical data in the image, it is possible to accurately detect uneven defects of various sizes.
[0008]
In the present invention, when an image is captured by imaging, data of the captured image preferably uses data of 12 bits or more for one pixel and has 4096 gradations or more.
By using such high-resolution image data, it becomes possible to detect uneven defects that could not be detected at 256 gradations, and since the luminance resolution is improved, the defect detection accuracy can be further improved. It becomes. Further, the accuracy of the evaluation value of the defect candidate described later is also improved.
[0009]
The background image is obtained by averaging a plurality of images captured by the same optical system and the same imaging system.
As a result, the component of the defect part randomly present in the inspection object is weakened, and only the defect-like luminance change caused by the illumination of the screen or the image pickup means, the lens characteristic, or the like, which is present at the same position every other time, is reduced. Is obtained.
[0010]
In the present invention, the step of creating an inspection image includes a step of extracting a display area from an image including a screen to be inspected, and applying a geometric deformation to the display area to make it a rectangle.
Since the inspection image is created based on an image obtained by capturing the screen of the inspection target, if the inspection target screen is an image projected on the screen by, for example, a projector, the image includes an edge portion of the screen. Or the display area may be oblique to the screen. Therefore, when an inspection image is created, a display area is extracted from an image including a screen to be inspected, and the display area is subjected to geometric deformation to form a rectangle, thereby distorting the image as described above. Deformation and the like can be corrected.
[0011]
Uneven defects include bright defects (white uneven defects) and dark defects (black uneven defects). In order to be able to detect both the bright defect and the dark defect, the threshold value of the defect candidate is determined for each of the bright defect and the dark defect.
The method further includes a step of calculating an evaluation value for the extracted defect candidate based on the luminance statistical data.
Therefore, defect candidates can be objectively and quantitatively evaluated, and accurate evaluation can be performed.
The evaluation value of the defect candidate is calculated for each of the bright defect and the dark defect, as in the case of the threshold value.
[0012]
The threshold is determined using the average value (average luminance data) of the luminance statistical data of the entire screen and the standard deviation. For example,
The white spot threshold for extracting a light defect is
Average luminance data + a1 × standard deviation The black unevenness threshold for extracting a dark defect is
It can be obtained from a calculation formula of average luminance data−a2 × standard deviation. Here, a1 and a2 are certain fixed constants. Also, the threshold calculated here is compared with the maximum value (maximum luminance data) and the minimum value (minimum luminance data) of the luminance statistical data of the entire screen to determine whether or not the inspection image has a non-uniform defect candidate. Do.
[0013]
The evaluation value of the defect candidate is calculated using the area, the maximum luminance value or the minimum luminance value of the defect candidate, the average value of the luminance statistical data over the entire screen, and the standard deviation.
The evaluation value is calculated for each white unevenness defect candidate and black unevenness defect candidate on each screen.
Therefore, it is possible to classify products by ranking unevenness defects.
These statistical data can be used for quality control.
[0014]
A device for detecting a mura defect according to the present invention uses the method for detecting a mura defect according to any one of claims 1 to 10.
More specifically, the inspection apparatus includes an imaging unit and an inspection apparatus main body that performs image processing. The inspection apparatus main body is generally configured by a computer. By incorporating an inspection program for performing each of the above-described processes into this computer, the unevenness defect can be automatically inspected.
[0015]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a configuration diagram of an unevenness defect detection apparatus according to an embodiment of the present invention.
In this embodiment, for example, the screen 10 to be inspected is a screen of a liquid crystal panel (also referred to as a liquid crystal light valve) 2 using a TFT element by the projector 1. When performing an inspection, the image 4 is projected on the screen 3 by the projector 1. The image 4 is drawn by giving a predetermined pattern to the liquid crystal panel 2 by the pattern generator 5. For example, an image 4 is picked up by a CCD camera 6 as an image pickup means, and the image signal is converted from an analog signal to a digital signal by an A / D converter (not shown) and is taken into a computer 7 which is a main body of the inspection apparatus. At this time, the image data is converted by the A / D converter into 12-bit data in which, for example, black is "0" and white is "4095" for each pixel, and is represented by a luminance value of 4096 gradations. Further, the computer 7 detects the light / dark unevenness defect by processing the image data of the image 4 taken into the image memory by a method described later. In detecting a defect, statistical processing of luminance information in the detected image is performed, a threshold is determined based on the statistical data, defect candidates are extracted, and an evaluation value for quantitatively evaluating the extracted defect candidate is determined. It is calculated. These inspection results are displayed on the display device 8.
[0016]
FIG. 2 is a flowchart used in the unevenness defect detection processing, and FIG. 3 is a schematic diagram of an image after image processing in each stage from an input image to an unevenness detection image. The detection processing of the unevenness defect is automatically performed according to the inspection program incorporated in the computer 7 or the image processing apparatus.
The processing procedure will be described according to the flowchart of FIG.
[0017]
(1) Display area extraction processing (step S1)
Extracting a display area means extracting only a screen portion to be inspected from an input image captured by imaging. For example, FIG. 3A shows an input image at the time of imaging. As shown in this figure, the input image 11 at the time of imaging includes the edge portion 31 of the screen 3 or the display area image 12 corresponding to the screen portion is not exactly rectangular but is distorted obliquely with respect to the screen 3. May go out. This is due to the fact that the screen 3 and the CCD camera 6 as the image pickup means are not strictly parallel to each other, or that distortion occurs due to the lens characteristics of the CCD camera 6 or the like. In addition, even when an image of a liquid crystal panel or the like is directly captured instead of indirectly captured as described above, distortion or deformation of the input image may occur. Of course, when the imaging unit or the screen of the inspection target is accurately set and imaged so that the edge portion 31 does not enter as described above, for example, the field of view is set so that the entire screen portion of the detection target falls within the field of view of the imaging unit. When set correctly, the display area extraction processing and the correction processing described below can be omitted, and the image captured by imaging directly becomes the original image.
[0018]
When the input image 11 is distorted as shown in FIG. 3A, only the display area image 12 of the screen portion is extracted as shown in FIG. A display area correction image 13 that is deformed and corrected to an accurate rectangle is created. Here, the display area correction image 13 is an original image to be actually inspected.
In the image correction processing by the geometric deformation, the coordinates of the four corners of the display area image 12 not including the edge portion 31 of the screen 3 are detected by pattern matching processing or the like, and the coordinates match the coordinates of the four corners of the rectangle. Is performed by performing coordinate conversion as described above. In this way, the original image 13 in which only the screen portion to be inspected is extracted can be created by correcting the distortion, deformation and the like of the input image 11. By setting the size of the rectangle set at this time to, for example, 1200 × 1000 pixel size, the file size of the original image 13 becomes, for example, 1200 × 1000 pixel size. This size is not particularly fixed, and may be set to a size equal to or close to the pixel size of the CCD camera.
[0019]
(2) Background image difference processing (step S2)
The background image difference process is a process of subtracting a previously created background image 14 as shown in FIG. 3C from the original image 13 created as described above. By subtracting the background image 14 from the original image 13, other than the defects existing in the liquid crystal panel to be inspected included in the original image 13, the defect-like brightness caused by the illumination of the imaging means, the lens characteristics, etc. Changes can be eliminated. In addition, in this difference processing, 2048 (4096 gradation ×× value) is added as an offset value at the time of the difference processing so that the luminance data does not become a negative value. The background image 14 is based on sample images of a plurality of (for example, about 20) liquid crystal panels with as few defects as possible, which are captured by using the same optical system and the same imaging system. It is created by averaging the luminance data, is created in advance by the same method as the original image 13, and is stored in the image memory of the computer 7.
As a result of the background image difference processing, a background difference image as shown in FIG. If so, the background difference image (inspection image) 15 has uneven defects 20 of various sizes.
[0020]
(3) Image size reduction processing (step S3)
In this process, since the background difference image (inspection image) 15 includes defects of various sizes, defects that are to be detected (corresponding to the size of the defect) and are smaller than the uneven defect size are detected. The inspection image 15 is reduced to an appropriate image size in order to remove (stains and streak defects). For example, the inspection image 15 is reduced to a quarter image size, that is, 300 × 250 pixels. Note that the reduction ratio is not limited to 1/4.
In the reduction processing, for example, as shown in FIG. 4, a method of repeating twice a 1/2 reduction processing of newly allocating the average value of the luminance data of four pixels of the original image (here, the inspection image 15) to one pixel is used. Creates a reduced image.
By reducing the image size of the inspection image 15 in this way, there is also an advantage that the processing time required for defect detection can be reduced.
In FIG. 3, all the images are shown in the same size, but the images (e) to (f) are actually images that are reduced to 1/4. Further, a quarter reduced image of the background difference image (inspection image) in (d) is not shown.
Here, reduction of the image size has been described, but of course, it is not always necessary to reduce the image size. For example, even if the filter size is adjusted by using a smoothing filter process, defects smaller than the uneven defect size to be detected are removed. Good.
[0021]
(4) Image duplication processing (step S4)
Since processing for separating the unevenness defect from other background parts (parts having no defect) is performed, at least three images are required as described later. A duplicate (copy) image of the / 4 reduced image is to be created. The copied image can also be used for detecting spots, spots, and line defects other than the unevenness defect.
[0022]
(5) Flattening process (Steps S5 and S6)
Next, a flattening process is performed on the background difference image (inspection image) 15 in a plurality of stages as a pre-process for separating the unevenness defect from the other background portion (the portion having no defect). Here, a two-stage flattening process is performed.
In the first flattening 1 process (step S5), a first flattened image 16 as shown in FIG. 3E is created based on a 縮小 reduced image of the inspection image 15 or its copy image. Is done.
In the first-stage flattening process, a process that applies a morphological process or the like is performed such that a large change in luminance is flattened while leaving a relatively small change in luminance.
In the second flattening 2 process (step S6), a second flattened image 17 that has been further flattened is created from the first flattened image 16 (see FIG. 3F).
In the second-stage flattening process, a process is performed on the first flattened image 16 from which a large luminance change has been removed so as to flatten a smaller luminance change by a process using a smoothing filter or the like.
[0023]
(6) Difference processing (Steps S7 and S8)
In the difference processing in step S7, the first flattened image 16 (the image (e) in FIG. 3) is subtracted from the 縮小 reduced image of the inspection image 15 or its copy image (the image (d) in FIG. 3). I do. By this processing, a first detection image 18 as shown in FIG. In the first detection image 18, the small size uneven defect is removed, and only the relatively large size uneven defect 20a remains.
Next, in the difference processing in step S8, the second flattened image 17 (the image (f) in FIG. 3) is subtracted from the first flattened image 16 (the image (e) in FIG. 3). Then, a second detection image 19 as shown in FIG. In the second detected image 19, only the small-sized uneven defect 20b remains.
[0024]
(7) Statistical calculation (Steps S9 and S10)
Next, using the luminance data of each pixel in the uneven defect detection images 18 and 19, a statistical amount calculation of the luminance data over the entire screen is performed. In the statistical calculation, the average value Lave, the standard deviation σ, the maximum value Lmax, and the minimum value Lmin of the luminance data are obtained. From these four pieces of luminance statistical data, the threshold values of white unevenness and black unevenness are determined as follows, for example.
White spot threshold: Lave + a1 × σ
Black unevenness threshold: Lave-a2 × σ
Here, a1 and a2 are certain fixed constants.
Therefore, the threshold value for detecting white unevenness and black unevenness can be automatically determined based on the brightness statistical data by statistically calculating the brightness data in the unevenness defect detection images 18 and 19. Therefore, the threshold value is not artificial, trial and error, but objective and relative.
[0025]
(8) Check 1: Judge whether there is a defect candidate (step S11)
Then, a first check is performed based on the threshold value to determine whether there is a white unevenness defect or a black unevenness defect in the unevenness defect detection images 18 and 19. In this check 1, it is checked whether or not there are white side and black side defect candidates based on each of the threshold values. First, it is checked whether Lmax exceeds the white unevenness threshold value, and if it exceeds, it is determined that there is a white unevenness defect candidate in the image. Next, it is checked whether or not Lmin is equal to or less than the black unevenness threshold value. If there are no defect candidates, it is determined at this stage that there is no unevenness defect, and the inspection ends.
[0026]
(9) Blob processing (steps S12 and S13)
When it is determined in the above check 1 that there is a defect candidate, the defect candidate is extracted using a threshold. At this time, if it is determined in the check 1 that there is a white unevenness defect candidate, a candidate having a white unevenness threshold or more is extracted as a defect candidate. Those that are equal to or less than the unevenness threshold are extracted as defect candidates. If both are determined, both are extracted. A blob process is performed on the extracted defect candidate, and the area (S (n)) of the extracted defect candidate and the maximum luminance value Lmax (n) of the white defect defect candidate and the black defect defect candidate Then, the minimum value Lmin (n) of the luminance value is obtained. A blob is a specific grayscale value or a “block” (area) having a range of values existing in an image cut out by binarization processing. The “blob processing” refers to obtaining various characteristic values in the extracted defect candidate area using a blob tool or the like. The characteristic value in the defect candidate area obtained here is used in calculation of an evaluation value described below.
[0027]
(10) Check 2: Calculation of evaluation value (step S14)
In the second (ie, final) check, for the uneven defect candidate detected as a defect candidate and subjected to the blob processing, the luminance statistical data (average value, standard deviation σ) and the characteristic value obtained by the blob processing are used. Using (area S (n), maximum value Lmax (n), minimum value Lmin (n)), an evaluation value is calculated by the following equation.

With these formulas, the white spot defect and the black spot defect detected as defect candidates can be quantitatively evaluated by objective data together with their size (area) and number (Blob number). Therefore, the detection accuracy of the unevenness defect is high.
[0028]
In addition, the calculated evaluation value is used not only for determining a defect, but also by classifying the product into several groups based on the magnitude of the value, thereby assigning a rank (grade) of the unevenness defect and a product grade. Is possible. For example, when a liquid crystal panel is used as a light valve of a projector for each of R (red), G (green), and B (blue) colors, the relative luminous efficiency is green (when the wavelength is around 555 nm). Since it is the highest, the product having the lowest evaluation value ranks as a product of G, the product with the next highest rank is a product of R, the next product is a product of B, and the product with a higher evaluation value is defective. , Can be graded.
[0029]
Since this embodiment is configured as described above, it is possible to automatically detect uneven defects existing on the screen of a display device such as a liquid crystal panel with high accuracy regardless of the size of the defect, and , The defects can be individually and quantitatively evaluated.
In addition, since unevenness defects are evaluated based on luminance statistical data, collecting and analyzing the quality data of products and parts can be useful for quality control, and a method aimed at further improving quality is developed. It is also possible to build.
[0030]
The present invention is not limited to a liquid crystal panel (liquid crystal light valve) using the above-described TFT element, but includes a liquid crystal panel using another diode element, a plasma display, an organic EL display, a DMD (direct mirror mirror). It can be used for inspecting display parts such as devices) and display devices and products using them, and even if they are used for these, they are not excluded from the scope of the present invention. Absent.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an unevenness defect detection device according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating a process of detecting an unevenness defect.
FIG. 3 is a schematic diagram of each image from an input image to an uneven defect detection image.
FIG. 4 is an explanatory diagram of a method for reducing an image size.
[Explanation of symbols]
Reference Signs List 1 projector, 2 liquid crystal panel, 3 screen, 4 images, 5 pattern generator, 6 CCD camera, 7 computer, 8 display device, 10 inspection target screen, 11 input image, 12 display area image, 13 original image (display area correction image) ), 14 background images, 15 inspection images (background difference images), 16 first flattened images, 17 second flattened images, 18 first detected images, 19 second detected images, 20 uneven defects, 20a Large size irregularity defect, 20b Small size irregularity defect, 31 Screen edge

Claims (11)

Imaging a screen to be inspected;
From the image captured by the imaging, a step of creating an inspection image by taking a difference from a background image created in advance,
Performing a plurality of levels of image flattening processing on the inspection image or a reduced image obtained by reducing the inspection image to create first and second flattened images;
Performing difference processing between the image of the inspection image or the reduced image of the inspection image and the first flattened image, and between the images of the first flattened image and the second flattened image. When,
Statistically calculating the luminance data of each pixel in the image obtained by the difference processing,
A step of determining a threshold based on the luminance statistical data and extracting a defect candidate;
A method for detecting unevenness defects. 2. The method according to claim 1, wherein the data of the captured image has 12 bits of 4096 gradations or more. 3. The method according to claim 1, wherein the background image is obtained by averaging a plurality of images captured by the same optical system and the same imaging system. The method according to claim 1, wherein the step of creating the inspection image includes a step of extracting a display area from an image including the screen to be inspected, and applying a geometric deformation to the display area to make the display area rectangular. 3. The method for detecting an uneven defect according to any one of 3. The method according to claim 1, wherein the threshold value of the defect candidate is determined for each of a light defect and a dark defect. The method according to claim 1, further comprising: calculating an evaluation value for the extracted defect candidate based on the luminance statistical data. 7. The method according to claim 1, wherein the evaluation value of the defect candidate is calculated for each of a bright defect and a dark defect. The method according to claim 5, wherein the threshold is determined using an average value and a standard deviation of the luminance statistical data. 7. The defect candidate evaluation value is calculated using an area of the defect candidate, a maximum value or a minimum value of luminance of the defect candidate, and an average value and a standard deviation of the luminance statistical data. Or a method for detecting unevenness defects according to item 7. 10. The method according to claim 9, wherein the evaluation value is classified into a plurality of ranks according to the degree of the defect. An unevenness defect detection apparatus using the unevenness defect detection method according to any one of claims 1 to 10.

Images