JP2002310937A - Method and apparatus for inspection of defect - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and an apparatus wherein a specimen on which a sample pattern is repeated can be inspected with high accuracy even when a pattern arrangement is asymmetric and even when a large memory capacity is not available. SOLUTION: When an inspection image obtained by imaging the specimen is processed so as to inspect a pattern defect, an isolated region whose luminance is different from that of a background is extracted. Regarding each isolated region, an area which uses a pixel as a unit and its center of gravity are calculated. The area of an isolated region to be noticed is compared with the area of an isolated region in the neighborhood. When the difference in an area between two or more isolated regions exceeds a reference area, it is judged that a defect exists. Inversely, when the area difference is less than the reference area, the minimum number of times of processing operations in which the reference area disappears when a contraction processing operation is performed by using the pixel as the unit is used as the reference number of times. An original image in the isolated region to be noticed, an opening image which is contracted in the same number of times after the original image has been expanded in the reference number of times and a closing image which is expanded in the same number of times after the original image has been expanded in the same number of times are compared respectively. When there exists a difference, it is determined that a very small defect exists.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a defect inspection method and apparatus, and more particularly to a defect inspection method and apparatus suitable for inspecting a pattern formed on a shadow mask, a color filter, a PDP, or the like.

[0002]

2. Description of the Related Art A shadow mask for a CRT of a color television, a color filter for a liquid crystal display, and
For a product (work) in which a fine pattern such as DP (Plasma Display Panel) is repeatedly formed, it is important that the pattern is correctly formed as designed. I have. Generally, for such an inspection, a method is adopted in which a work to be inspected is imaged by a camera or the like, and good or bad is determined based on the obtained image.

[0003] As inspection methods based on images obtained by imaging a workpiece in this way, there are known (1) a non-defective product comparison method, (2) a pitch shift method, and (3) a symmetric site comparison method. Have been.

(1) In the non-defective product comparison method, an inspection image obtained by imaging a target work, a non-defective image prepared by imaging a non-defective work in advance, and CAD data used in product design are used. This is a method for comparison and inspection.

(2) In the pitch shift method, an image of an inspection object (work) is captured by, for example, a CCD camera, and is binarized with a predetermined threshold value to obtain a binary image. An image is extracted for each fixed pitch, which is a unit of periodicity, and defects are obtained by taking the exclusive OR of an image between a certain pitch and an image that is supposed to be the same and shifted by a certain pitch from it. It is a method of detecting.

(3) The symmetric site comparison method is a method of comparing and inspecting symmetric sites when the inspection object has a geometrically symmetric structure. For example, in the case of a shadow mask, assuming that the entire image is shown in FIG. 12A, a part of the image is shown in an enlarged manner in FIG. Is formed under. As shown in FIG. 13, such a shadow mask has a line (axis) with respect to each of an X axis and a Y axis passing through the center.
Many products are symmetric or point symmetric about the center.

[0007] Inspection methods utilizing the symmetry of this structure include (i) a method of taking images of symmetrical parts of a product with two CCD cameras and comparing them, and (ii) one or more of them. There is a method in which images of a certain range are captured by two cameras, and symmetrical parts in the range are compared.

In the method (i), as shown in a perspective view of an image taken in FIG. 14A, a camera having a structure symmetric with respect to an alternate long and short dash line is moved by two cameras in the direction of the arrow. Scan sequentially and input images. FIG. 7B shows the relationship between the workpiece and the locus of the camera scan viewed from above in this case.
Shown in In the method (ii), for the case of one camera, an image of the entire inspection object is once input as shown in the perspective view of FIG. Thereafter, as shown in FIG. 3B, symmetrical points are compared on the memory of the apparatus.

[0009]

However, the conventional inspection methods (1) to (3) have the following problems.

(1) Problems of the non-defective product comparison method In order to perform this method, it is necessary to always store non-defective image data or design CAD data in the memory of the apparatus. If the size of the defect to be detected is very small (for example, several μm), it is necessary to take an image with high resolution. Therefore, when the size of the work is very large (for example, about 1000 mm × 1000 mm), a non-defective image of the entire work has a huge data amount. The same applies to CAD data. Therefore, when an image of a good product or CAD data is stored in a memory inside the apparatus and an inspection for comparing the image with an image of an inspection object is performed for a large size, a large memory capacity is required. When the size of the memory device is larger than the above, there is a problem that the scale of the memory device increases and the price of the device itself increases.

(2) Problems with the Pitch Shifting Method Some products having a fine pattern such as a shadow mask have different pitches at which patterns are arranged at a central portion and a peripheral portion. If a method of shifting the image by a certain pitch is applied to an image of such a product, erroneous detection occurs frequently.

(3) Problems with the Symmetric Site Comparison Method Asymmetrical inspection objects having no line-symmetric or point-symmetric structure in the pattern arrangement cannot be inspected. Also, even if it has a symmetric structure, if a specific pattern exists on the axis of symmetry and the point of symmetry and the defect is also axially symmetric and point symmetric, FIG.
No detection is possible as shown in FIG.

In the method (i) shown in FIG. 14, it is necessary that the fields of view of the two cameras are always accurately and symmetrically located on the work. Therefore, precise conveyance control is required so that the fields of view of the two cameras do not shift. In addition, inspection on the symmetry axis and the vicinity thereof is difficult because physical interference occurs in an input system such as a camera as shown in an image in FIG. Also, (ii) shown in FIG.
The method of (1) has a similar problem since it needs to handle a large amount of data as in the case of the non-defective product comparison method of (1).

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems, and a work in which the same pattern is repeatedly formed can have a large memory capacity even when the pattern arrangement has no symmetry. It is an object of the present invention to provide a defect inspection method and apparatus capable of inspecting a pattern with high accuracy without requiring a pattern inspection.

[0015]

SUMMARY OF THE INVENTION According to the present invention, an image of an inspection object on which substantially the same pattern is repeatedly formed is picked up, an inspection image is input, and the input inspection image is image-processed to form a pattern. In a defect inspection method for inspecting defects,
An isolated area different from the luminance of the background is extracted from the inspection image, and for each of the extracted isolated areas, the area and the center of gravity are calculated in units of pixels, and the area of any focused isolated area,
An area comparison is performed to compare the area of the isolated area near the center of gravity with the area of the nearby isolated area. If the area difference between the two or more isolated areas exceeds the reference area, the pattern corresponding to the target isolated area is defective. It is determined that there is, when the area difference is equal to or less than the reference area, when the contraction processing is performed in units of pixels, the minimum number of times the reference area disappears is set as the reference number, and the original image of the noted isolated area is set as Creates an opening image by performing the same number of times of expansion processing on a pixel basis and then performing the same number of times of expansion and contraction processing on the original image by the same number of times after creating the opening image by the same number of times on a pixel basis. Then, the original image is compared with the created opening image and the closing image, respectively. If there is an image difference in at least one of the images, the base image is added to the pattern corresponding to the isolated region of interest. By determining that there is a less defective area is obtained by solving the above problems.

According to the present invention, an inspection object in which substantially the same pattern is repeatedly formed is imaged, an inspection image is input, and the input inspection image is subjected to image processing to inspect the pattern for defects. In the defect inspection apparatus, extracting means for extracting an isolated area different from the luminance of the background from the inspection image; means for calculating an area and a center of gravity of each extracted isolated area in units of pixels; Means for comparing the area of the isolated area with the area of the isolated area near the center of gravity, and when the area difference between two or more isolated areas exceeds the reference area, Means for determining that there is a defect in the pattern corresponding to the reference area, and, when the area difference is equal to or smaller than the reference area, a reference number for setting the minimum number of processing times at which the reference area disappears when contraction processing is performed in units of pixels as the reference number. Means for performing an expansion process on the original image of the isolated region of interest in units of pixels, and then performing an erosion process the same number of times to create an opening image. And a means for expanding the same number of times to create a closing image, and comparing the original image with the created opening image and the closing image, respectively. The above-mentioned problem is also solved by providing image determining means for determining that the pattern corresponding to the region has a defect smaller than the reference area.

That is, in the present invention, the area is compared between an arbitrary isolated region of interest and two or more isolated regions located in the vicinity thereof, and if the area difference is larger than a predetermined reference area, When it is determined that the pattern corresponding to the region has a defect, and when the defect is a minute defect having a size equal to or smaller than the reference area, the opening image and the closing image are created by performing the image processing on the original image of the isolated region of interest. Since two processed images from which minute chips and protrusions have been removed are compared and these processed images are compared with the original image, a defect of an arbitrary size can be detected. Become. Therefore, an inspection object on which substantially the same pattern is formed can be inspected with high accuracy without requiring a large memory capacity, even when the pattern arrangement has no symmetry. .

[0018]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a schematic perspective view showing an entire defect inspection apparatus according to an embodiment of the present invention, and FIG. 2 is a block diagram showing an outline of a control system thereof.

The inspection apparatus according to the present embodiment includes an input buffer 10 in which a work W such as a shadow mask to be inspected is stored, and a robot (loader) that places the work W in the buffer 10 on an inspection stage 12. 14, an X-axis stage 16 for moving the inspection stage 12 on which the work W is placed in the direction of the arrow, and a line sensor camera (imaging means) for imaging the work W on the inspection stage 12 and inputting an inspection image. 18 and the camera 18 are connected to the X-axis stage 16.
Y-axis stage 20 that moves forward and backward in the direction of the arrow that is perpendicular to
And an illumination unit 22 that illuminates the field of view range from the camera side via an optical fiber during imaging by the line sensor camera 18,
An image processing device 24 that processes the inspection image input by the line sensor camera 18 to perform a defect inspection, a robot (unloader) 26 that discharges the inspection-completed work from the inspection stage 12, and the robot 26 discharges the work. An output buffer 28 for storing non-defective products W, a dust box 30 for storing defective products, and a control unit 32 for controlling the operation of the entire apparatus are provided. 2 includes the inspection stage 12, the robot 14, the X-axis stage 16, the Y-axis stage 20, and the robot 26. The image input unit includes the line sensor camera 18, and the image processing unit includes the image processing unit. Each of the processing devices 24 is included.

In the inspection apparatus of the present embodiment, the line sensor camera 18 captures an image of an inspection object (work) in which substantially the same pattern is repeatedly formed, inputs an inspection image, and inputs the inspection image. Each of the processes described below is performed on the image by the image processing device 24 to inspect the pattern for defects.

That is, the image processing device 24 generates a binary image by binarizing the inspection image with a predetermined threshold value (extraction means for extracting an isolated area different from the background luminance).
For each isolated region of the created binary image, means for calculating the area and the center of gravity in units of pixels, and compare the area of an arbitrary isolated region of interest with the area of an isolated region near the center of gravity near Means for determining that a pattern corresponding to the isolated area of interest has a defect when the area difference between the two or more isolated areas exceeds a reference area; and In the following cases, reference setting means for setting the minimum number of times that the reference area disappears when the contraction processing is performed in units of pixels as a reference number, and the original image of the isolated area of interest is subjected to the reference number of expansions in units of pixels. After that, a means for creating an opening image by performing the same number of contraction processings, a means for performing the reference number of times of contraction processing on the original image in units of pixels, and then performing the same number of times of expansion processing to create a closing image, Created Puningu image and closing the image and the image comparison respectively, the pattern in the case where there is an image difference in at least one corresponding to the interest isolated region,
Image determining means for determining that there is a defect having a size equal to or smaller than the reference area are realized by software.

In the present embodiment, the image processing device 24 further includes means for calculating the areas of all the extracted isolated areas, respectively. If the calculated area is significantly different from the area of the normal pattern, It is assumed that the isolated area corresponding to the defect corresponds to the defect, and means for excluding the isolated area from the processing target of the area comparison is realized by software.

Next, the operation of the present embodiment will be described with reference to the flowchart shown in FIG.

First, an inspection image is input from the work W (step 2). In the present embodiment, the inspection stage 12 is moved in the X direction by the X axis stage 16 and the line sensor camera 18 is moved in the Y direction by the Y axis stage 20, as shown in FIG. Meanwhile, the camera 18 scans the work W along a locus indicated by an arrow, and a reflected light image of the entire work W can be input as an inspection image under coaxial incident illumination by the illumination unit 22. It has become. Accordingly, the image processing described in detail below may be sequentially executed every time the line sensor camera 18 captures an image of a predetermined range.
It can also be executed for the entire range of inspection images input by imaging the entirety.

Here, for the sake of convenience, it is assumed that the inspection image input in step 2 is like the image shown in FIG. In FIG. 5, a large black circle represents a normal pattern, and if only the black circle is a good product, it is assumed that a small black or gray circle or a small white or gray circle in the black circle is a defect. Originally, if the work W is a shadow mask, the pattern is a through hole, so there should be no white or gray defects inside. However, consider the case of another work in which such defects may occur. However, here, description will be made as a defect.

Next, the input inspection image of FIG. 5 is binarized using a predetermined threshold value to create a binary image shown in FIG. 6 (step 4).

As the threshold value used in the binarization processing executed here, a value is used that does not lose the black defect and the white defect as well as the original pattern after the binarization. Specifically, a density histogram of an image created separately (not shown)
Is determined to be an intermediate value between the two local maximum points, or to an experimentally appropriate value.

Next, in the created binary image, pixels constituting a portion (isolated region) corresponding to a pattern or a defect are labeled with the same number (step 6). At this time, for each labeling portion (hereinafter, also referred to as an object) extracted as the same isolated region, the area in pixels and the coordinates of the center of gravity and the fillet diameter are calculated. FIG.
Is an image after labeling. In this figure, the center of gravity of each labeling portion (black area) extracted from the binary image of FIG. The rectangle is indicated by a dotted line. Although not used directly here, the fillet diameter is a dimension between opposed sides constituting a circumscribed rectangle.

Next, based on the determined area, an area determination I for sorting the labeling portion between normal and defective is performed (step 8). In this area determination I, the area of each labeling portion is clearly larger than the pattern without defects,
Two kinds of criteria, which are obviously small, are set in advance, so that they are not directly related to the original pattern.
Extremely large and small areas are excluded as defects at this stage. The criteria used here are:
For example, the area So of the object to be compared is 0.01 Sn <So <0.5 with respect to the area Sn of the normal pattern.
A defect is considered when Sn or 1.5Sn <So.

Even if the processing of the area determination I is omitted, the same result can be obtained by performing the next area determination II. However, it is more advantageous to remove the apparently abnormal portion first, because the processing time can be shortened and the possibility of erroneous detection can be reduced.

Next, area determination II is performed (step 1).
0). In this area determination II, the isolated area that was OK in the area determination I is targeted, and the area of an arbitrary isolated area of interest is compared with the area of a nearby isolated area near the center of gravity. Is larger than the reference area, it is determined that the pattern corresponding to the isolated region of interest has a defect. This will be described with reference to FIG. 8 which shows an image of the relationship between the area Sn of the object of interest (isolated region) and the areas Si (i = 1 to 6) of objects near six of the object of interest.

In this area determination II, an object adjacent to the object of interest is obtained from the barycentric coordinates of each object as described above, and the areas of adjacent objects are compared. If the area difference is larger than a predetermined reference area, the defect is determined. And At this time, the object to be compared with the object of interest is compared with two or more neighboring objects, such as the nearest object adjacent before and after, for example. The reason is that it is not clear which one has a defect in comparison with one object. Also, some shadow masks have a slightly different pattern size between the central portion and the peripheral portion, but there is no inconvenience because they are substantially the same as far as they are compared in the vicinity.

The reference area used for the determination is set in advance by directly using the number of pixels or by using a ratio to the area of the object of interest.

As a result of the area judgment II, if there is an object of interest whose area difference is equal to or smaller than the reference area, the minimum number of processings at which the reference area disappears when the contraction processing is performed in units of pixels is set as the reference number (step 12). Next, the original image of the object of interest is expanded by a reference number of times on a pixel-by-pixel basis and then contracted at the same time to create an opening image (step 14). After the processing, a closing image is created by performing the same number of expansions (step 16). Of course, the order of these two processes may be reversed.

FIG. 9 shows an opening image (B) and a closing image (C) by performing an opening process for first performing an expansion process and a closing process for performing an initial contraction process on one object of interest (A). (Hereinafter, both images are also referred to as processed images.) FIG. As shown in this figure,
Depending on the opening process, small chips are removed,
The minute projections are removed by the closing process.

Thereafter, the original image (A) is compared with the created opening image (B) and the created closing image (C) (Step 18). It is determined that the pattern corresponding to the attention isolated area has a defect smaller than the reference area (step 20).

By the way, in the example of FIG. 9 described above, the original image has two small protrusions and chippings smaller than the reference area which could not be detected by the area judgment II. Chips and protrusions can be detected as shown respectively.

As the image comparison, an exclusive OR (X
OR). By performing such an image comparison, it is possible to completely overlap both images other than the defective portion. Therefore, it is possible to reliably detect a minute defect as shown in FIGS. 9D and 9E. As a result, it is possible to improve the detection sensitivity for minute defects.

Since an extremely small isolated area is often caused by electrical noise, it is necessary to prevent the isolated area from being defective.
After applying the exclusive OR, a contraction process may be performed to remove such minute isolated points.

In this embodiment, the reference area used for the defect determination in the area determination II is S, and the reference number used in the opening processing and the closing processing (hereinafter also referred to as expansion / shrinkage processing) is N. As will be described in detail below, these two parameters are appropriately determined so that the detection of the defect size is not missed. Here, if one of the reference area and the reference number is changed, the other can be automatically changed.

The criterion used in the area determination II does not need to be directly set by the area itself, but needs to be a reference value that can be converted to the reference area S in pixels.

When the reference area S is set directly or indirectly as described above, the smallest square having a larger area is considered, and the minimum processing that can eliminate the square by the contraction processing in units of pixels is considered. The number of times is obtained and set as the reference number N. The relationship between the reference area S and the reference number N will be further described with a specific example.

Now, assuming that the area of the object of interest is 1000 pixels and the reference value of area determination II is 5% of the area of the object of interest, that is, the reference area S = 50 pixels, the smallest square larger than this area is 8 pixels. It means × 8 pixels.

The expansion and contraction processing is performed using a filter having a unit of 3 × 3, that is, a filter having a size of 3 × 3 pixels. For simplicity, this will be described by taking as an example a case where a binary image composed of 4 × 4 squares is contracted.
It will be like 0. FIG. 10A shows an image in which the entire periphery of a square formed of pixels having a pixel value of 1 has a pixel value of 0.
This shows a state in which a 3 × 3 contraction filter F indicated by a bold line is applied to one pixel at the top left. The contraction process is
If there is at least one 0 pixel in the filter, the process of changing the center 1 pixel to 0 is to be repeated clockwise or counterclockwise. When it becomes, the number of times of contraction processing is counted. That is, looking at the arrangement of the upper and lower or left and right pixels, the area of two pixels is reduced by one contraction process. Therefore, the 4 × 4 square disappears in a total of two contraction processes as shown in FIG.

The same principle is shown in FIG. 11 assuming that only pixels having a pixel value of 1 except for surrounding pixels having a pixel value of 0 correspond to one cell. A) ~
It turns out that it disappears in total four times until (E).

As described above, in this embodiment, the area judgment II
By setting a reference area of 50 pixels and setting a reference count of 4 in the dilation / shrinkage processing, it is possible to detect a defect of an arbitrary size as described above, and specifically detect a minute defect. It is possible to do.

In this embodiment, when the reference area S is set, if the smallest square having the area S or more is L × L pixels, the reference number N is set to the minimum integer of L / 2 or more. , Can be set automatically.

Conversely, when the reference number N is set, by setting the reference area S to an arbitrary number of pixels of (2N) ² ≧ S> (2N−1) ² , for example, an intermediate value, It can also be set automatically. By linking the parameters of the plurality of algorithms in this way, the setting can be facilitated, and it is possible to prevent a defect due to a human error from being overlooked.

According to the embodiment described above, even if the pattern arrangement has no regularity, any inspection object can be inspected as long as the pattern itself has substantially the same shape. it can. Therefore, it is possible to inspect products having different pitches between patterns at the central portion and the peripheral portion or even products having no symmetrical structure, and to inspect a pattern located on an axis of symmetry and a point of symmetry. Is possible.

Although the present invention has been specifically described above, the present invention is not limited to the above-described embodiment, and can be variously modified without departing from the gist thereof.

For example, the specific configuration of the defect inspection apparatus according to the present invention is not limited to the one shown in the above embodiment.

In the above-described embodiment, the case where an isolated region is extracted from a binary image has been described. However, a grayscale image region may be extracted from an inspection image before binarization. However, in this case, when comparing the image with the original image in the isolated area, a difference operation is performed instead of an exclusive OR operation.

Further, in the above-described embodiment, the case where one threshold value is set to perform binarization has been described.
Pixels having luminance within both ranges may be binarized. In this case, the small gray area shown in FIG. 5 can be selectively extracted.

The object to be inspected is not limited to the shadow mask, and the shape of the pattern and the shape of the area near the object are not limited to those described in the above embodiment.

In the above-described embodiment, the case where the reflected light image of the work is captured under the epi-illumination has been described. However, the transmitted light image may be captured by illuminating the work from the side opposite to the camera.
In this case, the isolated area (labeling portion) becomes white contrary to the above embodiment.

[0057]

As described above, according to the present invention,
A work in which the same pattern is repeatedly formed can be inspected with high accuracy without requiring a large memory capacity, even when the pattern arrangement has no symmetry.

[Brief description of the drawings]

FIG. 1 is a schematic perspective view showing a defect inspection apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an outline of a control system of the inspection apparatus.

FIG. 3 is a flowchart showing the operation of the embodiment.

FIG. 4 is an explanatory diagram showing a scanning order of a camera.

FIG. 5 is an explanatory diagram showing an image of an inspection image obtained by imaging a work;

FIG. 6 is an explanatory diagram showing an image of a binary image obtained from an inspection image.

FIG. 7 is an explanatory diagram showing an image of a labeling portion and a center of gravity on a binary image;

FIG. 8 is an explanatory diagram showing a relationship between an object of interest and objects near 6;

FIG. 9 is an explanatory diagram showing an image of an exclusive OR operation;

FIG. 10 is an explanatory diagram showing the principle of contraction processing.

FIG. 11 is an explanatory diagram showing a contraction process of a square of 8 × 8 pixels.

FIG. 12 is an explanatory view showing an image of a pattern formed repeatedly with a workpiece.

FIG. 13 is an explanatory diagram showing a geometrically symmetric structure of a work.

FIG. 14 is an explanatory diagram showing an image of a scanning order of two cameras.

FIG. 15 is an explanatory diagram showing an image of capturing an entire image by one camera.

FIG. 16 is an explanatory diagram showing an axisymmetric defect existing on the symmetry axis.

FIG. 17 is a schematic perspective view illustrating physical interference between two cameras.

[Explanation of symbols]

 10, 28 buffer 12 inspection stage 14, 26 robot 16 X-axis stage 18 line sensor camera 20 Y-axis stage 22 illumination unit 24 image processing unit 30 dust box 32 control unit

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ Identification FI FI Theme Court ゛ (Reference) H01J 9/42 H01J 9/42 A (72) Inventor Naoki Hata 1-chome, Ichigaya-cho, Shinjuku-ku, Tokyo No. 1 Inside Dai Nippon Printing Co., Ltd. (72) Inventor Masato Ushikusa 1-1-1 Ichigaya Kagacho, Shinjuku-ku, Tokyo F-term inside Dai Nippon Printing Co., Ltd. 2G051 AA90 AB02 AC21 CA03 DA03 DA07 EA11 ED15 ED21 ED23 5B057 AA01 BA02 CA12 CA16 DA03 DB02 DC04 5C012 AA02 BE03 5L096 BA03 CA02 EA02 FA09 FA52 FA59 JA11

Claims (14)

[Claims] 1. A defect inspection method for capturing an inspection object in which substantially the same pattern is repeatedly formed, inputting an inspection image, and performing image processing on the input inspection image to inspect a pattern for defects. Extracting an isolated region different from the luminance of the background from the inspection image; calculating an area and a center of gravity in units of pixels for each of the extracted isolated regions; An area comparison is performed to compare the area of the isolated area with the target area. If the area difference between the two or more isolated areas exceeds the reference area, it is determined that the pattern corresponding to the target isolated area has a defect. When the area difference is equal to or smaller than the reference area, the minimum processing number at which the reference area disappears when the contraction processing is performed in units of pixels is set as a reference number, and the original image of the isolated area of interest is set in units of pixels. Number After performing the expansion processing, the opening image is created by performing the same number of contractions, and the original image is subjected to the reference number of contractions in units of pixels, and then the same number of times is subjected to the same number of expansions to create a closing image. Comparing the created opening image and the closing image with each other, and determining that the pattern corresponding to the isolated region of interest has a defect smaller than the reference area if at least one of the images has an image difference. Defect inspection method. 2. The area of each of the extracted isolated regions is calculated, and if the calculated area is significantly different from the area of the normal pattern, the corresponding isolated region is excluded from the target of the area comparison. The defect inspection method according to claim 1, wherein: 3. The reference area is S pixels, and the reference number is N.
The minimum number of processing times required until a minimum square of S pixels or more disappears by contraction processing in units of pixels is set to a reference number N.
3. The defect inspection method according to 1. 4. When the reference area S is set, assuming that the smallest square not less than S is L × L pixels, the reference number N is automatically set to a minimum integer of L / 2 or more. When N is set, the reference area S is
4. The defect inspection method according to claim 3, wherein an integer greater than (2N-1) ^{2 and} equal to or less than (2N) ² is automatically set. 5. The method according to claim 1, wherein the inspection image is binarized by a predetermined threshold value to generate a binary image, and the isolated region is extracted from the generated binary image.
3. The defect inspection method according to 1. 6. A binary image is created from the inspection image by binarizing pixels having a luminance within a range of two preset threshold values, and the isolated area is extracted from the created binary image. The defect inspection method according to claim 1, wherein: 7. The defect inspection method according to claim 5, wherein the operation of comparing the images is an exclusive OR operation. 8. A defect inspection apparatus for capturing an inspection object in which substantially the same pattern is repeatedly formed, inputting an inspection image, and performing image processing on the input inspection image to inspect a pattern for defects. Extracting means for extracting an isolated area different from the luminance of the background from the inspection image; for each extracted isolated area, means for calculating an area and a center of gravity in units of pixels; Means for comparing the area of the isolated area near the center of gravity with the area of the nearby isolated area; and a pattern corresponding to the noted isolated area when an area difference between two or more isolated areas exceeds a reference area. Means for determining that there is a defect, and if the area difference is equal to or less than the reference area, a reference setting means for setting the minimum number of times the reference area disappears when the contraction processing is performed in units of pixels as the reference number The original image of the isolated area of interest is subjected to a reference number of expansions per pixel and then contracted the same number of times to create an opening image. Means for performing a number of times of expansion processing to create a closing image; and comparing the original image with the created opening image and the created closing image, respectively. A defect inspection apparatus for determining that the pattern has a defect having a size equal to or smaller than the reference area. 9. A means for calculating the area of each of the extracted isolated areas, and, if the calculated area is significantly different from the area of the normal pattern, the corresponding isolated area is excluded from the area comparison. The defect inspection apparatus according to claim 8, further comprising: 10. When the reference area is S pixels and the reference number is N, the reference setting means determines a minimum square required for a minimum square equal to or larger than S pixels to disappear by contraction processing in units of pixels. 9. The defect inspection apparatus according to claim 8, wherein the number of times of processing is set to a reference number of times N. 11. When the reference setting means sets the reference area S, if a minimum square equal to or larger than S is an L × L pixel, a minimum integer equal to or greater than L / 2 is automatically set in the reference number N. When setting the reference number N, the reference area S:
The defect inspection apparatus according to claim 10, wherein an integer greater than (2N-1) ^{2 and} equal to or less than (2N) ² is automatically set. 12. A function of the extraction means for forming a binary image by binarizing the inspection image with a predetermined threshold, and a function of extracting the isolated region from the generated binary image. 9. The defect inspection apparatus according to claim 8, comprising: 13. A function of the extracting means for binarizing pixels having a luminance within a range of two preset threshold values from the inspection image to create a binary image, and 9. The defect inspection apparatus according to claim 8, wherein the defect inspection apparatus has a function of extracting the isolated area from the data. 14. The defect inspection apparatus according to claim 12, wherein the operation of comparing the images by the image determining means is an exclusive OR operation.