JP2005172559A - Method and device for detecting line defect on panel - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and device for detecting line defects on a panel with enhanced detection accuracy for line defects as to a method and device for accurately and automatically detecting line defects on a panel, the line defects continuously arising in a lengthwise or crosswise direction in an inspection process in manufacturing a display device such as a liquid crystal panel and a projector which is an applied product thereof. <P>SOLUTION: This method comprises processes S6 and S7 for applying edge detection filters in different directions (horizontal and vertical) in order to stress line defects for detection to an image obtained by imaging an inspecting object panel, processes S8 and S9 for performing divided profile processing for dividing a luminance value of each pixel of an edge detection image in which line defects are stressed and the luminance values of respective pixels are integrated in direction in which line defects are stressed for each of divided areas to obtain integrated values, processes S10 and S11 for finding, by means of moving averages, an average value of integrated values in each of divided areas in which line defects are stressed and calculating statistical data by using the integrated values and the average value, and a process S12 for setting a threshold value of the integrated values based on the statistical data to extract defect candidates from the statistical data and the threshold value. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a panel line defect detection method for automatically and accurately detecting line defects appearing continuously in a vertical or horizontal direction in an inspection process in the manufacture of a display device such as a liquid crystal panel or a projector which is an application product thereof. Relates to the device.

  As a conventional LCD panel line defect detection method, a method of integrating pixel values in the vertical and horizontal directions using the directionality is generally used. For each pixel of interest, pixel values are added in the vertical direction, horizontal direction, and at least two diagonal directions within a predetermined range centered on the pixel of interest, and the maximum value among the calculation results Is used to generate a line unevenness extraction image by performing the process of setting the pixel value of the pixel of interest, and the line unevenness extraction image is binarized using a predetermined threshold value, thereby performing line unevenness extraction. Detecting the line unevenness in the LCD panel inspection. (Patent Document 1), “The process of imaging the entire screen to be inspected, the process of removing noise from the captured image captured by imaging, and the screen of the inspection target from the captured image” A step of extracting an inspected part image; a step of reconfiguring the inspected part image as a rectangular image by geometric deformation; a step of performing shading correction of the rectangular image; and a luminance value of the image after shading correction. A step of integrating in the horizontal direction and the vertical direction, respectively, and a step of detecting a line defect by comparing each of the integrated values in the horizontal direction and the vertical direction with a predetermined threshold value. Screen line defect detection method. (Patent Document 2) and the like have been proposed.

Japanese Patent Laid-Open No. 10-240933 JP 2003-168103 A

  In the conventional line defect detection method (Patent Document 1), for each target pixel of the gray image of the LCD panel, within a predetermined range centered on the target pixel, the vertical direction, the horizontal direction, and at least two oblique directions, A pixel value addition operation is performed, and the maximum value of each calculation result is processed to be the pixel value of the target pixel, thereby generating a line unevenness extraction image, and line unevenness having a higher luminance than the background is generated. Since detection is performed, lines with lower brightness than the background, such as black lines, cannot be detected, and there is a problem that detection accuracy is low.

  Further, the conventional line defect detection methods (Patent Documents 1 and 2) calculate the difference between the value of the target line defect and the average value of the average value of the entire screen when determining the presence or absence of the line defect. The absolute value of the difference is divided by the standard deviation. However, since the average value and the standard deviation are obtained over the entire inspection area including the line defect portion, the average value and the standard deviation are obtained. Therefore, since the values are not accurate, the detection accuracy is low. there were.

  The present invention has been made to solve such problems, and an object of the present invention is to obtain a panel line defect detection method and apparatus with improved line defect detection accuracy.

  The panel line defect detection method according to the present invention includes a step of applying an edge detection filter to different directions in order to emphasize and detect line defects in an image obtained by imaging a panel to be inspected, and a different direction. The luminance value of each pixel of the edge detection image in which the line defect is emphasized is divided into a plurality of rectangles, and the luminance value of each pixel is integrated in the direction in which the line defect is emphasized for each divided region to obtain an integrated value. A step of performing division profile processing, a step of calculating an average value of integrated values in each divided region in which line defects are emphasized by a moving average, and calculating statistical data using the integrated value and the average value, and statistical data The threshold value of the integrated value is set based on the above, and statistical data and a step of extracting defect candidates from the threshold value are included.

  In order to emphasize and detect line defects in an image obtained by imaging the panel to be inspected, edge detection filters are applied to the different directions, and line detection images in which the line defects are emphasized in different directions are each formed into a plurality of rectangles. The divided profile processing is performed to obtain the integrated value by integrating the luminance value of each pixel in the direction in which the line defect is emphasized in each divided region, and the integrated value in each divided region in which the line defect is emphasized is obtained. Statistical data is calculated, the threshold value of the integrated value is set based on the statistical data, and defect candidates are extracted from the statistical data and the threshold value. Judgment whether it is a non-defective product can be facilitated in a short time. In addition, since the average value of the integrated values in the divided area where the line defect is emphasized is obtained by moving average, and the statistical data is calculated using the integrated value and the average value (moving average), it exists in the divided profile. It is possible to avoid the influence of other peaks. In addition, even if an area showing a high luminance value changes gently, it is possible to suppress the influence of the height of the luminance value in subsequent processing. For this reason, the detection accuracy of the line defect is improved.

  In the line defect detection method for a panel according to the present invention, the calculation process for obtaining the moving average is performed by removing the central portion of the area for which the moving average is obtained from the calculation within a predetermined range. When the peak part is included in the central part, it is possible to eliminate the influence, and as a result, since the average part does not include the peak part, the average value is obtained low. (Luminance value−average value) at the time of determination is a large value and the peak portion is emphasized, so that the detection accuracy of the line defect is improved.

  Further, the line defect detection method for a panel according to the present invention extracts an image portion of the inspected portion from the captured image, and takes a difference between the extracted image of the inspected portion and the background image to inspect the inspected portion. And performing a pre-process for creating a background difference image that removes a defect-like luminance change caused by a non-uniformity, and a step for performing a flattening process for removing the influence of unevenness in the background difference image. Thereafter, the process proceeds to the step of applying the edge detection filter. Since the image to which the edge detection filter is applied has no defect-like luminance change caused by other than the part to be inspected and is not affected by unevenness, only the line defect is emphasized in the image obtained by the process of applying the edge detection filter. Become.

  Further, in the line defect detection method for a panel according to the present invention, the step of applying the edge detection filter in the different direction includes a horizontal edge detection filter for performing horizontal edge enhancement processing to obtain a horizontal line detection image, and vertical edge enhancement processing. And a vertical edge detection filter to obtain a vertical line detection image. Line defects appearing horizontally and vertically on the panel can be emphasized.

  In the line defect detection method for a panel according to the present invention, a step of adding the median gradation value of the image as an offset value to the image subjected to the edge detection filter at the same time as applying the edge detection filter is provided. It is a thing. With respect to the image subjected to the edge detection filter processing, the edge information on both sides of the line can be grasped by one image, and not only the white line but also the black line can be detected.

  In the line defect detection method for a panel according to the present invention, the step of performing the division profile processing divides the horizontal line detection image subjected to the edge detection processing in the horizontal direction into a plurality of rectangles in the vertical direction, and A process of performing horizontal profile processing for accumulating luminance values of pixels in the horizontal direction to obtain integrated values, and dividing a vertical line detection image subjected to vertical edge detection processing into a plurality of rectangles in the horizontal direction, And a step of performing vertical profile processing for accumulating the luminance values of the respective pixels in the divided region in the vertical direction to obtain the integrated value. The integrated value of each divided area divided into a plurality of rectangles acquired in the process of performing horizontal profile processing is not easily affected by the brightness fluctuation of the entire panel, and thin horizontal line defects can be detected and divided. Horizontal line defects that are relatively short compared to the width of the entire panel can also be detected. In addition, the integrated value of each divided area divided into a plurality of rectangles acquired in the process of performing vertical profile processing is not easily affected by the brightness variation of the entire panel, and thin vertical line defects can be detected and divided. This makes it possible to detect vertical line defects that are shorter than the overall height of the panel.

  Further, in the line defect detection method for a panel according to the present invention, the statistical data in the step of calculating the statistical data includes an average value, a standard deviation, a maximum value, and an average value obtained from the integrated values acquired from each divided region in the divided profile step. The threshold value which is the minimum value and is set based on the statistical data in the defect candidate extraction step is calculated by a predetermined formula based on the average value and the standard deviation obtained from the integrated values acquired from each divided region by the divided profile processing. Is. Candidates having line defects can be extracted by statistical processing in the defect candidate extraction step.

  In the line defect detection method for a panel according to the present invention, a line defect evaluation value is calculated by a predetermined formula based on statistical data of defect candidates, and the defect evaluation value exceeds a predetermined threshold value set in advance. Even if there is a line defect candidate depending on whether or not, a step of determining whether or not the line defect candidate is at a non-defective level is provided. A line defect evaluation value is calculated by a predetermined formula based on statistical data of an image determined to have a line defect candidate, and the line defect candidate is determined by whether or not the defect evaluation value exceeds a predetermined threshold value. Even if there is a defect, it can be determined whether it is a non-defective product level, so that the defect rank of the line defect can be determined, and the defect rank is determined only by defect candidates, so that the calculation time is It can be short, and defects can be ranked in a short time.

  In the line defect detection method for a panel according to the present invention, the line defect evaluation value in the step of determining whether or not the non-defective level is an average value obtained for each divided region of the image determined to have a line defect candidate. Since the standard deviation, the maximum value, and the minimum value are calculated by a predetermined formula, the line defect evaluation value in the defect candidate can be accurately obtained from the predetermined formula.

  The line defect detection apparatus for a panel according to the present invention includes an image pickup means for picking up an image of a panel to be inspected and a calculation means for performing calculation processing of the above-described line defect detection method for a panel.

Embodiment 1
FIG. 1 is a block diagram showing a configuration of a panel line defect detection apparatus according to Embodiment 1 of the present invention. In the first embodiment, for example, the screen 10 to be inspected is a screen of the liquid crystal light valve 2 using a TFT element by the projector 1. When performing the inspection, the projector 4 projects an image 4 on the screen 3. The image 4 is drawn by giving a predetermined pattern to the liquid crystal light valve 2 by the pattern generator 5. For example, the image 4 is picked up by the CCD camera 6 as the image pickup means, and the image signal is converted from an analog signal to a digital signal by an A / D converter (not shown), taken into the computer 7 and stored in the storage device 7b. The computer 7 performs image input, display area extraction, background image difference processing, flattening processing, image duplication processing, horizontal / vertical edge detection filter processing, horizontal / vertical profile processing for the image data of the image 4 stored in the storage device 7b. Various processes such as detection processing, calculation of statistical data, defect candidate extraction processing, non-defective product rank, evaluation value calculation, defect rank classification, and the like are performed, and the evaluation results are stored in the storage device 7b or displayed on the display device 8. Details of the above-described various arithmetic processes by the computer 7 will be described with reference to the flowchart of FIG.

  FIG. 2 is a flowchart showing the operation of the line defect detection apparatus of FIG. This operation is mainly performed by a CPU (arithmetic unit) 7a of the computer 7, which is performed according to a program stored in the storage device 7b.

(S1: Image input) First, the image 4 projected on the screen 3 is photographed by the CCD camera 6, and the computer 7 captures the image data of the image 4. At this time, the imaging data is converted into 12-bit digital data, for example, “0” for black and “4095” for white for each pixel by an A / D converter (not shown). It is represented by
(S2: Display Area Extraction) The CPU 7a of the computer 7 extracts a display area for extracting only the screen portion which is the display area of the inspected part from the captured image data. In this extraction screen, the coordinates of the four corners of the image to be inspected are subjected to pattern matching processing (four corner reference images prepared in advance for four small regions of several tens of pixels by several tens of pixels near the four corners of image data) The pattern can be extracted by performing a pattern matching process and specifying the coordinates of the four corners).

(S3: Background Image Difference Processing) The CPU 7a performs background image difference processing for removing defective luminance changes caused by things other than the liquid crystal light valve 2, such as illumination and lenses.
FIG. 3 is an explanatory diagram of the background image difference processing. This background image difference processing is performed by using the background image shown in FIG. 3B from the image (image before background difference processing) in which only the screen portion of the inspected portion of the inspection target image data shown in FIG. Is subtracted to create the background difference image shown in FIG. 3C, and the background difference image is a difference image for each pixel corresponding to the two images. The background image is created by capturing a plurality of samples having as few defects as possible, creating an averaged image, and extracting only the screen portion of the inspected part from the image.

(S4: Flattening process) The CPU 7a performs a flattening process on the background difference image subjected to the flattening process. This flattening process is a process of flattening a gradual change in luminance of the background difference image that has been subjected to the background image difference process, and is performed to remove the influence of unevenness over a relatively wide range. This flattening process is performed by a process applying a filter process or a morphology process.

(S5: Image Duplicating Process) The CPU 7a creates a duplicate of the image that has been flattened, and prepares for a horizontal edge detection process and a vertical edge detection process to be described later.
(S6: Horizontal edge detection process) The CPU 7a performs a horizontal edge detection process by a horizontal edge detection filter to create a horizontal line detection image in which the horizontal edge is emphasized. In this horizontal edge detection filter process, since it is difficult to detect white and black line defects at a minute level in a flattened image, the horizontal edge detection filter is applied to perform line defect enhancement calculation.
(S7: Vertical Edge Detection Process) The CPU 7a performs a vertical edge detection process using a vertical edge detection filter to create a vertical line detection image in which the vertical edges are emphasized. Also in this vertical edge detection filter processing, since it is difficult to detect white and black line defects at a minute level in a flattened image, the vertical edge detection filter is applied to perform line defect enhancement calculation.

  FIGS. 4A and 4B show examples of a horizontal edge detection filter for detecting a horizontal edge and a vertical edge detection filter for detecting a vertical edge, respectively. These horizontal / vertical edge detection filters detect whether or not there is an edge component in a small area of several pixels × several pixels including the pixel of interest. This is a filter in which each pixel is weighted so as to be emphasized by performing a convolution operation from the relationship between the luminance values of the pixel and surrounding pixels.

  4A and 4B, the image subjected to the horizontal / vertical edge detection filter has a positive gradation on one side and a negative gradation on the opposite side on both sides of the line (depending on the state of the edge). Appears as a value. In the image processing format, the image data can usually only have a positive value, so that the negative component is 0 as it is, and is excluded from the processing target. Therefore, so that both edges can be detected from the same image, when the screen is represented by 4096 gradations of 12 bits, a process of adding half the value of 2048 as an offset value is performed. As a result, a negative component also appears as a positive gradation, so that both edge components can be detected by a single filtering process. When the image format is an 8-bit gray scale, there are 256 gradations, and half of the 128 values are added to the filter processing result as an offset value.

(S8: Horizontal profile processing) The CPU 7a divides the horizontal line detection image subjected to the horizontal edge detection process into four parts in the vertical direction, and integrates the luminance values of the respective pixels in the divided areas in the horizontal direction to obtain integrated values. Perform horizontal profile processing.
(S9: Vertical profile processing) Similarly, the CPU 7a divides the vertical line detection image subjected to the vertical edge detection processing into four in the horizontal direction, and integrates the luminance values of the respective pixels in the divided areas in the vertical direction. Performs vertical profile processing to obtain integrated values. In the following description, the luminance value, the maximum value, and the minimum value are all targeted for the integrated value.
FIGS. 6A and 6B respectively show horizontal profile data obtained by horizontal profile processing and vertical profile data obtained by vertical profile processing.
As described above, the horizontal line detection image is divided into four in the vertical direction during the horizontal profile processing, and the vertical line detection image is divided into four in the vertical direction during the vertical profile processing. In the detection image or the vertical line detection image, there is a possibility that the thin line defect is buried due to the fluctuation of the brightness, and in the short line defect, other parts are integrated, and therefore the component of the line defect may be thinned. However, thin line defects are not affected by brightness fluctuations in small divided images, so thin line defects can be detected. This is because the component is not so thin and even a short line defect can be detected.

(S10: Calculation of statistical data) The CPU 7a obtains the average value (moving average), standard deviation, maximum value, and minimum value of the luminance values of the divided area screen from the integrated value of each row in each divided area subjected to the horizontal profile processing.
(S11: Statistical data calculation) The CPU 7a calculates the average value (moving average), standard deviation, maximum value, and minimum value of the luminance values of the divided area screen from the integrated value of each column in each divided area subjected to the vertical profile processing. Perform statistical data calculation processing.

Here, a method for calculating the average value (moving average) and the standard deviation of the luminance values of the divided area screen in the above processing (S10, S11) will be described.
FIG. 7 is an explanatory view conceptually showing a method of calculating an average value (moving average). The normal method of calculating the moving average is obtained by moving the area for which the average value is obtained. In this embodiment, in addition to this, the center position of the area is as shown in FIG. In addition, the defect determination value calculation region is excluded except for a predetermined width Δ. An average value Ave (x) is obtained based on the luminance value in the defect determination value calculation area. By performing the processing in this way, it is possible to eliminate the influence of a peak portion in the predetermined width Δ. As a result, as shown in FIG. 7B, the luminance value-average value Ave (x) is not affected by the peak portion. Note that there are cases where the peak portion is included in the defect determination value calculation region by moving the center position (x), but the peak portion is obtained by the horizontal edge detection process (S6) or the vertical edge detection process (S7). One line defect can rise and fall, and a maximum value and a minimum value appear. The appearance of the minimum value suppresses an increase in the average value of the peak portion, and temporarily assumes the center position (x Even if the average value in () increases, since the peak portion is not included in the center position (x), there is no significant influence on the subsequent processing (brightness value−average value Ave (x)). Therefore, the detection capability is improved as a whole.

  FIG. 8 is an explanatory diagram showing an example (comparative example) when the defect determination value calculation region is set over the entire area for obtaining the average value. As shown in FIG. 8A, when the defect determination value calculation region is set over the entire region, the peak portion is included, and the average value Ave is inevitably high. Therefore, the luminance value-average value Ave is included. When obtaining the peak portion A, the peak portion A appearing in FIG. 7B may be buried as shown in FIG. 8B.

Here, the process for obtaining the moving average value of FIG. 7 will be described in more detail with reference to FIG.
FIG. 9 is a diagram showing a line profile (original data). In this line profile, if the moving average value at an arbitrary point (x) is to be calculated, n pixels on both sides of the arbitrary point (x) are excluded from the calculation, and m pixels on both sides of each pixel are defective. As the determination value calculation area, the average value Ave (x) and the standard deviation σ (x) are obtained by the following equations, respectively. The average value Ave (x) and its standard deviation σ (x) are both functions of the pixel position (x), and the average value Ave (x) and its standard deviation σ are moved while moving the pixel position (x). (X) will be obtained respectively. In the following equation, x is the length direction of the divided elongated area, and y (= 1 to 4) is the unit of each divided area.

Moving average value Ave (x) = {p [x- (n + m), y] + p [x- (n + m-1), y]
+ P [x− (n + m−2), y] +... + P [x− (n + m− (m−1)), y]
+ P [x + (n + m− (m−1)), y] +... + P [x + (n + m−2), y]
+ P [x + (n + m-1), y] + p [x (n + m), y]} / 2 * m
... (1)
Standard deviation σ (x)
S ² (x) = 1 / (2 * m) {(p [x− (n + m), y] −Ave (x)) ²
+ (P [x− (n + m−1), y] −Ave (x)) ² +...
+ (P [x− (n + m− (m−1)), y] −Ave (x)) ²
+ (P [x + (n + m− (m−1)), y] −Ave (x)) ² +...
+ (P [x + (n + m−1), y] −Ave (x)) ²
+ (P [x + (n + m), y] −Ave (x) ² }
σ (x) = {S ² (x)} ^1/2 (2)

(S12: Defect candidate extraction processing) The CPU 7a calculates the average value Ave ₁ (x), standard deviation σ ₁ (x), maximum value and minimum value of the integrated value of each row of each divided region subjected to horizontal profile processing, and the vertical profile. average Ave ₂ of the integrated values of each column of processing each divided region, the standard deviation sigma ₂ (x), by respectively based on the maximum and minimum values the following equation, horizontal threshold (horizontal line) and the vertical line threshold ( (Vertical line) is obtained by calculation. In each equation, a ₁ and a ₂ are certain constants.
Horizontal line threshold (x) = Ave ₁ (x) ± a ₁ * σ ₁ (x)
Vertical line threshold (x) = Ave ₂ (x) ± a ₂ * σ ₂ (x)
As a calculation result of one expression, two values of a certain pixel position (x) are calculated as threshold values. The threshold value obtained by + becomes the threshold value of the edge emphasized to the plus side by the filter processing, and the division profile processing Of these images, those exceeding the threshold are detected as line defect candidates. Also, the threshold value obtained with-becomes the threshold value of the edge emphasized to the negative side by the filter processing, and the image below this threshold value is detected as a line defect candidate in the image subjected to the division profile processing.

  For example, for each divided region image that has been subjected to horizontal profile processing, if the maximum integrated value of the image exceeds the horizontal line threshold (+ calculation), the next evaluation value processing is performed as an image having horizontal line defect candidates. Proceed to (S14). If the minimum integrated value of the image does not exceed the horizontal line threshold (−calculation), the process proceeds to the next evaluation value process as an image having horizontal line defect candidates (S14). Therefore, if the maximum integrated value of the image does not exceed the horizontal line threshold (+ calculation) and the minimum integrated value exceeds the horizontal line threshold (−calculation), there is no horizontal line defect. Since it is a non-defective product (S13), evaluation value processing for evaluation value calculation described later is not performed.

  For each divided region image that has undergone vertical profile processing, the next evaluation value is determined as an image having a vertical defect candidate when the maximum integrated value of the image exceeds the vertical line threshold (+ calculation). It progresses to a process (S14). If the minimum integrated value of the image does not exceed the vertical line threshold (−calculation), the process proceeds to the next evaluation value process (S14) as an image having a vertical line defect candidate. Therefore, if the maximum integrated value of the image does not exceed the vertical line threshold (+ calculation) and the minimum integrated value exceeds the vertical line threshold (−calculation), a vertical line defect No evaluation value processing for evaluation value calculation which will be described later is performed.

(S13: non-defective product rank) When the CPU 7a determines that there is no horizontal or vertical line defect in the determination process (S12), the CPU 7a ranks the liquid crystal light valve to be inspected as a non-defective product. To do.

(S14: Calculation of evaluation value) The CPU 7a extracts, for each divided region image that has undergone horizontal profile processing, the maximum value of the integrated value of the image extracted as the horizontal line threshold (+ calculation) or more, and the integrated value of the image. For the horizontal line defect candidate extracted as not exceeding the horizontal line threshold (−calculation) and the vertical profile processed divided region images, the maximum integrated value of the images is the vertical line threshold. (+ Calculation) Extracted as above, and vertical line defect candidates extracted as the minimum integrated value of the image does not exceed the vertical line threshold (−calculation), each divided region Based on the average value Ave (x), standard deviation σ (x), maximum value and minimum value obtained from the image, the line evaluation value H (x) and the line evaluation value B (x ) Each seek. These are determined for each of the horizontal profile and the vertical profile.
Line evaluation value H (x) = (maximum value−Ave (x)) / σ (x)
Line evaluation value B (x) = (Ave (x) −minimum value) / σ (x)

(S15: Defect rank classification) The line evaluation value H (x) and the line evaluation value B (x) are set as threshold values that are line defect candidates but have no problem as non-defective products. The CPU 7a then extracts the maximum value in each divided region image that has been subjected to the horizontal profile processing when the line evaluation value H (x)> the threshold value or the line evaluation value B (x)> the threshold value. Is evaluated as having a level of horizontal line defects, and other than that, it is a horizontal line defect candidate, but it is evaluated as a level having no problem as a non-defective product. For the case where the maximum value in each divided region image subjected to vertical profile processing is extracted as being equal to or greater than the vertical line threshold value, when line evaluation value H (x)> threshold value or line evaluation value B (x)> threshold value, It is evaluated as a level having a vertical line defect, and other than that, it is a vertical line defect candidate, but it is evaluated as a level having no problem as a non-defective product.

10A, 10B, and 10C are characteristic diagrams showing, for example, a horizontal profile line profile (original data), data obtained by subtracting the average value Ave (x) from the line profile (original data), and an evaluation value. It is.
FIGS. 11A, 11B, and 11C show the same data as described above, but are examples (comparative examples) in the case where the average value of the entire in-plane is obtained without performing the moving average.
When comparing the data of FIG. 10 with the data of FIG. 11, the data of the portion B of the original data is close to zero in the evaluation value of FIG. 10C, and for example, the influence of shading is eliminated. The evaluation value in FIG. 11C is a large value, and it can be seen that this is a cause of erroneous detection.

  As described above, according to the first embodiment, as described above, the horizontal line threshold value and the vertical line defect candidate for determining the presence or absence of a horizontal line defect candidate from the average value and the standard deviation obtained by the statistical data calculation process. A vertical line threshold value for determining the presence or absence of a vertical line is set, and whether or not the maximum value of the integrated value of each divided area subjected to horizontal profile processing and vertical profile processing exceeds the horizontal line and vertical line threshold values for determining line defects is displayed on the screen. A primary determination is made as to whether or not there is a line defect, and whether or not the screen has a line defect depending on whether or not the minimum value of the integrated value of the screen exceeds the horizontal and vertical line thresholds for determining another line defect candidate. Thus, it is possible to easily determine whether each divided region image is a non-defective product having no white line or black line or a defect candidate having a white line or black line defect in a short time.

  For the defect candidate image determined to have a line defect, the line evaluation value H and the line evaluation are calculated according to a predetermined formula based on the average value, standard deviation, maximum value, and minimum value of the integrated values of the divided areas. Whether the evaluation value calculation process for calculating the value B is performed, and the line evaluation value H obtained by the evaluation value calculation process and the line evaluation value B are preset line defect candidates, but do not exceed a threshold value of a level at which there is no problem Since the defect rank of the line defect candidate is determined, it is possible to detect a thin line defect and a short line defect with high accuracy, and the evaluation value calculation target is only the defect candidate. Therefore, the calculation time is short, and the defect ranking can be performed in a short time.

  In addition, since the average value is obtained by moving average when obtaining the above average value, an appropriate average value can be obtained without being affected by the luminance of other parts (for example, the luminance value of the peak portion). Since this is possible, the subsequent arithmetic processing is also appropriate. Moreover, when calculating the moving average, the calculation is performed in a state in which the predetermined width of the central region is excluded, so that the influence when the peak value is included in the predetermined width can be eliminated. Furthermore, since the average value is expressed as a function of the pixel position, the standard deviation, threshold value, and evaluation value are also expressed as a function of the pixel position. For this reason, the threshold value and the evaluation value become dynamic (will change according to the pixel position). For example, when there is shading in the surface, the threshold value dynamically changes. Shading effects can be eliminated. Further, the influence of the maximum value and the minimum value due to the edge enhancement process, and the influence when the peak value is included in the average value can be suppressed by making the threshold dynamic.

Embodiment 2. FIG.
In the first embodiment, the example in which the original data is divided into four parts when obtaining the horizontal profile and the vertical profile has been described. However, the number of divisions is not limited to that number, and two, It may be 3, 5, or 6 or more. In the first embodiment, the inspection target is described as a liquid crystal light valve. However, display body parts such as a liquid crystal panel, a plasma display, an organic EL display, and a DMD (direct mirror device), and a display device using them.・ Applicable to product inspection. Furthermore, although the example in which the panel to be inspected is enlarged and displayed by the projector has been described, the display state of the panel to be inspected may be directly taken in via, for example, a microscope.

The block diagram which shows the structure of the line defect detection apparatus of Embodiment 1 of this invention. The flowchart for demonstrating operation | movement of a collinear defect detection apparatus. Explanatory drawing of a background image difference process. The figure which shows the example of an edge detection filter. The figure which shows the line defect detection image by which the edge detection filter process was carried out. The figure which shows the profile data by which the division profile process was carried out. Explanatory drawing which showed notionally the calculation method of an average value (moving average). Explanatory drawing which showed notionally the calculation method (comparative example) of an average value. The figure which showed the line profile (original data). The figure which showed the data of the process which calculates | requires an evaluation value. The figure which showed the data of the process which calculates | requires an evaluation value (comparative example).

Explanation of symbols

1 projector, 2 liquid crystal light valve, 3 screen, 4 image, 5 pattern generator, 6 CCD camera, 7 computer, 7a CPU, 8 display device, 10 screen to be inspected.

Claims (10)

Applying an edge detection filter to each of different directions in order to emphasize and detect line defects in an image obtained by imaging the panel to be inspected;
The luminance value of each pixel of the edge detection image in which the line defect is emphasized in different directions is divided into a plurality of rectangles, and the luminance value of each pixel is integrated in the direction in which the line defect is emphasized for each divided region. Performing a split profile process for obtaining
Obtaining an average value of integrated values in each divided region in which line defects are emphasized by a moving average, and calculating statistical data using the integrated value and the average value;
A step of setting a threshold value of an integrated value based on the statistical data, and extracting defect candidates from the statistical data and the threshold value;
A line defect detection method for a panel, comprising:   2. The line defect detection method for a panel according to claim 1, wherein the calculation processing for obtaining the moving average is performed by removing a central portion of a region for which the moving average is obtained from a calculation within a predetermined range. A background difference obtained by extracting an image portion of the inspected portion from the captured image and removing a difference between the extracted image of the inspected portion and the background image to remove a defective luminance change caused by other than the inspected portion. A step of pre-processing to create an image;
3. A step of performing a flattening process for removing the influence of unevenness in a background differential image, and the process proceeds to the step of applying the edge detection filter after the flattening process. Panel line defect detection method.   The step of applying the edge detection filter in the different directions includes a horizontal edge detection filter that performs horizontal edge enhancement processing to obtain a horizontal line detection image, and a vertical edge detection filter that performs vertical edge enhancement processing to obtain a vertical line detection image, respectively. The line defect detection method for a panel according to claim 1, wherein the line defect detection method is performed.   5. The step of applying an edge detection filter adds the median value of the gradation of the image as an offset value to the image at the same time as applying the edge detection filter. Panel line defect detection method.   The step of performing the division profile process divides the horizontal line detection image subjected to the edge detection process in the horizontal direction into a plurality of rectangles in the vertical direction, and integrates the luminance value of each pixel in each divided area in the horizontal direction to obtain an integrated value. The vertical line detection image that has been subjected to the horizontal profile processing and the vertical edge detection processing is divided into a plurality of rectangles in the horizontal direction, and the luminance value of each pixel in each divided area is integrated in the vertical direction. The method for detecting a line defect in a panel according to claim 1, further comprising a step of performing vertical profile processing for obtaining an integrated value.   Statistical data in the step of calculating the statistical data is the average value, standard deviation, maximum value, and minimum value obtained from the integrated values acquired from each divided region in the division profile step, and statistical data in the defect candidate extraction step. The threshold value set based on is calculated by a predetermined formula based on an average value and a standard deviation obtained from an integrated value acquired from each divided region by a divided profile process. The line defect detection method for a panel according to any one of the above.   For an image determined to have a line defect candidate, a line defect evaluation value is calculated by a predetermined formula based on the defect candidate statistical data, and whether or not the defect evaluation value exceeds a predetermined threshold value 8. The method for detecting a line defect in a panel according to claim 1, further comprising a step of determining whether or not a defect candidate is at a non-defective level.   The line defect evaluation value in the step of determining whether the product is at the non-defective level is predetermined based on the average value, standard deviation, maximum value, and minimum value obtained for each divided region of the image determined to have a line defect candidate. The line defect detection method for a panel according to claim 8, wherein the line defect detection method is calculated by the following formula. Imaging means for imaging the panel to be inspected;
An apparatus for detecting a line defect in a panel, comprising: an arithmetic unit that performs an arithmetic process of the line defect detection method for a panel according to claim 1.

Images