JP2011008482A - Defect detection method, defect detection device and defect detection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a defect detection method, a defect detection device and a defect detection program, for detecting a stripe defect with high accuracy.SOLUTION: The defect detection method includes: a defect enhancement processing step having an inspection pixel selection step of sequentially selecting a target pixel from an image obtained by imaging an object to be inspected, a comparison pixel group setting step of setting a comparison pixel group, a minimum luminance difference calculation step of obtaining a minimum luminance difference, and a defect enhancement value calculation step of obtaining a defect enhancement value; and a defect detection step having a defect candidate extraction step ST31 of extracting a defect candidate pixel from the defect enhancement value, a circumscribing quadrangle setting step ST33 of detecting an inspection circumscribing quadrangle that circumscribes a defect candidate pixel group, a dimensional ratio calculation step ST34 of calculating a dimensional ratio of a short side to a long side of the inspection circumscribing quadrangle, an area ratio calculation step ST35 of calculating an area ratio of the defect candidate pixel group to the inspection circumscribing quadrangle, and a defect specification step ST36 of specifying a stripe defect based on the dimensional ratio and the area ratio.

Description

  The present invention relates to a defect detection method, a defect detection apparatus, and a defect detection program for detecting a streak-like defect of an inspection object by processing an image obtained by imaging the inspection object.

  Conventionally, in the defect detection of foreign matters and scratches on the wiring sealing surface of a flexible printed circuit board, defect detection is performed using a filter process which is one of image processing techniques (see, for example, Patent Documents 1 and 2).

In the inspection method described in Patent Document 1, noise is removed from a captured image of an inspection object using a smoothing filter, and a streaky uneven defect is detected by using a primary differential filter.
In the image inspection method described in Patent Document 2, an image obtained by binarizing a predetermined luminance threshold after applying a second-order differential filter to a captured image and a binary image of a transmission image are binarized with respect to the predetermined luminance threshold. A secondary feature amount of the image or an image obtained by synthesizing these binarized images is calculated, and a defect is detected from the calculated secondary feature amount.

  In addition, the applicant obtains the difference in luminance value between the inspection target point and a plurality of comparison target points arranged around the inspection target point, and among the luminance value difference data, the one with the maximum value is the inspection target point. A defect detection method for extracting defect candidates based on the defect enhancement value is proposed (see Patent Document 3).

JP 2006-189293 A JP 2007-309679 A JP 2007-285753 A

However, the defect detection method using the first derivative as described in Patent Document 1 has a problem that it is difficult to selectively detect only a streak-like defect because even a spot-like defect is detected. Further, in the defect detection method of Patent Document 2, defect detection is performed from an image obtained by binarizing a captured image, and an image obtained by applying a secondary differential filter to the captured image and then binarizing the image. When the differential filter is used, there is a risk of erroneous detection due to the influence of shading, and there is a problem that the defect detection accuracy deteriorates.
On the other hand, in the defect detection method of Patent Document 3 proposed by the present applicant, a linear streak defect can be detected well, but when the angle changes within the defect, it is not detected as a streak defect. In some cases, it is necessary to provide a separate detection method. Also, since both streak-like defects and spot-like defects are detected, it is difficult to distinguish them.

  In view of the above problems, an object of the present invention is to provide a defect detection method, a defect detection apparatus, and a defect detection program for accurately detecting streak defects.

  The defect detection method of the present invention includes a defect enhancement process step of performing defect enhancement processing on a captured image obtained by imaging an inspection object using a defect enhancement filter, and defect enhancement of each pixel obtained in the defect enhancement processing step. A defect detection step for detecting a defect based on a value, and the defect enhancement processing step includes: an inspection target pixel selection step for sequentially selecting inspection target pixels with respect to the captured image; and a selected inspection target pixel A plurality of comparison target pixels separated by a predetermined distance from the center of the pixel are arranged in a substantially circular shape around the inspection target pixel, and a pair of these comparison target pixels are arranged at point-symmetric positions with respect to the inspection target pixel. The comparison target pixel group setting step for setting a plurality of comparison target pixel groups with the comparison target pixel as a set, and the difference between the luminance value of each comparison target pixel included in the comparison target pixel group and the luminance value of the inspection target pixel Ah The luminance difference data is obtained, and among the luminance difference data, a minimum luminance difference calculation step for obtaining a minimum luminance difference with a minimum value for each comparison target pixel group, and a minimum luminance difference calculated for each comparison target pixel group. A defect enhancement value calculating step in which a minimum luminance difference having a maximum value is a defect enhancement value of the inspection target pixel, and the defect detection step has the defect enhancement value equal to or greater than a predetermined defect threshold value. A defect candidate extraction step of extracting pixels as defect candidate pixels, and the defect candidate pixels adjacent to each other as a defect candidate pixel group, and among circumscribed rectangles circumscribing the defect candidate pixel group, a circumscribed rectangle having the smallest area is inspected A circumscribed rectangle setting step for setting as a circumscribed rectangle, a long side dimension of the inspection circumscribed rectangle along the inertial axis direction of the defect candidate pixel, and the inspection circumscribed line along the width direction orthogonal to the inertial axis direction Measuring the short side dimension of the shape, calculating a ratio of the short side dimension to the long side dimension, and calculating an area ratio of the defect candidate pixel group to the inspection circumscribed rectangle When the calculation step and the dimensional ratio calculated in the dimensional ratio calculation step are less than a predetermined dimensional ratio threshold, or the area ratio calculated in the area ratio calculation step is less than a predetermined area ratio threshold And a defect specifying step of specifying the defect candidate pixel group as a streak-like defect.

In the present invention, in the defect enhancement processing step using the defect enhancement filter, the selected inspection target pixel and a plurality of comparison target pixels arranged in a substantially circular shape around the selected inspection target pixel are set. Then, the comparison target pixels are divided into a plurality of comparison target pixel groups that are a set of a pair of comparison symmetric pixels arranged at point-symmetric positions with respect to the inspection symmetric pixels, and are included in these comparison target pixel groups. From the luminance difference data between the comparison target pixel and the inspection target pixel, a minimum luminance difference having a minimum value is selected, and among the minimum luminance differences selected for each of the plurality of comparison target pixel groups, the value is The smallest minimum luminance difference is used as the defect enhancement value of the inspection target pixel. As a result, it is possible to detect a defect that includes the inspection target pixel and does not include any of the comparison target pixel groups. The defect emphasis filter of the present invention indicates an algorithm for performing the defect emphasis process, that is, an inspection target pixel selection process, a comparison target pixel group setting process, a minimum luminance calculation process, and a defect emphasis value calculation process. This process is performed for each pixel of the captured image, and only a certain spatial frequency component is selectively left as an emphasis value and other spatial frequency components are suppressed.
In such a defect emphasis processing step, when there is no defect in the pixel portion to be inspected and there is no luminance difference with surrounding pixels, the minimum luminance difference becomes a very small value. Further, even if the inspection target pixel has a defect, if the defect extends to any of the comparison target pixel portions, there is almost no luminance difference between the inspection target pixel and the comparison target pixel included in the defective portion. The minimum luminance difference is also a very small value. On the other hand, when the inspection target pixel has a defect and any of the surrounding comparison target pixel groups have no defect, the luminance value of the inspection target pixel is the luminance value of any comparison target pixel of the comparison symmetric pixel group. Therefore, the minimum luminance difference is a relatively large value. As a result, when there is a defect having a size that fits inside one of the comparison target pixel groups, the minimum luminance difference becomes a relatively large value, and the defect is emphasized.
Further, at this time, the plurality of comparison target pixels are divided into a plurality of comparison target pixel groups in which one set of comparison target pixels is arranged in a point-symmetrical position with respect to the inspection target pixel. Also, streak-like defects can be detected appropriately. That is, when the comparison target pixel is not divided into a plurality of comparison target pixel groups, and there is a streak defect that overlaps at least one of the plurality of comparison target pixels and the detection target pixel, each pixel on the streak defect Since the difference in luminance value is small, the minimum luminance difference is also a small value, and a defect cannot be detected. On the other hand, when the comparison target pixels arranged in a plurality of substantially circular shapes are divided into a plurality of comparison target pixel groups and the minimum luminance difference is calculated for each comparison target pixel group as in the present invention, Because the pixel positions are different, the comparison target pixel of one comparison target pixel group overlaps the comparison target pixel, and even if the streak defect cannot be detected, the comparison target pixel of the other comparison target pixel group is not the streak defect. The streak defect can be detected without overlapping.

As described above, by using the defect emphasis value, it is possible to emphasize both a spot-like defect and a streak-like defect, and a defect portion can be extracted with high accuracy. In the present invention, in the defect detection step, defect candidate pixels are picked up based on the defect enhancement value detected as described above. Then, defect candidate pixels adjacent to each other are defined as a defect candidate pixel group, and the size ratio and the area ratio are calculated based on the inspection circumscribed rectangle circumscribing these defect candidate pixels. Here, when the area ratio is less than the area ratio threshold, or when the dimension ratio is less than the dimension ratio threshold, it is detected as a streak defect.
That is, in the straight streak-like defect, the inertial axis direction and the long side direction of the inspection circumscribed rectangle substantially coincide with each other, so that the short side dimension of the inspection circumscribed rectangle and the width dimension of the streak defect are substantially the same dimension. . On the other hand, in the case of a spot-like defect, since the width dimension becomes large, the dimension ratio of the short side dimension to the long side dimension of the inspection circumscribed rectangle becomes a value close to 1. On the other hand, in the stripe defect, the width dimension is small, and the dimensional ratio of the short side to the long side of the inspection circumscribed rectangle is also small, for example, a value close to 0. That is, by setting a value close to 0 as the dimension ratio threshold value, it is possible to detect a thin, substantially linear stripe-like defect.
In addition, the streak defect is not limited to a straight defect, but a broken line defect, a curved defect, an annular defect, or the like whose angle changes in the middle is also conceivable. In this case, the short side of the inspection circumscribed rectangle The length dimension also increases. For such streak defects, the area ratio of the defect candidate pixel group to the inspection circumscribed rectangle is used. That is, in the case of a spot-like defect, the area ratio of the defect candidate pixel group to the inspection circumscribed rectangle is large and is a value close to 1. On the other hand, the streak-like defect is linear and has a small area, so that the area ratio is close to zero. Therefore, a defect candidate pixel group whose area ratio is smaller than the area ratio threshold can be detected as a streak defect. As a result, only the streak defect can be detected well from the defect candidate pixel group in which the spot-like defect and the streak defect are mixed.

  In the defect detection method of the present invention, the minimum luminance difference calculation step includes a light defect from light defect luminance difference data obtained by subtracting the luminance value of each comparison target pixel included in the comparison target pixel group from the luminance value of the inspection target pixel. Bright defect minimum brightness difference calculating step for calculating a minimum brightness difference, and dark defect minimum from dark defect brightness difference data obtained by subtracting the brightness value of the inspection target pixel from the brightness value of each comparison target pixel included in the comparison target pixel group A dark defect minimum luminance difference calculation step for calculating a luminance difference, and the defect enhancement value calculation step includes a bright defect enhancement value and a dark defect enhancement based on the bright defect minimum luminance difference and the dark defect minimum luminance difference. It is preferable that a value is obtained, and the defect detection step detects a streak-like bright defect and a streak-like dark defect based on the bright defect enhancement value and the dark defect enhancement value, respectively.

In the present invention, the minimum luminance difference calculation step inspects the bright defect minimum luminance difference, which is the minimum value of the luminance value data obtained by subtracting the luminance value of the comparison target pixel from the luminance value of the inspection target pixel, and the luminance value of the comparison target pixel. A dark defect minimum luminance difference, which is the minimum value of the luminance value data obtained by subtracting the luminance value of the target pixel, is obtained, and the subsequent steps are performed for each of the light defect minimum luminance difference and the dark defect minimum luminance difference.
When only the value obtained by subtracting the luminance value of the comparison target pixel from the luminance value of the inspection target pixel is used when calculating the minimum luminance difference, a dark defect cannot be detected, and from the luminance value of the comparison target pixel When only the value obtained by subtracting the luminance value of the pixel to be inspected is used, a bright defect cannot be detected. On the other hand, although a method of obtaining a defect enhancement value using the absolute value of the difference between the luminance value of the inspection target pixel and the luminance value of the comparison target pixel is also conceivable, in this case, it is difficult to distinguish between a bright defect and a dark defect, For example, when it is desired to detect only bright defects or only dark defects, it is necessary to provide a separate process for discriminating these. On the other hand, in the present invention, in the minimum luminance difference calculating step, the light defect minimum luminance difference and the dark defect minimum luminance difference are calculated, respectively, and for these light defect minimum luminance difference and dark defect minimum luminance difference, By performing the defect emphasis value calculation step and the defect detection step, it is possible to clearly detect bright defects and dark defects. Therefore, even when it is desired to selectively detect only one of the bright defect and the dark defect, the defect detection process can be easily performed.

  In the defect detection method of the present invention, it is preferable that the defect enhancement processing step performs the defect enhancement processing using a plurality of defect enhancement filters having different distances from the inspection target pixel to the comparison target pixel.

  In the present invention, in the comparison pixel setting step, the comparison target pixel that is arranged in a substantially circular shape is set apart from the inspection target pixel by a predetermined pixel. In such a method, since a plurality of defect emphasis filters having different distances from the inspection target pixel to the comparison target pixel are used, it is possible to detect stripe-shaped defects of various sizes. In addition, by synthesizing the detected streak defects based on a plurality of defect emphasis filters set by changing the distance between the inspection target pixel and the comparison target pixel, each size streak defect is, for example, one A defect detection result image can be displayed together. In addition, since the defect emphasis value for a larger number of comparison target pixels is calculated, it is possible to extract a defect with higher accuracy and further improve the reliability of defect detection.

  In the defect detection method of the present invention, the defect emphasis processing step is a defect in which a distance from the inspection target pixel to the comparison target pixel is set according to a width dimension of the streak defect detected by the defect detection step. It is preferable that the defect enhancement process is performed using an enhancement filter.

In the present invention, the defect enhancement process is performed using a defect enhancement filter in which the distance from the inspection target pixel to the comparison target pixel is set in accordance with the width dimension of the streak defect detected as a defect.
That is, when the comparison target pixel is selected using a defect enhancement filter in which the distance from the inspection target pixel to the comparison target pixel is set to a predetermined distance (comparison pixel selection distance), that is, twice the comparison pixel selection distance, A streak defect having a line width larger than the distance between the comparison target pixels facing each other across the inspection target pixel is not detected, and a streak defect smaller than the distance between the comparison target pixels is detected. Therefore, by performing defect emphasis processing using a defect emphasis filter in which a comparison pixel selection distance is set according to the line width of the line-shaped defect to be detected, it is possible to accurately detect a line-shaped defect having a desired line width. Can do.

  At this time, in the defect enhancement processing step, the distance from the inspection target pixel to the comparison target pixel is a distance larger by a predetermined pixel than the width dimension of the streak defect detected by the defect detection step. It is preferable to carry out the defect enhancement process using a defect enhancement filter.

  In the present invention, it is preferable to use a defect enhancement filter in which the comparison pixel selection distance is set to be, for example, 2 to 3 pixels larger than the detected line width of the streak-like defect. By using such a defect emphasis filter, it is possible to prevent a detection defect or overdetection of a streak-like defect having a target line width, and to perform highly accurate defect detection.

  The defect detection apparatus of the present invention includes defect enhancement processing means for performing defect enhancement processing on a captured image obtained by imaging an inspection object using a defect enhancement filter, and defect enhancement of each pixel obtained in the defect enhancement processing step. Defect detection means for detecting a defect based on a value, wherein the defect enhancement processing means selects inspection target pixels for sequentially selecting inspection target pixels for the captured image, and selected inspection target pixels A plurality of comparison target pixels separated by a predetermined distance from the center of the pixel are arranged in a substantially circular shape around the inspection target pixel, and a pair of these comparison target pixels are arranged at point-symmetric positions with respect to the inspection target pixel. A comparison target pixel group setting unit that sets a plurality of comparison target pixel groups each including a comparison target pixel, and a difference between a luminance value of each comparison target pixel included in the comparison target pixel group and a luminance value of the inspection target pixel Ah The luminance difference data is obtained, and among the luminance difference data, the minimum luminance difference calculating means for obtaining the smallest luminance difference with the smallest value for each comparison target pixel group, and the minimum luminance difference calculated for each comparison target pixel group A defect enhancement value calculation unit that uses a minimum luminance difference having a maximum value as a defect enhancement value of the inspection target pixel, and the defect detection unit has the defect enhancement value equal to or greater than a predetermined defect threshold value. A defect candidate extraction unit that extracts pixels as defect candidate pixels and the defect candidate pixels adjacent to each other as a defect candidate pixel group, and a circumscribed rectangle having a minimum area among the circumscribed rectangles circumscribing the defect candidate pixel group is inspected A circumscribed rectangle setting means for setting as a circumscribed rectangle, a long side dimension of the inspection circumscribed rectangle along the inertial axis direction of the defect candidate pixel, and the inspection circumscribed line along the width direction orthogonal to the inertial axis direction Dimension ratio calculating means for measuring the short side dimension of the shape and calculating a dimensional ratio which is a ratio of the short side dimension to the long side dimension, and an area ratio for calculating the area ratio of the defect candidate pixel group to the inspection circumscribed rectangle When the dimensional ratio calculated in the calculating means and the dimensional ratio calculating step is less than a predetermined dimensional ratio threshold, or the area ratio calculated in the area ratio calculating step is less than a predetermined area ratio threshold And defect specifying means for specifying the defect candidate pixel group as a streak-like defect.

  The present invention is a defect detection apparatus that performs the above-described defect detection method, can enjoy the same operations and effects as the defect detection method, and extract defect candidate pixels with high accuracy based on the defect enhancement value. From these defect candidate pixels, only streak defects can be detected with high accuracy.

  The defect detection program of the present invention is a defect detection program that is read and calculated by a calculation means, and the defect detection program uses a defect enhancement filter for a captured image obtained by imaging an inspection object. A defect enhancement processing unit that performs processing, and a defect detection unit that detects a defect based on a defect enhancement value of each pixel obtained in the defect enhancement processing step, and the defect enhancement processing unit adds to the captured image. On the other hand, the inspection target pixel selection means for sequentially selecting the inspection target pixels, and a plurality of comparison target pixels that are separated by a predetermined distance from the center of the selected inspection target pixels are arranged in a substantially circular shape around the inspection target pixels. Comparison target image for setting a plurality of comparison target pixel groups, each of which is a set of a pair of comparison target pixels arranged at point symmetry with respect to the inspection target pixel among the comparison target pixels A luminance setting data which is a difference between a luminance value of each comparison target pixel included in the comparison target pixel group and a luminance value of the inspection target pixel, and a value of the luminance difference data is the smallest A minimum luminance difference calculating means for obtaining a minimum luminance difference for each comparison target pixel group, and among the minimum luminance differences calculated for each comparison target pixel group, the minimum luminance difference having the maximum value is determined as a defect of the inspection target pixel. A defect enhancement value calculating unit that sets an enhancement value, and the defect detection unit is adjacent to a defect candidate extraction unit that extracts a pixel having the defect enhancement value equal to or higher than a predetermined defect threshold as a defect candidate pixel. The defect candidate pixel is defined as a defect candidate pixel group, and a circumscribed rectangle setting unit that sets a circumscribed rectangle having a minimum area as a test circumscribed rectangle among circumscribed rectangles circumscribing the defect candidate pixel group; A dimension which is a ratio of a short side dimension to a long side dimension while measuring a long side dimension of the inspection circumscribed square along the axial direction and a short side dimension of the inspection circumscribed square along a width direction orthogonal to the inertial axis direction. A size ratio calculating means for calculating a ratio, an area ratio calculating means for calculating an area ratio of the defect candidate pixel group with respect to the inspection circumscribed rectangle, and a size ratio calculated in the size ratio calculating step is a predetermined size ratio threshold value Defect specifying means for specifying the defect candidate pixel group as a streak-like defect when the area ratio is less than or less than a predetermined area ratio threshold value. It is characterized by.

  The present invention is a program used to implement the above-described defect detection method, and can enjoy the same operations and effects as those of the above-described defect detection method, and extract defect candidate pixels with high accuracy based on the defect emphasis value. From these defect candidate pixels, only streak defects can be accurately detected.

The block diagram which shows the structure of the defect detection apparatus which concerns on one embodiment of this invention. The block diagram which shows schematic structure of a control apparatus. The figure which shows an example of the comparison object pixel arrangement | positioning of a defect emphasis filter. The figure which shows the test object pixel and comparison object pixel which are set on the image data of a captured image. The figure which shows an example of a linear stripe defect. The figure which shows an example of a broken line-like stripe defect. The figure which shows an example of a cyclic | annular stripe-like defect. The figure which shows an example of a spot-like defect. The flowchart which shows operation | movement of the defect detection apparatus of this Embodiment. The flowchart which shows the process of a defect emphasis processing process. The flowchart which shows the process of a defect detection process. The figure which shows an example of a captured image. The figure which shows the bright defect emphasis image with respect to the captured image of FIG. The figure which shows the dark defect emphasis image with respect to the captured image of FIG. The figure which shows the bright defect specific image with respect to FIG. The figure which shows the dark defect specific image with respect to FIG.

[Configuration of defect detection device]
FIG. 1 is a block diagram showing a configuration of a defect detection apparatus according to an embodiment of the present invention.
The defect detection apparatus 100 according to the present embodiment detects a defect of an inspection object 1 such as a flexible substrate, a liquid crystal panel (TFT panel), or a semiconductor wafer. The inspection object 1 is placed on the XY stage 2 and configured to be movable in a plane.
The defect detection device 100 includes a microscope 4, a CCD camera 5, a control device 6, and a display device 7.

The microscope 4 is provided for enlarging the inspection object 1 and photographing it with the CCD camera 5, and a microscope having a sufficient magnification for detecting a defect of the inspection object 1 is used.
The CCD camera 5 is an imaging unit that images the inspection object 1 through the microscope 4.
The control device 6 is an image processing unit that controls the CCD camera 5 and detects the inspection object 1. The display device 7 is a display device such as a liquid crystal display connected to the control device 6.

FIG. 2 is a block diagram illustrating a schematic configuration of the control device 6.
For example, a personal computer or the like can be used as the control device 6, and a captured image captured by the CCD camera 5 is processed to detect a defect in the inspection object 1. As shown in FIG. 2, the control device 6 includes an image input unit 60, a defect enhancement processing unit 61, and a defect detection unit 62. In the present embodiment, the control device 6 includes storage means such as an HDD and a memory, and defect emphasis processing means 61 and defect detection means 62 are stored as programs recorded in the storage means. Then, a CPU (Central Processing Unit) provided in the control device 6 reads programs such as the defect emphasis processing means 61 and the defect detection means 62 from the storage means, and a defect detection process is executed by arithmetic processing.
In the present embodiment, as described above, the defect emphasis processing means 61 and the defect detection means 62 are stored in the storage means as programs, and a configuration in which the processing is executed by being appropriately read out by the CPU is illustrated. It is not limited to this. That is, the defect emphasis processing means 61 and the defect detection means 62 are configured by an integrated circuit such as an IC chip, and may be configured to detect defects by appropriately processing image data of a captured image input from the image input means 60. Good.

  The image input means 60 receives image data of a captured image taken from the CCD camera 5 and stores the image data in a storage means (not shown).

  The defect enhancement processing unit 61 performs a defect enhancement processing step of performing defect enhancement processing on the acquired image using a defect enhancement filter. The defect enhancement processing unit 61 and the comparison target pixel group setting are performed. Means 612, minimum luminance difference calculation means 613, and defect enhancement value calculation means 614 are provided.

The inspection target pixel selection unit 611 performs an inspection target pixel selection step of sequentially selecting inspection target pixels in the captured image.
The comparison target pixel group setting unit 612 has a substantially circular shape around the inspection target pixel. The comparison target pixel is separated from the center of the inspection target pixel selected by the inspection target pixel selection unit 611 by a predetermined distance (comparison pixel selection distance). A comparison target pixel group setting step is performed in which a plurality of comparison target pixels are arranged and set in a plurality of comparison target pixel groups. Alternatively, the comparison target pixel may be selected using a defect enhancement filter configuration in which the positional relationship between the inspection target pixel and the comparison target pixel is obtained in advance. FIG. 3 is a diagram illustrating an example of a defect enhancement filter configuration in which the comparison target pixel selection distance is set to 7 pixels. As shown in FIG. 3, the defect enhancement filter configuration 70 includes a central pixel 71 corresponding to the inspection target pixel and a plurality of comparison target pixels arranged in a substantially annular shape at a position separated from the central pixel 71 by 7 pixels. And a comparison setting pixel 72 corresponding to. The comparison target pixel group setting unit 612 then sets the defect enhancement filter configuration 70 so that the inspection target pixel O ₁ (see FIG. 4) selected by the inspection target pixel selection unit 611 from the captured image and the center pixel 71 coincide with each other. And pixels that overlap with the comparison setting pixel 72 on the captured image may be set as comparison target pixels S _{1 to} S ₅₆ (see FIG. 4).

In FIG. 3, as the defect enhancement filter configuration 70, the comparison pixel selection distance from the center pixel 71 to the comparison setting pixel 72 is set as 7 pixels. It is set as appropriate according to the line width of the defect. That is, the storage unit stores a plurality of defect enhancement filter configurations 70 having different comparison pixel selection distances, and the comparison target pixel group setting unit 612, for example, performs defect enhancement according to the setting items set and input by the inspector. The comparison target pixel is set by switching the filter configuration 70. For example, in the comparison target pixel selected by the defect enhancement filter configuration 70 as illustrated in FIG. 3, the distance between a pair of comparison target pixels that constitute one comparison target pixel group and that are opposed to each other with the inspection target pixel interposed therebetween is 14. In this case, the streak defect having a line width of 12 pixels or less can be accurately detected. At this time, if the inspector inputs a request signal for detecting a stripe-like defect having a narrower line width, for example, a setting input for detecting a stripe-like defect having a line width of about 5 pixels, the defect enhancement filter The configuration 70 reads out the defect enhancement filter configuration 70 in which the distance between the comparison target pixels is about 7 to 8 pixels, that is, the comparison pixel selection distance is 3 to 4 pixels, and the comparison target pixel is set by the defect enhancement filter configuration 70. .
The comparison target pixel group setting unit 612 divides the set comparison target pixel into a plurality of comparison target pixel groups. In the present embodiment, the comparison target pixel group setting unit 612 sets a comparison target pixel group that includes a pair of comparison target pixels that are arranged at positions that are point-symmetric with respect to the inspection target pixel. For example, when the defect enhancement filter configuration 70 as shown in FIG. 3 is used, 56 comparison target pixels as shown in FIG. 4 are set for the inspection target pixel. In this case, the comparison target pixel S _{n + 28} (where n = 1, 2, 3... 28) having a point-symmetrical positional relationship across the comparison target pixel _Sn and the inspection target pixel is set as the comparison target pixel group.

The minimum luminance difference calculating unit 613 obtains luminance difference data that is a difference between the luminance value of each comparison target pixel included in the comparison target pixel group and the luminance value of the inspection target pixel, and among the luminance difference data, A minimum luminance difference calculating step for obtaining a minimum luminance difference with a minimum value for each comparison target pixel group is performed.
Specifically, the minimum luminance difference calculating unit 613 calculates a bright defect minimum luminance difference and a dark defect minimum luminance difference for each comparison target pixel group.
In calculating the bright defect minimum luminance difference, first, the bright defect luminance difference F _wsn is calculated based on the following equations (1) and (2).

Thereafter, the minimum luminance difference calculation means 613 calculates the bright defect minimum luminance difference D as a smaller one of the bright defect luminance differences F _wsn calculated by the equations (1) and (2) according to the following equation (3). _Calculate as _wsn .

On the other hand, the dark defect minimum luminance difference is calculated by calculating the dark defect luminance difference F _bsn based on the following equations (4) and (5).

Then, the minimum luminance difference calculation unit 613 calculates the dark defect minimum luminance difference D _{bsn from} the dark defect luminance differences F _bsn calculated by the equations (4) and (5) according to the following equation (6). Calculate as

As shown in Expression (7), the defect enhancement value calculation unit 614 has a maximum value among the bright defect minimum luminance difference D _wsn and the dark defect minimum luminance difference D _bsn calculated for each comparison target pixel group. To obtain a bright defect enhancement value R1 and a dark defect enhancement value R2, respectively.

  The defect detection means 62 extracts defect candidate pixels from the bright defect enhancement value R1 and the dark defect enhancement value R2 that are the results processed by the defect enhancement processing means 61, and only the streak-like defects are extracted from these defect candidate pixels. Is detected. As shown in FIG. 2, the defect detection unit 62 includes a defect candidate extraction unit 621, a circumscribed rectangle setting unit 622, a size ratio calculation unit 623, an area ratio calculation unit 624, and a defect specification unit 625. ing.

  The defect candidate extraction unit 621 performs a defect candidate extraction step of extracting defect candidate pixels by comparing the defect enhancement values R1 and R2 calculated by the defect enhancement processing unit 61 with a predetermined threshold. As this threshold value, a bright defect threshold value wslevel and a dark defect threshold value bslevel as shown in the following equations (8) and (9) are set and compared with the bright defect emphasis value R1 and the dark defect emphasis value R2, respectively. A defect candidate pixel and a dark defect candidate pixel are extracted.

  In the above formulas (8) and (9), avr (bright) and avr (dark) are the average value of the bright defect enhancement value R1 and the average value of the dark defect enhancement value R2, respectively, and σ (bright) and σ (dark) ) Are standard deviations of the bright defect emphasis value R1 and the dark defect emphasis value R2, and α1, α2, β1, and β2 are arbitrary constants, and are determined according to the state of the captured image to be inspected. Further, there are pixels in which the bright defect enhancement value R1 and the dark defect enhancement value R2 are negative, but when the bright defect enhancement value R1 is a negative value, it indicates that the defect is a dark defect, and the dark defect enhancement value R2 When N is a negative value, it indicates a bright defect. Therefore, in the calculation of avr (bright), avr (dark) and σ (bright), σ (dark), these defect emphasis values R1, R2 Each threshold value is calculated by omitting the negative value of.

  Alternatively, the defect candidate pixel extraction processing by the defect candidate extraction unit 621 may be performed after applying a median filter or the like to the image processed by the defect enhancement processing unit 61 to perform noise removal processing.

  When there are defect candidate pixels in adjacent pixels among the defect candidate pixels extracted by the defect candidate extraction unit 621, the circumscribed rectangle setting unit 622 connects the defect candidate pixel group 81 (see FIG. 5). To (see FIG. 8). Here, the circumscribed rectangle setting unit 622 inspects, for example, eight pixels around the defect candidate pixel, and recognizes a defect candidate pixel adjacent to each other when there is a defect candidate pixel in the eight pixels. To detect.

Then, the circumscribed rectangle setting unit 622 performs a circumscribed rectangle setting step of setting an inspection circumscribed rectangle 82 circumscribed to the detected defect candidate pixel group 81.
Specifically, the circumscribed rectangle setting unit 622 detects outer peripheral defect candidate pixels arranged on the outer peripheral portion of the defect candidate pixel group 81, and detects a circumscribed rectangle that is in contact with at least three points of the outer peripheral defect candidate pixels. Then, the circumscribed rectangle setting unit 622 sets the circumscribed rectangle having the smallest area among these detected circumscribed rectangles as the inspection circumscribed rectangle 82. Here, one direction along the long side of the inspection circumscribed rectangle 82 is the inertial principal axis direction of the defect candidate pixel group.

The dimension ratio calculation means 623 measures the long side length dimension l and the short side length dimension h of the inspection circumscribed square 82 set by the circumscribed square setting means 622, and measures the short side dimension with respect to the long side dimension. A dimension ratio calculating step for calculating a dimension ratio M1 (= h / l), which is a ratio of
The area ratio calculation unit 624 measures the area of the defect candidate pixel group 81 in the inspection circumscribed rectangle 82 set by the circumscribed rectangle setting unit 622. For example, the number m of defect candidate pixels forming the defect candidate pixel group 81 is measured and calculated by multiplying the area a by one pixel. Then, the area ratio calculating means 624 performs an area ratio calculating step of calculating the area ratio M2 (= ma / lh) of the defect candidate pixel group 81 with respect to the inspection circumscribed rectangle 82.

The defect identification unit 625 identifies the streak defect by comparing the dimension ratio M1 and the area ratio M2 calculated by the dimension ratio calculation unit 623 and the area ratio calculation unit 624 with the dimension ratio threshold N1 and the area ratio threshold N2. Perform the defect identification process.
Here, FIGS. 5 to 8 show examples of the defect candidate pixel group 81. FIG. 5 is a diagram illustrating an example of a linear streak defect. FIG. 6 is a diagram illustrating an example of a broken line-like streak defect. FIG. 7 is a diagram illustrating an example of an annular streak defect. FIG. 8 is a diagram illustrating an example of a spot-like defect.
In a linear streak defect as shown in FIG. 5, the ratio of the length dimension in the width direction perpendicular to the inertial principal axis direction to the length dimension along the inertial principal axis direction, that is, the long side of the inspection circumscribed rectangle 82 The dimension ratio M1 of the short side with respect to is a small value (M1≈0). On the other hand, in a spot-like defect as shown in FIG. 8, the dimension ratio M1 of the short side to the long side of the inspection circumscribed rectangle 82 is large (for example, M1≈1 in FIG. 8). Therefore, the defect specifying means 625 determines whether or not the defect candidate pixel group 81 is a linear streak defect by determining whether or not the dimension ratio M1 is smaller than the predetermined dimension ratio threshold N1. Is possible.

Further, in the case of curved streaky defects, broken line shapes, and annular streaky defects, when the curvature is small or the bending angle is small, the dimension ratio M1 is sufficiently small as in the case of the straight streaky defects described above. As shown in FIG. 6, when the dimensional difference between the long side and the short side of the inspection circumscribed square 82 is small, the value of the dimensional ratio M1 is large. In such a case, even if it is a streak-like defect only with the dimensional ratio M1, it is not detected as a spot-like defect. On the other hand, in the defect detection apparatus 100 of the present embodiment, the defect identification unit 625 also performs defect detection based on the area ratio.
That is, in the spot-like defect as shown in FIG. 8, the area ratio M2 of the defect candidate pixel group 81 with respect to the inspection circumscribed rectangle 82 is large, for example, M2≈1 in FIG. On the other hand, in the case of a streak-like defect on a curve as shown in FIG. 6 or an annular streak-like defect as shown in FIG. 7, the area ratio M2 of the defect candidate pixel group 81 to the inspection circumscribed rectangle 82 is a sufficiently small value ( M2≈0). Therefore, the defect specifying unit 625 determines whether or not the defect candidate pixel group 81 is a streak defect having an angle change by determining whether or not the area ratio M2 is smaller than the predetermined area ratio threshold N2. It becomes possible.
As described above, the defect specifying unit 625 applies the defect candidate pixel group 81 when the dimension ratio M1 is smaller than the dimension ratio threshold N1 or when the area ratio M2 is smaller than the area ratio threshold N2. Is identified as a streak defect.

[Operation of defect detection device]
Next, the defect detection method by the defect detection apparatus of this Embodiment is demonstrated.
FIG. 9 is a flowchart for explaining the operation of the defect detection apparatus 100 of the present embodiment.

First, when the inspection object 1 is set on the XY stage 2, the image input means 60 of the control device 6 captures an image of the inspection object 1 with the CCD camera 5 and obtains image data of the captured image. An acquisition process (imaging process) is performed (ST1). At this time, the image data of the photographed image is taken into the control device 6 as, for example, 256 gradation (8 bits) digital data by an A / D converter (not shown).
When the inspection object 1 is a display panel such as a liquid crystal panel, a specific image pattern may be displayed on the display panel so that defects can be easily detected. For example, there are an all-white screen pattern that displays the entire screen in white so that dark defects can be easily detected, an all-black screen pattern that displays all screens in black so that bright defects can be easily detected, and a halftone screen pattern. What is necessary is just to set suitably according to the defect kind to want.

  Next, the defect emphasis processing means 61 performs a defect emphasis processing step for emphasizing defects on the acquired image (ST2). In this defect enhancement processing step ST2, since it is difficult to detect a low-contrast defect as it is, only the defect component in the image is enhanced. The defect enhancement processing step ST2 is performed according to the processing flow shown in FIG.

First, the defect enhancement processing means 61 executes an inspection target pixel selection step of selecting an inspection target pixel to be inspected by the inspection target pixel selection means 611 (ST21).
In the present embodiment, the target pixel is selected for each imaging pixel unit of the CCD camera 5.

Next, the defect enhancement processing means 61 executes the comparison target pixel group setting step by the comparison target pixel group setting means 612 (ST22).
That is, the comparison target pixel group setting means 612 uses a defect emphasis filter configuration 70 as shown in FIG. 3 and, as shown in FIG. 4, 56 comparison targets in the circumferential direction centering on the inspection target pixel O _1. Pixels S _{1 to} S ₅₆ are set.

The comparison target pixels S _{1 to} S ₅₆ are set by the defect enhancement filter configuration 70 in which the distance between the inspection target pixel O ₁ and the comparison target pixels S _{1 to} S ₅₆ is 7 pixels. The defect emphasis filter configuration 70 is appropriately selected according to the width dimension of the target stripe defect. That is, in the present embodiment, the defect is emphasized by the luminance difference between the inspection target pixel O ₁ and the comparison target pixels S _{1 to} S ₅₆ , so that the streak defect is within the area surrounded by the comparison target pixels S _{1 to} S ₅₆ . It can be detected only if the width is within the range. Therefore, the distance between the inspection target pixel O ₁ and the comparison target pixels S _{1 to} S ₅₆ may be set according to the size of the stripe defect to be detected.

The comparison target pixel group setting unit 612 then sets these comparison target pixels S _{1 to} S ₅₆ as a set of 28 comparison targets that are a pair of comparison target pixels that are symmetric about the inspection target pixel O _1. Set by dividing into pixel groups.

Next, the defect enhancement processing means 61 performs a minimum luminance difference calculation step by the minimum luminance difference calculation means 613 (ST23). Specifically, the minimum luminance difference calculation unit 613 first selects a pair of comparison target pixels in the first comparison target pixel group (S ₁ , S ₂₉ ) in order, as shown in Expressions (1) and (2). , obtaining the inspection target pixel each comparison target pixel _S 1 from the luminance value of _{O 1,} bright defect luminance difference by subtracting the luminance value of _{S 29 F ws1, F ws29.} Further, the minimum luminance difference calculation means 613 subtracts the luminance value of the inspection target pixel O ₁ from the luminance values of the comparison target pixels S ₁ and S ₂₉ to obtain the dark defect luminance differences F _bs1 and F _bs29 .

Next, the minimum luminance difference calculation unit 613, using Equation (3) and (6), the first comparison the bright defect luminance difference of the target pixel group _{F ws1,} among _{F Ws29,} the value is minimum bright defect minimum luminance difference _{D ws1,} and, among the dark defect luminance difference _{F bs1, F bs29,} obtaining the dark defect minimum luminance difference _{D bs1} value is minimized.

Next, the minimum luminance difference calculating unit 613 sequentially selects each pixel of the second comparison target pixel group one by one, and in the same manner as the above processing, calculates each comparison target pixel S from the luminance value of the inspection target pixel O _{1. 2} and the bright defect luminance difference F _ws2 and F _ws30 obtained by subtracting the luminance values of S ₃₀ and the dark defect luminance difference F obtained by subtracting the luminance value of the inspection target pixel O ₁ from the luminance values of the comparison target pixels S ₂ and S _{30. bs2} and _Fbs30 are obtained. Then, as in the first comparative pixel group, determine these minimum and is bright defect minimum luminance difference D _ws1 value _and dark defect minimum luminance difference D _bs1.
The minimum luminance difference calculation unit 613 performs the same processing as that for the first comparison pixel group and the second comparison pixel group for n = 3 to ₂₈ , and _determines the bright defect minimum luminance difference D _wsn and darkness. The defect minimum luminance difference D _bsn is obtained sequentially.

Next, the defect emphasis processing means 61 performs a defect emphasis value calculation step of setting the defect emphasis value of the inspection target pixel by the defect emphasis value calculation means 614 (ST24). Specifically, the defect emphasis value calculation means 614 uses the bright defect minimum luminance difference D _wsn calculated for each comparison target pixel group as the bright defect at the position of the inspection target pixel O ₁ with the maximum value. The brightness value minimum brightness difference D _wsn calculated for each comparison target pixel group is set as the enhancement value R1, and the one with the maximum value is set as the dark defect enhancement value R2 at the position of the inspection target pixel O _1. .

As shown in FIG. 10, the defect enhancement processing means 61 determines whether or not the defect enhancement processing has been completed for the entire acquired image (ST25). If not, the inspection target pixel O ₁ is moved. Then, another inspection target pixel O ₁ is selected (ST21), and a comparison target pixel group setting step ST22, a minimum luminance difference calculation step ST23, and a defect enhancement value calculation step ST24 are performed. That is, since the inspection target pixel O ₁ is set for each imaging pixel of the CCD camera 5, the inspection target pixel O ₁ may be sequentially moved for each imaging pixel, and the steps ST21 to ST24 may be sequentially performed.

On the other hand, if the processing has been completed in ST25, the defect enhancement processing means 61 generates a defect enhanced image using the defect enhancement value calculated for each pixel (ST26). That is, the defect enhancement processing means 61 generates a bright defect enhanced image based on the bright defect enhanced value R1, and generates a dark defect enhanced image based on the dark defect enhanced value R2. Here, in the present embodiment, the defect enhancement processing means 61 generates a bright defect enhancement image and a dark defect enhancement image with the defect enhancement values R1, R2 as luminance values. For example, the defect enhancement values R1, R2 It is also possible to set a level value in accordance with, and generate an emphasized image in which the luminance value and chromaticity are set in accordance with the level value. Here, as sample data of the defect detection processing in the present embodiment, FIG. 12 shows an example of a captured image, FIG. 13 shows a bright defect enhanced image with respect to the captured image of FIG. 12, and FIG. The dark defect emphasis image with respect to the captured image of is shown.
In addition, the defect emphasis processing unit 61 generates only the defect emphasis image based on the calculated defect emphasis value when only the defect emphasis value of either the bright defect or the dark defect is calculated.
Thus, the defect enhancement processing step ST2 is completed.

  When the defect enhancement processing step ST2 is completed, as shown in FIG. 9, the defect detection means 62 executes a defect detection step of detecting streak defects based on the defect enhancement image obtained in the defect enhancement processing step ST2 ( ST3). FIG. 11 is a flowchart showing the defect detection process by the defect detection means 62.

In the defect detection process, first, the defect detection means 62 performs a defect candidate extraction step of extracting defect candidate pixels by the defect candidate extraction means 621 (ST31).
For this purpose, the defect candidate extraction unit 621 of the defect detection unit 62 sets a bright defect threshold for cutting out a bright defect as shown in Expression (8) for the bright defect emphasized image, and for the dark defect emphasized image. A dark defect threshold value for cutting out a dark defect as shown in Expression (9) is set. Then, the defect candidate extraction unit 621 compares the brightness values of the bright defect enhanced image and the dark defect enhanced image, that is, the bright defect enhanced value R1 and the dark defect enhanced value R2, with the set bright defect threshold wslevel and dark defect threshold bslevel. Then, pixels that are equal to or greater than the threshold are cut out as defect candidate pixels. At this time, the defect candidate extraction unit 621 detects pixels having the bright defect threshold wslevel or higher as the bright defect candidate pixels for the bright defect enhancement result, and detects pixels having the dark defect threshold bslevel or higher for the dark defect enhancement result. It is detected as a dark defect candidate pixel.

  Next, the defect detection means 62 carries out a defect candidate pixel group extraction step for setting the defect candidate pixel group 81 by the circumscribed rectangle setting means 622 (ST32), and further, the inspection circumscribing that circumscribes the defect candidate pixel group 81 A circumscribed rectangle setting process for setting the rectangle 82 is performed (ST33).

  Specifically, in the defect candidate pixel group extraction step, the circumscribed rectangle setting unit 622 determines whether or not there is another defect candidate pixel in the surrounding eight pixel range for each defect candidate pixel extracted in ST31. to decide. When there are other defect candidate pixels in the eight pixels around the defect candidate pixels, these defect candidate pixels are associated with each other and extracted as a defect candidate pixel group 81. Here, when the number of defect candidate pixels included in the defect candidate pixel group 81 is less than the predetermined threshold, it is considered that the defect candidate pixel is noise. You may perform operation | movement which does not implement a process.

  In the circumscribed rectangle setting step of ST33, the circumscribed rectangle setting means 622 detects the peripheral defect candidate pixels included in the defect candidate pixel group 81, and the area of the circumscribed rectangle circumscribing these peripheral defect candidate pixels is the smallest. The circumscribed rectangle is set as the inspection circumscribed rectangle 82.

  Thereafter, the defect detection means 62 measures the length dimension l of the long side and the length dimension h of the short side of the inspection circumscribed rectangle 82 set in the circumscribed rectangle setting process of ST33 by the dimension ratio calculating means 623. Then, the dimension ratio calculating means 623 performs a dimension ratio calculating step of calculating the dimension ratio M1, which is the ratio of the length of the short side to the length of the long side (ST34).

  Further, the defect detection means 62 performs an area ratio calculation step of calculating the area ratio of the defect candidate pixel group 81 with respect to the inspection circumscribed rectangle 82 in ST33 by the area ratio calculation means 624 (ST35). Specifically, the area ratio calculation unit 624 measures, for example, the number m of defect candidate pixels included in the defect candidate pixel group 81 and multiplies the area a for one pixel ( ma) is calculated, and the area ratio (ma / lh) to the area (lh) of the inspection circumscribed rectangle 82 is calculated.

Next, the defect detection means 62 carries out a defect specification step of specifying only a streak-like defect from the defect candidate pixel group 81 by the defect specification means 625 (ST36).
Specifically, the defect specifying means 625 compares the dimension ratio M1 calculated in the dimension ratio calculation process of ST34 with a predetermined dimension ratio threshold N1 (for example, N1 = 0.1), and the dimension ratio M1 is the dimension. When it is less than the ratio threshold N1, the defect candidate pixel group 81 is specified as a streak defect. Further, the defect identification unit 625 compares the area ratio M2 calculated in the area ratio calculation process of ST35 with a predetermined area ratio threshold N2 (for example, N2 = 0.1), and the area ratio M2 is the area ratio threshold N2. If it is less, the defect candidate pixel group 81 is specified as a streak-like defect. That is, the defect specifying means 625 detects a defect candidate pixel group 81 having a dimension ratio M1 less than the dimension ratio threshold N1, or a defect candidate pixel group 81 having an area ratio M2 less than the area ratio threshold N2, and creates a streak-like defect. Identify.

Moreover, the defect detection means 62 produces | generates the defect specific image based on the specified stripe-shaped defect. Here, the defect detection means 62 generates a defect specific image for each of the bright defect detection enhanced image and the dark defect enhanced image. That is, the defect detection means 62 generates a bright defect specifying image as shown in FIG. 15 based on the streak defect detected for the bright defect emphasized image as shown in FIG. 13, and also shown in FIG. Based on the streak defect detected with respect to such a dark defect enhanced image, a dark defect specific image as shown in FIG. 16 is generated.
Note that a process of combining these bright defect specific image and dark defect specific image into one defect specific image may be performed.

  In the above-described embodiment, the example in which the inter-pixel distance from the center pixel 71 to the comparison setting pixel 72 is 7 pixels is shown as the defect emphasis filter configuration 70, but the present invention is not limited to this. For example, when the width dimension of the stripe defect to be inspected is not particularly set, a plurality of defect enhancement filter configurations 70 having different inter-pixel distances from the center pixel 71 to the comparison setting pixel 72 are used, and each defect enhancement filter configuration Based on the defect emphasis values R1 and R2 set by 70, a defect detection process for specifying a streak defect may be performed. In this case, it becomes possible to detect streak-like defects of various sizes corresponding to each defect enhancement filter configuration 70. Moreover, the defect specific image which displayed the stripe-shaped defect of various width dimensions can be obtained by synthesize | combining the stripe-shaped defect detected in this way, and making it one defect specific image, for example.

[Effects of Embodiment]
As described above, in the defect detection apparatus 100 of the present embodiment, the defect enhancement processing unit 61 arranges the plurality of comparison target pixels S _{1 to} S ₅₆ around the inspection target pixel O ₁ , and these The comparison target pixels S _{1 to} S ₅₆ are divided into a plurality of comparison target pixel groups and set, and the difference between the luminance value of the inspection target pixel O _{1 and} the luminance value of each comparison target pixel S _{1 to} S ₅₆ is obtained. Then, the minimum luminance differences D _wsn and D _bsn having the smallest luminance difference data are calculated for each comparison target pixel group, and among these minimum luminance differences D _wsn and D _bsn , the largest value is determined as the defect of the inspection target pixel O ₁ . Emphasized values R1 and R2.
Then, the defect detection means 62 extracts defect candidate pixels based on the defect enhancement values R1 and R2 calculated as described above by the defect candidate extraction means 621, and the defect candidate adjacent to each other by the circumscribed rectangle setting means 622. An inspection circumscribed rectangle 82 of the defect candidate pixel group 81 in which the pixels are connected is set. Thereafter, the defect detection means 62 calculates the size ratio M1 of the short side with respect to the long side of the inspection circumscribed rectangle by the dimension ratio calculation means 623, and the area of the defect candidate pixel group 81 for the inspection circumscribed rectangle by the area ratio calculation means 624. The ratio M2 is calculated. Then, the defect specifying unit 625 extracts a defect candidate pixel group 81 in which the size ratio M1 is less than the size ratio threshold N1, or a defect candidate pixel group 81 in which the area ratio M2 is less than the area ratio threshold N2, and these defect candidates. The pixel group 81 is detected as a streak defect.

Therefore, the defect detection apparatus 100, by obtaining the minimum luminance difference at each comparison pixel group, in each comparison pixel group includes the inspection target pixel O _1, and, within the area enclosed by the comparison pixel A certain spot-like defect and a streak-like defect passing through a region surrounded by the comparison target pixel can be emphasized. At this time, since each comparison target pixel group is different in position of the comparison target pixel, defects that cannot be emphasized in one comparison target pixel group can be emphasized in the other comparison target pixel group. By setting a larger value among the luminance differences as the defect emphasis value of the inspection target pixel O ₁ , it is possible to surely emphasize and detect the defect candidate pixel of the spot-like defect and the streak-like defect. The detection omission due to the angle of can be prevented, and the defect detection sensitivity can be improved.
Then, the defect candidate pixel group 81 and the inspection circumscribed rectangle 82 are set for the defect candidate pixels detected as described above, and the dimension ratio M1 for the inspection circumscribed rectangle 82 and the defect candidate pixel group for the inspection circumscribed rectangle 82 are set. An area ratio M2 of 81 is calculated. In the detection of the stripe defect using the inspection circumscribed rectangle 82, when the defect candidate pixel group 81 is a linear stripe defect, the area ratio M2 is M2≈1, but the length of the inspection circumscribed rectangle 82 is long. The length dimension of the short side with respect to the side becomes small, and the dimension ratio M1 becomes a value close to zero. Therefore, the defect specifying means 625 can easily detect a linear streak defect from the defect candidate pixel group 81 by comparing the dimension ratio M1 with the dimension ratio threshold N1. In addition, when the defect candidate pixel group 81 is a curved, broken line, or annular streak defect, the dimension ratio M1 may be a value close to 1, but the area ratio M2 is a value close to 0. Therefore, the defect specifying means 625 can easily detect a curved, bent, or annular streak defect by comparing the area ratio M2 with the area ratio threshold N2.
As described above, the defect enhancement processing unit 61 can detect the defect candidate pixels with high accuracy, and the defect detection unit 62 can accurately identify the streak-like defects from these defect candidate pixels. Therefore, in the inspection for detecting only the streak-like defect from the inspection object 1, the streak-like defect can be easily and accurately detected.

Further, the minimum luminance difference calculation means 613 uses the equations (1), (2), and (4), (5) in calculating the minimum luminance difference, so that the bright defect minimum luminance difference D _wsn and the dark defect minimum luminance difference are _calculated. D _bsn is obtained, and the defect emphasis value calculating means 614 _obtains the light defect emphasis value R1 and the dark defect emphasis value R2 from the light defect minimum luminance difference D _wsn and the dark defect minimum luminance difference D _bsn , respectively. At 62, a streak-like bright defect and a streak-like dark defect are detected based on the bright defect enhancement value R1 and the dark defect enhancement value R2, respectively.
For this reason, a streak-like defect can be detected separately for a bright defect and a dark defect, and the defect detection accuracy can be further improved. Further, it is possible to easily cope with a case where it is desired to separately detect and detect a light defect and a dark defect in the inspection object 1 or a case where any one of a light defect and a dark defect is detected. It is possible to detect a streak defect and a dark defect.

The comparison target pixel group setting unit 612 of the defect enhancement processing unit 61 selects the defect enhancement filter configuration 70 according to the width dimension of the stripe defect to be detected, and sets the comparison target pixel and the comparison target pixel group. .
For this reason, for example, by selecting a defect enhancement filter configuration 70 having a shorter inter-pixel distance from the center pixel 71 to the comparison setting pixel 72 by a user's setting input, only finer streak-like defects can be detected. . Therefore, the width dimension of the streak-like defect to be detected can be easily changed, and appropriate defect detection according to the inspection purpose can be performed.

Further, at this time, the comparison target pixel group setting unit 612 has a distance between pixels of a pair of comparison setting pixels 72 arranged with the center pixel 71 interposed therebetween with respect to the width dimension of the stripe defect to be detected. The comparison target pixel and the comparison target pixel group are set using the defect emphasis filter configuration 70 that is larger by the pixel.
When the distance between the pair of comparison setting pixels 72 arranged with the center pixel 71 of the defect enhancement filter configuration 70 sandwiched therebetween is equal to or larger than the width dimension of the stripe-like defect to be detected, Although a streak-like defect can be detected, for example, when a part of the streak-like defect has a part with an increased width dimension, the defect emphasis value is not calculated correctly, and detection omission may occur. On the other hand, as described above, the defect enhancement filter in which the inter-pixel distance between the pair of comparison setting pixels 72 arranged with the center pixel 71 interposed therebetween is two to three pixels larger than the width dimension of the stripe defect to be detected. By using the configuration 70, the stripe defect can be detected with higher accuracy. Note that the defect emphasis filter configuration 70 in which the inter-pixel distance between the pair of comparison setting pixels 72 arranged symmetrically with respect to the center pixel 71 may be larger than the width dimension of the stripe defect to be detected. In this case, since a non-defective non-defective part may be detected as a defect, for example, an over-detection may be considered. Therefore, as described above, the center pixel 71 is disposed with respect to the width dimension of the stripe-shaped defect to be detected. It is preferable to use a defect enhancement filter configuration 70 in which the distance between the pair of comparison setting pixels 72 is larger by 2 to 3 pixels.

The circumscribed rectangle setting unit 622 of the defect detecting unit 62 sets the circumscribed rectangle that circumscribes the defect candidate pixel group 81 as the inspection circumscribed rectangle 82 that has the smallest area. Then, the dimension ratio calculating unit 623 and the area ratio calculating unit 624 calculate the dimension ratio M1 and the area ratio M2 based on the inspection circumscribed rectangle 82, respectively.
By setting such an inspection circumscribed rectangle 82, when there is a linear streak defect, it is possible to make the inertial axis direction of the streak defect coincide with the inertial axis direction (direction along the long side) of the inspection circumscribed rectangle. it can. That is, the length dimension along the inertial axis direction of the linear streak defect and the long side dimension of the inspection circumscribed rectangle 82 are substantially the same dimension, and the width dimension of the linear streak defect and the short side dimension of the inspection circumscribed rectangle 82 are the same. And have substantially the same dimensions. Therefore, by comparing the dimension ratio M1 of the inspection circumscribed rectangle 82 with the dimension ratio threshold N1, it is possible to easily identify the linear streak defect.
Further, when detecting a line-like defect such as a curved line, a broken line, or a ring using the area ratio M2, the detection accuracy can be improved by using the inspection circumscribed rectangle 82 as described above. That is, among the circumscribed rectangles circumscribing the defect candidate pixel group 81, for example, in the circumscribed rectangle having a large area, the ratio of the area of the defect candidate pixel to the area of the circumscribed rectangle may be small even for a spot-like defect. It is conceivable that only the streak defect cannot be detected with high accuracy. On the other hand, when the circumscribed rectangle having the smallest area among the plurality of circumscribed rectangles is the inspection circumscribed rectangle 82, the area ratio M2 is reduced only in the case of a line-shaped defect such as a curved line, a polygonal line, or a ring, In the spot-like defect, the area ratio M2 is a large value close to 1. Therefore, by comparing the area ratio M2 using the inspection circumscribed rectangle 82 and the area ratio threshold N2, it is possible to easily and accurately detect a line-shaped defect such as a curved line, a polygonal line, or a ring. it can.

[Other Embodiments]
It should be noted that the present invention is not limited to the above-described embodiments, and modifications, improvements, and the like within the scope that can achieve the object of the present invention are included in the present invention.
For example, in the above embodiment, the defect emphasis processing unit 61 and the defect detection unit 62 are stored in the storage unit, and have been shown as examples configured as programs that are appropriately read out by the CPU to execute arithmetic processing. However, the present invention is not limited to this, and for example, an integrated circuit such as an IC chip and various hardware may be used.

  The present invention can detect if the captured image of the object to be inspected 1 has a portion having a luminance difference from the surrounding. For this reason, the present invention can be widely used for detection of foreign matter defects on a flexible substrate, detection of scratches and dirt on the surface of an object to be inspected, detection of streak defects of various display devices, and the like.

  In addition, the specific structure and procedure for carrying out the present invention can be changed as appropriate to other structures and the like within the scope of achieving the object of the present invention.

  DESCRIPTION OF SYMBOLS 1 ... Inspection object, 61 ... Defect emphasis processing means, 62 ... Defect detection means, 70 ... Defect emphasis filter configuration, 81 ... Defect candidate pixel group, 82 ... Inspection circumscribed rectangle, 100 ... Defect detection apparatus, 611 ... Inspection object pixel Selection means, 612 ... Comparison target pixel group setting means, 613 ... Minimum luminance difference calculation means, 614 ... Defect enhancement value calculation means, 621 ... Defect candidate extraction means, 622 ... circumscribed rectangle setting means, 623 ... dimensional ratio calculation means, 624 ... Area ratio calculation means, 625 ... Defect identification means.

Claims (7)

A defect enhancement processing step of performing defect enhancement processing using a defect enhancement filter on a captured image obtained by imaging the inspection object; and
A defect detection step of detecting a defect based on the defect enhancement value of each pixel obtained in the defect enhancement processing step, and
The defect emphasis processing step includes
An inspection target pixel selection step for sequentially selecting inspection target pixels for the captured image;
A plurality of comparison target pixels that are separated from the center of the selected inspection target pixel by a predetermined distance are arranged in a substantially circular shape around the inspection target pixel, and are located symmetrically with respect to each other across the inspection target pixel among these comparison target pixels A comparison target pixel group setting step of setting a plurality of comparison target pixel groups in which a pair of comparison target pixels arranged in a set is set;
The luminance difference data that is the difference between the luminance value of each comparison target pixel included in the comparison target pixel group and the luminance value of the inspection target pixel is obtained, and the minimum luminance difference that minimizes the value among the luminance difference data Calculating a minimum luminance difference for each comparison target pixel group,
A defect enhancement value calculating step in which a minimum luminance difference having a maximum value among the minimum luminance differences calculated for each comparison target pixel group is a defect enhancement value of the inspection target pixel; and
Have
The defect detection step includes
A defect candidate extraction step of extracting pixels with the defect enhancement value equal to or higher than a predetermined defect threshold as defect candidate pixels;
A circumscribed rectangle setting step for setting the defect candidate pixels adjacent to each other as a defect candidate pixel group, and setting a circumscribed rectangle having a minimum area as an inspection circumscribed rectangle among circumscribed rectangles circumscribing the defect candidate pixel group;
The long side dimension of the inspection circumscribed rectangle along the inertial axis direction of the defect candidate pixel and the short side dimension of the inspection circumscribed rectangle along the width direction orthogonal to the inertial axis direction are measured, and the short side with respect to the long side dimension is measured. A dimensional ratio calculating step of calculating a dimensional ratio which is a ratio of dimensions;
An area ratio calculating step of calculating an area ratio of the defect candidate pixel group with respect to the inspection circumscribed rectangle;
The defect when the dimensional ratio calculated in the dimensional ratio calculating step is less than a predetermined dimensional ratio threshold or when the area ratio calculated in the area ratio calculating step is less than a predetermined area ratio threshold. A defect identification step for identifying the candidate pixel group as a streak defect;
A defect detection method characterized by comprising: The defect detection method according to claim 1,
The minimum luminance difference calculating step calculates a light defect minimum luminance difference from light defect luminance difference data obtained by subtracting the luminance value of each comparison target pixel included in the comparison target pixel group from the luminance value of the inspection target pixel. Minimum dark difference calculation step, and dark defect minimum for calculating a dark defect minimum brightness difference from dark defect brightness difference data obtained by subtracting the brightness value of the inspection target pixel from the brightness value of each comparison target pixel included in the comparison target pixel group A luminance difference calculating step,
The defect enhancement value calculation step obtains a bright defect enhancement value and a dark defect enhancement value based on the bright defect minimum luminance difference and the dark defect minimum luminance difference,
The defect detection step detects a streak-like bright defect and a streak-like dark defect based on the bright defect enhancement value and the dark defect enhancement value, respectively. In the defect detection method according to claim 1 or 2,
In the defect enhancement processing step, the defect enhancement processing is performed using a plurality of defect enhancement filters having different distances from the inspection target pixel to the comparison target pixel. The defect detection method according to any one of claims 1 to 3,
The defect enhancement processing step uses the defect enhancement filter in which a distance from the inspection target pixel to the comparison target pixel is set in accordance with the width dimension of the streak defect detected by the defect detection step. A defect detection method characterized by performing processing. The defect detection method according to claim 4,
The defect enhancement processing step uses a defect enhancement filter in which a distance from the inspection target pixel to the comparison target pixel is a distance larger by a predetermined pixel than a width dimension of the streak defect detected by the defect detection step. And performing the defect enhancement process. Defect enhancement processing means for performing defect enhancement processing using a defect enhancement filter on a captured image obtained by imaging the inspection object;
A defect detection means for detecting a defect based on the defect enhancement value of each pixel obtained in the defect enhancement processing step,
The defect enhancement processing means includes
Inspection target pixel selection means for sequentially selecting inspection target pixels for the captured image;
A plurality of comparison target pixels that are separated from the center of the selected inspection target pixel by a predetermined distance are arranged in a substantially circular shape around the inspection target pixel, and are located symmetrically with respect to each other across the inspection target pixel among these comparison target pixels Comparison target pixel group setting means for setting a plurality of comparison target pixel groups in which a pair of comparison target pixels arranged in a set is set;
The luminance difference data that is the difference between the luminance value of each comparison target pixel included in the comparison target pixel group and the luminance value of the inspection target pixel is obtained, and the minimum luminance difference that minimizes the value among the luminance difference data A minimum luminance difference calculating means for obtaining for each comparison target pixel group,
A defect enhancement value calculation means for setting a minimum luminance difference having a maximum value among the minimum luminance differences calculated for each comparison target pixel group as a defect enhancement value of the inspection target pixel;
Have
The defect detection means includes
A defect candidate extracting means for extracting, as a defect candidate pixel, a pixel having the defect enhancement value equal to or greater than a predetermined defect threshold;
A circumscribed rectangle setting means for setting the defect candidate pixels adjacent to each other as a defect candidate pixel group, and setting a circumscribed rectangle having a minimum area as an inspection circumscribed rectangle among circumscribed rectangles circumscribing the defect candidate pixel group;
The long side dimension of the inspection circumscribed rectangle along the inertial axis direction of the defect candidate pixel and the short side dimension of the inspection circumscribed rectangle along the width direction orthogonal to the inertial axis direction are measured, and the short side with respect to the long side dimension is measured. A dimensional ratio calculating means for calculating a dimensional ratio which is a dimensional ratio;
An area ratio calculating means for calculating an area ratio of the defect candidate pixel group with respect to the inspection circumscribed rectangle;
The defect when the dimensional ratio calculated in the dimensional ratio calculating step is less than a predetermined dimensional ratio threshold or when the area ratio calculated in the area ratio calculating step is less than a predetermined area ratio threshold. A defect specifying means for specifying a candidate pixel group as a streak-like defect;
A defect detection apparatus comprising: A defect detection program that is read and calculated by a calculation means,
The defect detection program includes defect enhancement processing means for performing defect enhancement processing using a defect enhancement filter for a captured image obtained by imaging an inspection object;
A defect detection means for detecting a defect based on the defect enhancement value of each pixel obtained in the defect enhancement processing step,
The defect enhancement processing means includes
Inspection target pixel selection means for sequentially selecting inspection target pixels for the captured image;
A plurality of comparison target pixels that are separated from the center of the selected inspection target pixel by a predetermined distance are arranged in a substantially circular shape around the inspection target pixel, and are located symmetrically with respect to each other across the inspection target pixel among these comparison target pixels Comparison target pixel group setting means for setting a plurality of comparison target pixel groups in which a pair of comparison target pixels arranged in a set is set;
The luminance difference data that is the difference between the luminance value of each comparison target pixel included in the comparison target pixel group and the luminance value of the inspection target pixel is obtained, and the minimum luminance difference that minimizes the value among the luminance difference data A minimum luminance difference calculating means for obtaining for each comparison target pixel group,
A defect enhancement value calculation means for setting a minimum luminance difference having a maximum value among the minimum luminance differences calculated for each comparison target pixel group as a defect enhancement value of the inspection target pixel;
Have
The defect detection means includes
A defect candidate extracting means for extracting, as a defect candidate pixel, a pixel having the defect enhancement value equal to or greater than a predetermined defect threshold;
A circumscribed rectangle setting means for setting the defect candidate pixels adjacent to each other as a defect candidate pixel group, and setting a circumscribed rectangle having a minimum area as an inspection circumscribed rectangle among circumscribed rectangles circumscribing the defect candidate pixel group;
The long side dimension of the inspection circumscribed rectangle along the inertial axis direction of the defect candidate pixel and the short side dimension of the inspection circumscribed rectangle along the width direction orthogonal to the inertial axis direction are measured, and the short side with respect to the long side dimension is measured. A dimensional ratio calculating means for calculating a dimensional ratio which is a dimensional ratio;
An area ratio calculating means for calculating an area ratio of the defect candidate pixel group with respect to the inspection circumscribed rectangle;
The defect when the dimensional ratio calculated in the dimensional ratio calculating step is less than a predetermined dimensional ratio threshold or when the area ratio calculated in the area ratio calculating step is less than a predetermined area ratio threshold. A defect specifying means for specifying a candidate pixel group as a streak-like defect;
A defect detection program characterized by comprising: