JP2006266752A - Defect detection method, defect inspection method, defect detection device, defect inspection device, defect detection program and recording medium for recording program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an optical panel inspection method capable of performing properly rank determination based on only display defects having a prescribed area or more. <P>SOLUTION: An image by a liquid crystal panel is imaged, to thereby acquire image data (inspection image acquisition process). Brightness data of each pixel of a point defect or a strain defect having a smaller area than an irregular defect in the image data acquired in the inspection image acquisition process is replaced with an interpolation value based on brightness data of a peripheral pixel, to thereby repair the non-evaluation object defect species (point/strain defect repair process ST200). An irregular defect is detected in image data wherein the point/strain defect is repaired in the point/strain defect repair process ST200 (irregular defect detection process ST300). Rank determination of the liquid crystal panel is performed based on a defect area of the irregular defect and average brightness detected in the irregular defect detection process (ST300) (rank determination process ST400). <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a defect detection method, a defect inspection method, a defect detection apparatus, a defect inspection apparatus, a defect detection program, and a recording medium on which the program is recorded. For example, when inspecting an optical panel such as a liquid crystal panel, the present invention relates to a method for detecting a mura defect which is a display defect having a predetermined area or more.

As an optical panel for displaying an image, a liquid crystal panel having a liquid crystal cell disposed in each pixel on a display surface and performing display control by controlling a voltage applied to the liquid crystal cell is known.
In such a liquid crystal panel, display defects such as luminance unevenness may occur due to variations in the thickness of the liquid crystal cell, etc., so an image display for inspecting the liquid crystal panel for display defects during the manufacturing process, etc. Inspection is performed.
Conventionally, in an image display inspection of a liquid crystal panel, humans have actually visually checked the presence or absence of defects. However, in recent years, in order to increase the production amount and improve the inspection efficiency, the human eyes No automatic inspection has been performed (Patent Document 1).

  In the automatic inspection, for example, a display image on the liquid crystal panel is picked up by a CCD camera, and the luminance for each pixel is calculated based on the picked-up data. Further, a predetermined defect emphasis filter is applied to the entire display image, and an emphasis luminance value that emphasizes a luminance difference from surrounding pixels is calculated for each pixel. Then, based on the emphasized luminance value for each pixel, an area where a predetermined number or more of pixels having a luminance difference equal to or greater than a predetermined threshold with respect to the surroundings is detected as a display defect.

Here, point defects 710, spot defects 720, unevenness defects 730, and the like are known as display defects to be detected. FIG. 25 is a diagram illustrating an example of a point defect 710, a spot defect 720, and a mura defect 730.
A point defect 710 is a defect having a very high contrast although it has a dot shape and an extremely small area (about 1 to 5 pixels). As the point defect 710, there are a bright spot defect 711 having higher luminance than the surroundings and a black point defect having lower luminance than the surroundings.
The spot defect is a defect having a larger area than that of the point defect 710 (up to about 48 pixels in diameter), and varies from a high-contrast stain defect to a low-contrast stain defect. As the spot defect, there are a white spot defect having a higher brightness than the surroundings and a black spot defect having a lower brightness than the surroundings.
The mura defect 730 has a relatively large area and varies from a mura defect having a high contrast to a mura defect having a low contrast. As the mura defect 730, there are a white mura defect having higher luminance than the surroundings and a black mura defect having lower luminance than the surroundings.

  When each defect is detected in this way, factors such as the area of the defect and contrast are comprehensively determined to determine whether the liquid crystal panel is good or bad, and rank A to D among non-defective products. Is determined.

  Such good / bad judgment and ranking need to be a judgment that approximates an impression when an actual user looks at the display screen. For example, even if the defect contrast is low, the defect area is large. For example, since a large defect is conspicuous in the appearance of the user, such a liquid crystal panel is a defective product. On the other hand, even if the defect contrast is somewhat high, if the defect area is small, the defect is not conspicuous, and is judged as a non-defective product.

For example, FIG. 26 shows an example of a threshold curve 610 set as the determination threshold based on the relationship between the defect area and contrast.
The area and contrast of the detected defects are plotted on a graph 610 (FIG. 26), and whether the liquid crystal panel to be inspected is in the non-defective range or, in the case of the non-defective range, which region from A rank to D rank Is determined.

JP 2004-212311 A

As an aspect of the defect generated on the display screen of the liquid crystal panel, the mura defect 730 and the point defect 710 overlap each other, or the mura defect 730 and the spot defect 720 overlap each other, so that the point defect is within the mura defect 730 region. 710 and a spot defect 720 may exist.
For example, the white spot defect 721 is included in the white spot defect 731 (A in FIG. 27), the black spot defect 722 is included in the white spot defect 731 (B in FIG. 27), or the black spot defect 732 is present. May have a bright spot defect 711 (C in FIG. 27).
Even if the mura defect 730 and the spot / spot defect (710, 720) overlap as described above, if the defect is detected by the defect enhancement filter, the mura defect 730 area, the spot defect 720 area, and the point defect 710 are detected. Each of these areas can be detected.

  Here, for example, when the white spot defect 721 is included in the white spot defect 731 (A in FIG. 27), in performing rank determination based on the spot defect 730, the defect contrast 730 is simply determined as the defect contrast in the region of the spot defect 730. It is conceivable that an average value of contrast is calculated, and rank determination is performed by applying the defect area and defect contrast to the threshold curve 610. However, since the spot defect 720 is present in the region of the mura defect 730, the average contrast value includes the contrast of the white spot defect 721, and the rank evaluation based on the mura defect 730 has a contrast evaluation value. It will be too expensive.

That is, if rank determination based on display defects (point defect 710, spot defect 720, unevenness defect 730) is performed by human eyes, each defect is recognized as a separate defect even if the defects overlap. Rank determination is performed for each type (point defect 710, spot defect 720, mura defect 730), and the contrast of the mura defect 730 is used as the reference for the defect contrast in the rank determination based on the mura defect 730.
Then, in the automatic inspection, when the average contrast is simply calculated including the white spot defect 721 in the region of the uneven defect 730, the contrast of the uneven defect 730 is more than the recognition of the human eye because the white spot defect 721 is included. As a result, the rank determination result differs between rank determination by human eyes and rank determination by automatic inspection.

  Further, when the black spot defect 722 is included in the white spot defect 731 (B in FIG. 27), when the average contrast is calculated in the area of the spot defect 730, the contrast between the white spot defect 731 and the black spot defect 722 is obtained. This cancels out, and the contrast of the mura defect 730 cannot be detected. Similarly, when the black spot defect 732 includes the bright spot defect 711 (C in FIG. 27), the contrast is offset and the contrast of the spot defect 730 cannot be detected.

Here, in the rank determination, the point defect 710 and the spot defect 720 existing in the area of the uneven defect 730 may be removed, but the area of the point defect 710 and the spot defect 720 is simply removed from the area of the uneven defect 730. If the point defect 710 and the spot defect 720 are not removed, the area of the mura defect becomes small. Then, the evaluation value of the defect area is underestimated, and the result of the rank determination is not appropriate. Here, FIG. 28 is a conceptual diagram showing a state where the spot defect 720 existing in the unevenness defect 730 is removed.
In addition, it is conceivable to correct the contrast of the point defect 710 and the spot defect 720 existing in the unevenness defect 730 by substituting the contrast with the average contrast of the entire image (see FIG. 29). Since the difference in contrast between the portion and the portion to be corrected is too large, it is still possible to make an appropriate rank determination.

  An object of the present invention is to detect a defect in an inspection object, a defect detection method, a defect inspection method, a defect detection device, a defect inspection device, a defect detection program, and a program for recording the program that can appropriately detect a defect larger than a predetermined area. Is to provide a recording medium.

  The defect detection method of the present invention is a defect type having an area less than a predetermined size in the image data acquired in the inspection image acquisition step of imaging the inspection object and acquiring the image data. A non-evaluation target defect type repair process for repairing the non-evaluation target defect type by replacing the luminance data of each pixel of the non-evaluation target defect type with an interpolated value based on the luminance data of surrounding pixels, and repairing the non-evaluation target defect type And an evaluation target defect type detection step of detecting an evaluation target defect type whose area is a defect type having a predetermined size or more in the image data in which the non-evaluation target defect type is repaired in the process.

In such a configuration, first, an inspection object is imaged and its image data is acquired (image data acquisition step).
Next, the non-evaluation target defect type is repaired by replacing each pixel of the non-evaluation target defect type present in the acquired image data with an interpolation value based on luminance data of surrounding pixels. As a result, the luminance data of each pixel of the non-evaluation target defect type is replaced so as to follow the luminance data change of the surrounding pixels, and what protrudes or is recessed from the luminance of the surrounding pixels smoothly adapts to the surrounding luminance change. As repaired.
In this way, the evaluation target defect type is detected from the image data in which the non-evaluation defect type is corrected (evaluation target defect type detection step).
Then, for example, the inspection object is evaluated based on the detected evaluation object defect type. Examples of such evaluation include good / bad determination evaluation based on quantitative evaluation values such as defect area and defect luminance (for example, average luminance of defective portions).

Here, when an evaluation based only on display defects (evaluation target defect type) having a predetermined area or more is desired, a display defect (non-evaluation target display defect type) having an area smaller than the evaluation target defect type overlaps with the evaluation target defect type. It may exist in the state.
In such a case, the evaluation target defect type detection process is simply performed on the image data acquired in the inspection image acquisition process, and only the evaluation target defect type is evaluated based on the detected display defect type. Cannot make an evaluation based on. For example, when calculating the average luminance within the area of the detected evaluation target defect type, the average luminance value including the luminance of the non-evaluation target defect type that overlaps the evaluation target defect type will be calculated. This is because when the evaluation target defect type and the non-evaluation target defect type have different luminances, the average luminance based only on the evaluation target defect type is not obtained.

  In this regard, in the present invention, the non-evaluation target defect type is repaired in advance in the non-evaluation target defect type repair process, and the evaluation target defect type is detected from the image data in which the non-evaluation target defect type is repaired. Even if the non-evaluation target defect type overlaps with the evaluation target defect type in the image data, the repaired non-evaluation defect type is not detected and only the evaluation target defect type is detected appropriately. .

Then, when repairing the non-evaluation target defect type, the luminance data of the pixel of the non-evaluation defect type is replaced with an interpolation value, and the interpolation value in the non-evaluation target defect type repair process is the luminance of the pixels located around each pixel. It is a value based on data and reflects surrounding trends.
For example, when the luminance data of each pixel of the non-evaluation target defect type that overlaps the evaluation target defect type is replaced with an interpolated value, the non-evaluation target defect type region becomes luminance data that follows the tendency of the luminance change of the evaluation target defect type. Is replaced.
As a result, the average luminance of the evaluation target defect type after the pixel luminance data of the non-evaluation target defect type is replaced with the interpolation value in this way approximates the average luminance based only on the evaluation target defect type, Furthermore, the average luminance is based on the assumption that the defect type to be evaluated continues in that region instead of the non-evaluated defect type.

  Here, for example, if the luminance data of each pixel of the non-evaluation target defect type is simply replaced with the average value of the entire image data, the luminance value becomes extremely different from the surroundings, so the luminance value of the non-evaluation target defect type region Is far away from the surroundings, and the average luminance value in the defect type to be evaluated varies greatly.

  In this regard, in the non-evaluation defect repair process of the present invention, since the interpolation value reflects the luminance data of the surrounding pixels, it is detected in the image after repairing the non-evaluation defect type that overlaps the defect type to be evaluated. The average luminance in the region of the defect type to be evaluated is exactly the average luminance based on the defect type to be evaluated.

In addition, in repairing the non-evaluation defect type, the area of the non-evaluation defect type is not deleted from the image data because the area of the non-evaluation defect type has not existed from the beginning. The luminance value is replaced with the interpolation value.
Therefore, the area does not change before and after the repair, and the area of the defect type to be evaluated changes even if the defect type of the non-evaluation target defect is repaired at the portion where the defect type to be evaluated and the non-evaluation target defect type overlap. There is no.
Therefore, the evaluation target defect type is detected from the image data after the non-evaluation defect type repair process is performed in the non-evaluation defect type repair process (evaluation target defect type detection process). As a quantitative evaluation value, for example, when a defect area or average luminance is calculated, the defect area does not change before and after repair of a non-evaluation target defect type, and the average luminance is a value based only on the evaluation target defect type. It becomes.

That is, if the defect type to be evaluated is detected in the image data after repairing the non-evaluation defect type and the area and average brightness of the detected defect type is simply calculated, the human eye will recognize it. Values corresponding to the defect area and the average luminance can be obtained.
As a result, for example, it is possible to appropriately evaluate the inspection object based on the defect area and average luminance of the defect type to be evaluated, and to obtain an evaluation that approximates the impression of human appearance.

  In the present invention, in the non-evaluation target defect type repairing step, luminance data of pixels satisfying a predetermined condition around each pixel of the non-evaluation target defect type tends to change toward each pixel of the non-evaluation target defect type. It is preferable to calculate the interpolation value based on

In such a configuration, when calculating the interpolation value in the non-evaluation target defect type repair process, a value based on the tendency of luminance data change of pixels around each pixel of the non-evaluation target defect type is used as the interpolation value of each pixel. calculate. This interpolated value is a value based on the tendency (change rate) of surrounding pixels to change toward each pixel of the non-evaluation target defect type, and is a value along the surrounding tendency.
According to such a configuration, when the non-evaluation target defect type overlaps the evaluation target defect type, the luminance of each pixel of the non-evaluation target defect type is replaced with an interpolated value obtained by extending the surrounding luminance change. Therefore, the average brightness calculated based only on the defect type to be evaluated and the average brightness of all areas of the defect type to be evaluated including the area of the non-evaluation target defect type repaired after repairing the non-evaluation target defect type Is an approximate value, and it is an average luminance when it is assumed that the defect type to be evaluated continues not continuously in the non-evaluated defect type but also in the region.

  That is, if the defect type to be evaluated is detected in the image data after repairing the non-evaluation defect type and the area and average brightness of the detected defect type is simply calculated, the human eye will recognize it. Values corresponding to the defect area and the average luminance can be obtained. An evaluation result that is the same as that of the human eye can be obtained by performing an evaluation that is compared with a predetermined threshold value based on the quantitative evaluation values such as the area of the defect type to be evaluated and the average luminance calculated in this way.

  In the present invention, the non-evaluation defect type repair step includes a non-evaluation target defect type detection step of detecting the non-evaluation target defect type in the image data acquired in the image data acquisition step, and the interpolation value as the interpolation value. An interpolation step that calculates for each pixel of the non-evaluation target defect type and replaces the luminance data of each pixel with the interpolation value, and the interpolation step is performed on the outer edge of the detected non-evaluation target defect type. The interpolated values of the pixels closer to the inner side are calculated starting from the located pixels, and when obtaining the interpolated values of the pixels located on the outer edge, the luminance data of normal pixels adjacent to the pixel and having normal luminance are obtained. The interpolated value is calculated based on the inner pixel and the interpolated value of the pixel closer to the inner side is already calculated as the luminance data of the normal pixel and the interpolated value of the pixel closer to the outer edge. It is preferable to calculate the interpolated value based on those adjacent to the pixel which is the target for calculating the interpolated value of the between values.

In such a configuration, in performing the non-evaluation target defect type repair process for repairing the non-evaluation target defect type, first, the non-evaluation target defect type having a smaller area than the evaluation target defect type is detected by the non-evaluation target defect type detection process. To detect. Then, an interpolation value of each pixel of the detected non-evaluation target defect type is calculated, and luminance data of each pixel of the non-evaluation target defect type is replaced with this interpolation value (interpolation process).
As a result, the luminance data of each pixel of the non-evaluation target defect type is replaced so as to follow the luminance data change of the surrounding pixels, and what protrudes or is recessed from the luminance of the surrounding pixels smoothly adapts to the surrounding luminance change. As repaired. At this time, when the luminance data of each pixel of the non-evaluation target defect type is replaced with the interpolation value in the interpolation process, first, an interpolation value is calculated for each pixel located at the outer edge of the non-evaluation target defect type.
Since there is a normal pixel that has normal brightness and is not included in the non-evaluation target defect type at a position adjacent to the outer edge pixel of the non-evaluation target defect type, it is adjacent to the outer edge of the non-evaluation target defect type pixel. The interpolated value calculated based on the luminance data of the normal pixels is a value along the luminance data change of these normal pixels.

That is, the outer edge pixel is replaced with an interpolated value reflecting the luminance change tendency of the normal pixel located just outside the non-evaluation target defect type.
Then, after correcting the pixel located at the outer edge with the interpolation value, the interpolation value is sequentially calculated for each pixel located inside the non-evaluation target defect type.
When calculating the interpolated value of the inner pixel, the interpolated value is calculated based on the luminance data of the normal pixel and the interpolated value of the already corrected pixel adjacent to the pixel for which the interpolated value is to be calculated. calculate.
If an interpolation value is calculated based on adjacent normal pixels, it goes without saying that this interpolation value becomes a value along the luminance data change of normal pixels not included in the non-evaluation target defect type.

Here, there is not a normal pixel adjacent to all pixels of the detected non-evaluation defect type, and in particular, only a defective pixel exists in the vicinity of the pixel closer to the non-evaluation target defect type. There is also a possibility. In this case, since the interpolation value is calculated from the pixels closer to the outer edge in the interpolation step, when calculating the interpolation value for the inner pixel, the interpolation value is already calculated for the outer pixel adjacent to this pixel. It will be.
Therefore, when the interpolation value is calculated for the inner pixel, the interpolation value can be calculated using the value already calculated as the interpolation value for the outer pixel. The interpolation value of the outer pixel is an interpolation value calculated for the outer pixel, whereas the interpolation value calculated for the outermost pixel is calculated based on normal pixels located outside the non-evaluation target defect type. Value. Therefore, by calculating the interpolation value in order from the outside to the inside, the interpolation value calculated based on the already calculated interpolation value also changes the luminance data of normal pixels outside the non-evaluation target defect type. It becomes a value that reflects and relays.

  Thus, in calculating the interpolation value, the interpolation value is calculated in order from the outside to the inside of the detected non-evaluation target defect type, thereby reflecting the luminance data change of the normal pixels outside the non-evaluation target defect type. Interpolation values can be calculated in order, and the non-evaluation target defect type can be repaired by replacing the luminance data of each pixel of the non-evaluation target defect type with the interpolation value thus calculated. Since the interpolation value of the inner pixel can be calculated using the interpolation value already calculated for the outer pixel even for the pixel existing on the inner side of the non-evaluation target defect type, Interpolation values can be calculated and replaced for all pixels.

  In the present invention, the non-evaluation target defect type detection step includes an image reduction step of reducing the image data to be inspected, and a defect enhancement filter that emphasizes a difference in luminance data between the pixel of interest and pixels around the pixel of interest. A filtering step of detecting display defects over the image data reduced in the image reduction step, the non-evaluation target defect type repairing step filtering the image data reduced to a predetermined size in the image reduction step After the interpolation value of the pixel of the non-evaluation target defect type detected in the process is calculated in the interpolation process and the non-evaluation target defect type is repaired, the image reduction process is further performed on the repaired image data. It is preferable that the non-evaluation target defect type of the area to be repaired is repaired sequentially by repeating the filtering process and the interpolation process.

  In such a configuration, the non-evaluation target defect type detection step reduces the image data to a predetermined size in the image reduction step, and then applies a defect enhancement filter for defect detection to the reduced image data. The non-evaluation target defect type is detected. As a method for reducing the image data, for example, four or six pixels in the image data are combined into one pixel, and the luminance data of the combined pixels is an average of the original four or six pixels. As an example.

Here, the defect emphasis filter has a predetermined size, and the size of the display defect detected by the defect emphasis filter of this size matches the defect emphasis filter.
Therefore, in the non-evaluation target defect type detection step, first, the image data is reduced in the image reduction step.
Then, the non-evaluation target defect type that is larger than the detection size of the defect enhancement filter in the initial image size is reduced to a size that matches the defect enhancement filter.

When a defect enhancement filter is applied to the image data reduced in this way, a display defect having a size matching the defect enhancement filter is detected.
When image reduction is performed at several stages of reduction ratios, the sizes of non-evaluation target defect types that have various sizes in the initial image data match the size of the defect enhancement filter at any reduction ratio.
Therefore, by applying a defect emphasis filter to the reduced image data while reducing the image data in several stages, it is possible to detect all non-evaluation target defect types of various sizes in the image data.

Further, a defect emphasis filter is applied to the image data reduced to a predetermined size to detect a display defect of a predetermined size, and the detected display defect is repaired by an interpolation process. At this stage, the display defect that does not apply to the defect emphasis filter at this reduction rate has not been repaired yet, and further image reduction is performed on the image data with the display defect repaired at a predetermined reduction rate.
When a defect emphasis filter is applied to the reduced image data, display defects that have not been repaired are applied to the defect detection filter, so that the detected display defects are repaired in an interpolation process.
In this way, the image reduction process, the filtering process, and the interpolation process are repeated, and the reduction rate of the image data is determined so that all display defects of the size of the non-evaluation target defect type are applied to the defect detection filter. All of these are repaired.

  In the present invention, the defect type to be evaluated is a mura defect, the defect type to be non-evaluated is a spot defect having a smaller area than the mura defect and a point defect having a smaller area than the spot defect, and the non-evaluation In the target defect type repairing step, the point defect is detected in the filtering step for the image data reduced in the first image reduction step, and the interpolation value of the point defect is calculated in the interpolation step. The image data in which the point defect is repaired is reduced in the image reduction process, the spot defect is detected in the filtering process, and an interpolation value of the spot defect is calculated in the interpolation process. It is preferable to repair.

In such a configuration, the defect type to be evaluated is a mura defect, and the defect type to be evaluated is a spot defect and a point defect.
Conventionally, when it is desired to evaluate an optical panel based on a mura defect (defect type to be evaluated), if a spot defect or a point defect (non-evaluation type) that is smaller than the mura defect exists on the mura defect, Since the defect area and the average luminance cannot be calculated as a quantitative evaluation value based only on the defect, appropriate evaluation based only on the unevenness defect cannot be performed.
In this regard, in the present invention, it is possible to repair point defects and spot defects by repeating the image reduction process, the filtering process, and the interpolation process.

That is, the point defect in the image data is matched with the size of the defect enhancement filter by the reduction of the image data in the first image reduction process, and is detected by the defect enhancement filter.
The point defect is repaired by replacing the luminance data of each pixel of the detected point defect with the interpolation value calculated in the interpolation process (interpolation process).
Then, the image data with the point defects corrected is further reduced in an image reduction process, and a defect enhancement filter is applied to the reduced image data to detect a spot defect. The brightness data of each pixel of the detected spot defect is replaced with the interpolation value calculated in the interpolation process and repaired.
Then, image data in which point defects and spot defects, which are non-evaluation target defects, are repaired is obtained.
In this way, spot defects and spot defects are detected by detecting the mura defect, which is the defect type to be evaluated, from the image data corrected, and evaluating the optical panel based on the detected mura defect. The optical panel can be appropriately evaluated based only on the mura defect that is the evaluation target defect type without being affected.

  In the present invention, the non-evaluation target defect type detection step compares the luminance data of each pixel of the image data with a predetermined threshold value, and a defective pixel whose luminance data is greater than or equal to the threshold value and a normal pixel whose luminance data is less than the threshold value, And a binarization step for generating a binarized array space in which the defective pixel is 1 and the normal pixel is 0, and the pre-interpolation step is detected in the non-evaluation target defect type detection step. Further, a repair target pixel search step for sequentially searching for a pixel to be repaired among defective pixels of the non-evaluation target defect type, and an interpolation value of the pixel to be repaired searched in the repair target pixel search step is calculated. An interpolated value calculating step, wherein a 1 is arranged in a central cell of interest in a matrix arrangement of a predetermined size, and two or more 0s are continuously arranged in at least 5 or more of the surrounding 8 directions To do A plurality of matching filters are prepared, and the repair target pixel search step sequentially applies the matching filter to the binarized array space to find an array portion of 1 and 0 that matches one of the matching filters. Searching in the binarized array space, recognizing in the image data a pixel corresponding to the target cell of the matching filter at the array location that matches the matching filter as a pixel to be repaired, and the interpolation value calculating step includes: The interpolation value of the pixel to be repaired is calculated based on the rate of change of the luminance data of the pixel in the direction in which 0 is continuous for 2 or more, and the interpolation value is calculated in this interpolation value calculating step so that the luminance data becomes the interpolation value. For the replaced pixel, the corresponding cell is rewritten from 1 to 0 in the binary array space, and the It is preferred that the cells of the binarized sequence space is repeated repairing target pixel search process and the interpolation value calculating step until all 0.

In such a configuration, when detecting a non-evaluation target defect type from the image data acquired in the inspection image acquisition process (non-evaluation target defect type detection process), a defective pixel that has abnormal luminance and generates a display defect type A predetermined threshold value for distinguishing normal pixels having normal luminance is prepared, and defective pixels and normal pixels are detected based on this threshold value.
Then, a binarized array space having a size corresponding to the image data is generated. In this binarized array space, a position corresponding to the defective pixel is “1”, and a position corresponding to the normal pixel is “1”. 0 ″ is arranged (binarization process).
Then, a cell in which 1 is arranged in the binarized array space corresponds to a defective pixel, and a cell in which 0 is arranged corresponds to a normal pixel.

Next, when calculating the interpolation value of the defective pixel in the interpolation process, the interpolation value is not calculated for all defective pixels at the same time, but in order from the outer edge side to the inner side among the non-evaluation target defect types. When the interpolation value is calculated, the pixels to be repaired are searched in order in the repair target pixel search step.
In a repair target pixel search process for searching for a repair target pixel for which an interpolation value is to be calculated, a predetermined array of 1 and 0 is searched in order by applying a matching filter to the binarized array space.

Here, the matching filter covers a pattern in which 1 is arranged in the center target cell in a matrix arrangement of a predetermined size, and 0 is continuously arranged in five or more directions among the eight directions around the matching filter. Are prepared.
Here, “1” corresponds to a defective pixel, and “0” corresponds to a normal pixel. However, searching in the binary array space for an array location that matches the 1 and 0 array of the matching filter A defective pixel having a predetermined number or more of normal pixels is searched.

When the matching filter is applied to the binarized array space and an array location matching the matching filter array is found in the binarized array space, the cell corresponding to the cell of interest at the center of the matching filter is the object to be repaired It becomes.
Therefore, a position corresponding to the repair target found in the binary array space is searched from the image data, and the pixel is recognized as the repair target pixel. Furthermore, since the direction in which 0 continues is known from the matching filter array, the interpolation value of the repair target pixel is calculated based on the luminance data of the pixel in the direction in which 0 continues.
Then, an interpolation value of a pixel (defective pixel) searched as a repair target pixel is obtained based on the brightness data of a pixel adjacent to the repair target pixel and having normal brightness data.
The pixel to be repaired is repaired by replacing the luminance data of the pixel to be repaired with the calculated interpolation value.

Further, for the pixel corrected by the interpolation value calculated in this way, the corresponding binary array space cell is rewritten from 1 to 0.
As described above, the pixel corrected by replacing the luminance data with the interpolated value is replaced with 0 in the binarized array space. Therefore, when the matching filter is applied next, the pixel corrected with the interpolated value is also replaced. This is used as a basis for calculating the interpolation value.
In addition, by replacing the pixels corrected with the interpolation value with 0 in the binarized array space, the periphery is initially surrounded by defective pixels, and there are no more than a predetermined number of normal pixels in the periphery. As for the defective pixels, the surrounding pixels of the defective pixels are corrected with the interpolation values and replaced with 0 in the binarized array space, so that the surrounding 0s are gradually arranged. It starts to search for.
Therefore, all the defective pixels initially detected are sequentially searched by the matching filter.

Then, by applying a matching filter to the binarized array space and repeating the search for the pixel to be repaired and the calculation of the interpolation value, all the cells in the binarized array space become 0, and the defective pixel initially detected Repeat until all the repairs are made.
With such a configuration, the defective pixels in the image data are searched for in order as the repair target pixels for defective pixels having a predetermined number or more of luminance data of normal pixels or already calculated interpolation values. By replacing the luminance data of the repair target pixel with an appropriate interpolation value, it is possible to repair all detected defective pixels.

  In the present invention, it is preferable that in the repair target pixel search step, the binarized array space is sequentially applied from a matching filter having a small direction in which two or more zeros continue around the target cell.

  In such a configuration, when a matching filter is applied to the binarized array space, 1 is arranged in the center target cell in a matrix array of a predetermined size as a matching filter, and five of the surrounding eight directions are arranged. When a plurality of patterns in which 0s are continuously arranged in the above direction are prepared, the matching filter is sequentially applied to the binarized array space from the matching filter having few directions in which two or more 0s continue around the target cell.

Here, since the defect type non-evaluation target defect type to be repaired is generally configured by collecting defective pixels in a certain area, there are many defects around the target cell from the beginning in the binary array space. It is considered that there are few patterns in which 0s continue in the direction.
In such a situation, if priority is given from a matching filter in which 0 continues in many directions around the target cell, an array pattern that matches the matching filter is not found at first, and a pixel to be repaired is found. Only the matching process is wasted without being performed. In addition, since the matching process has to see matching / mismatching between a plurality of cells, the burden required for one matching process is very large, and a considerable time is required.
In this respect, in the present invention, since the filter is applied from a filter with few directions of zeros around the cell of interest, it is possible to increase the possibility of matching from the beginning, and to improve the processing efficiency.

  In the present invention, the interpolation value calculating step is twice the luminance data of the pixel adjacent to the repair target pixel in each direction in which two or more zeros are adjacent to the repair target pixel in the binarized array space. After calculating the basic interpolation value obtained by subtracting the luminance data of the outer pixels from the average value, the average value obtained by averaging the basic interpolation values calculated in each direction may be used as the interpolation value of the repair target pixel. preferable.

In such a configuration, for example, assuming that the luminance data of the pixels adjacent to the repair target pixel are L _α and L _{β in} order from the repair target pixel, the rate of change at which these two luminance data change toward the repair target pixel. Is (L _α -L _β ). Then, this change rate (L _α -L _β ) is added to L _α that is the luminance data of the pixel adjacent to the repair target pixel, thereby performing interpolation based on the change rates of L _α and L _β at the position of the repair target pixel. A basic value “L _α + (L _α −L _β )” is obtained.
By averaging the interpolation basic values calculated in each direction in this way to obtain the interpolation value of the pixel to be repaired, an interpolation value having an appropriate value reflecting the luminance change of the pixels located around the pixel to be repaired is obtained. Obtainable.

  The defect inspection method of the present invention includes a defect detection step for detecting an evaluation target defect type in the inspection target by the defect detection method, and a quantitative evaluation value of the evaluation target defect type detected in the defect detection step. And an evaluation process for evaluating the inspection object based on the evaluation object.

  According to such a configuration, first, in the defect detection step, the defect type to be evaluated can be detected from the image data after the non-evaluated defect type repairing means has repaired the non-evaluated defect type. And as a quantitative evaluation value of the detected evaluation target defect type, for example, when calculating the defect area and average luminance, the defect area does not change before and after repair of the non-evaluation target defect type, and The average luminance is a value based on the defect type to be evaluated. As a result, in evaluating the inspection object in the evaluation process, it is possible to perform an appropriate evaluation based on the defect type of the evaluation object, and to obtain an evaluation result that approximates the impression of human appearance.

  The defect detection apparatus of the present invention is an inspection image acquisition unit that captures an image of an inspection object and acquires the image data, and is a defect type whose area is less than a predetermined size in the image data acquired in the inspection image acquisition step. Non-evaluation target defect type repair means for repairing the non-evaluation target defect type by replacing the luminance data of each pixel of the non-evaluation target defect type with an interpolation value based on the luminance data of surrounding pixels, and the non-evaluation target defect type repair And an evaluation target defect type detection means for detecting an evaluation target defect type whose area is a defect type having a predetermined size or more in the image data in which the non-evaluation target defect type is repaired by the means.

  The defect detection program of the present invention incorporates a computer into a defect detection apparatus that detects a defect type to be evaluated that is a defect type having a predetermined area or more in an inspection target, and the computer images the inspection target and captures the image data thereof. Luminance data of each pixel of the non-evaluation target defect type which is a defect type whose area is less than a predetermined size in the image data acquired in the inspection image acquisition step The non-evaluation target defect type repairing means for repairing the non-evaluation target defect type by replacing with the interpolation value based on the non-evaluation target defect type repairing means, and the area in the image data in which the non-evaluation target defect type repairing means is repaired It functions as an evaluation target defect type detection means for detecting an evaluation target defect type that is a defect type of a predetermined size or larger.

  The recording medium of the present invention is characterized in that the defect detection program is recorded in a computer-readable manner.

According to such a configuration, the same effects as those of the above-described invention can be achieved.
In other words, the evaluation target defect type is detected from the image data after the non-evaluation defect type repairing means is repaired by the non-evaluation defect type repairing means (evaluation target defect type detection means), and the detected evaluation target defect type is quantitatively detected. For example, when the defect area and the average luminance are calculated as the evaluation value, the defect area does not change before and after the repair of the non-evaluation target defect type, and the average luminance is a value based on the evaluation target defect type. As a result, for example, when evaluating the inspection object, it is possible to perform an appropriate evaluation based on the defect type to be evaluated, and to obtain an evaluation result that approximates the impression of human appearance.

  Further, a CPU or a memory may be arranged in the defect detection apparatus so as to function as a computer, and a predetermined optical panel inspection program may be incorporated in the computer to function as each functional unit. In order to install a predetermined optical panel inspection program, a memory card, a CD-ROM or the like may be directly inserted, or a device for reading these storage media may be connected externally. Furthermore, the program may be supplied and installed by communication by connecting a LAN cable, a telephone line, or the like, or the program may be supplied and installed wirelessly.

  If the defect detection program of the present invention provided by such a recording medium or communication means such as the Internet is incorporated in the defect detection apparatus, it is easy to change predetermined parameters.

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments of the present invention will be illustrated and described with reference to reference numerals attached to respective elements in the drawings.
(First embodiment)
An optical panel inspection method as a first embodiment according to the defect inspection method of the present invention will be described.
FIG. 1 is a diagram illustrating an overall configuration of an optical panel inspection apparatus (defect inspection apparatus) that performs an optical panel inspection method.
The optical panel inspection apparatus 100 includes a liquid crystal light valve (optical panel) 112 to be inspected, an image projection unit 110 that projects an image, a pattern generator 120 that drives the liquid crystal light valve 112, and an image projection unit 110. A projection unit 130 having a screen 134 for projecting an image to be projected, a CCD camera (imaging means) 140 for imaging an image projected on the screen 134, and a mura defect (based on image data captured by the CCD camera 140) And an arithmetic processing unit 200 that performs rank determination (evaluation) on the basis of the detected nonuniformity defect 730.

The image projection unit 110 is a light source 111, a liquid crystal light valve (optical panel) 112 that is driven by the pattern generator 120 and modulates the light from the light source 111, and a projection lens that collects and projects the light from the liquid crystal light valve 112. 113 and a mirror 114 that reflects light from the projection lens 113 toward the screen 134.
The liquid crystal light valve 112 is a liquid crystal panel provided with a liquid crystal cell for each pixel. If each liquid crystal cell has a polarization failure or the like, a display defect occurs in a projected image. Examples of the display defect include a point defect (non-evaluation target defect type) 710, a spot defect (non-evaluation target defect type) 720, and a mura defect (evaluation target defect type) 730.
The image projecting unit 110 imitates a projection type image display device such as a projector, for example.

The pattern generator 120 applies a drive signal to each switching element that drives each liquid crystal cell of the liquid crystal light valve 112 to drive the liquid crystal light valve 112 in a predetermined pattern.
Here, in performing the image display inspection, it is inspected whether all the display pixels exhibit the same luminance at the same input level, and the pattern generator 120 drives all the liquid crystal cells at the same input level.

  The projection unit 130 includes a dark box 131 and a screen 134 provided in the dark box 131. The dark box 131 has an opening 132 into which light from the image projection unit 110 is incident, and an opening 133 for capturing an image on the screen 134 with the CCD camera 140.

The CCD camera 140 captures an image projected on the screen 134 and outputs the captured image data to the arithmetic processing unit 200.
Here, since the display defect of the liquid crystal light valve 112 is detected based on the image data captured by the CCD camera 140, the CCD camera 140 captures an image with the number of gradations exceeding the number of gradations of the liquid crystal light valve 112. It is necessary to have a function. For example, when 256 gradations are expressed by the liquid crystal light valve 112, the image data of the CCD camera 140 has the gradation number of 12-bit data in which “0” is black and “4095” is white.

The arithmetic processing unit 200 has a CPU (central processing unit) and a predetermined memory (storage device), and by incorporating an optical panel inspection program, as shown in FIG. The function as a spot defect repairing means (non-evaluation target defect type repairing means) 400, a mura defect detection means (evaluation target defect type detection means) 500, and a rank determination means (evaluation means) 600 is realized.
Here, the preprocessing unit 300 includes an inspection image acquisition unit 310, a background difference unit 320, an area extraction unit 330, and a luminance value calculation unit 340. The point / stain defect repair unit 400 includes a point / spot defect correction unit 400. A defect detection unit (non-evaluation target defect type detection unit) 410 and an interpolation unit 450; the point / stain defect detection unit 410 includes an image reduction unit 420, a filtering unit 430, and a binarization unit 440; The interpolation unit 450 includes a repair target pixel search unit 460 and an interpolation value calculation unit 470.
The operation of each means will be described with reference to the flowcharts of FIGS. 3, 4, and 5.

(Optical panel inspection method)
Next, the optical panel inspection method will be described with reference to the flowcharts of FIGS.
FIG. 3 is a flowchart showing a processing procedure of the optical panel inspection method.
As shown in the flowchart of FIG. 3, the optical panel inspection method has, as a whole structure, a preprocessing step (ST100) for acquiring image data by the liquid crystal light valve 112 to be inspected and preprocessing the image data, and preprocessing. The point defect 710 and the spot defect 720 are repaired by replacing the luminance data of each pixel of the point defect 710 and the spot defect 720 in the image data preprocessed in the step (ST100) with an interpolated value based on the luminance data of the surrounding pixels. In this image data, based on the point / spot defect repairing process (non-evaluation target defect type repairing process, ST200), and the point defect 710 and spot defect 720 repaired by the point / stain defect repairing process (ST200). Mura defect detection step (evaluation target defect type detection step, ST300) for detecting mura defect 730, and mura Recessed detection step (ST 300) rank determination step (evaluation step, ST400) for performing rank determination of the liquid crystal light valve 112 (Evaluation) Based on the quantitative evaluation value of the detected regions of nonuniformity defects 730 in provided with a.

FIG. 4 is a flowchart showing the processing procedure of the preprocessing step (ST100).
The pretreatment process of ST100 will be described with reference to the flowchart of FIG.
In the preprocessing step (ST100), first, an inspection image to be subjected to defect detection in ST110 is acquired (inspection image acquisition step). In obtaining the inspection image, the liquid crystal light valve 112 is set on the image projection unit 110 and an image is projected toward the screen 134, and this image is captured by the CCD camera 140. The captured image data is output to the inspection image acquisition unit 310 and stored.

  Next, in ST120, the background difference unit 320 performs a background difference step of subtracting the background image from the image data to be inspected. In the background difference unit 320, data of an image projected on the screen 134 using the liquid crystal light valve 112 that is known not to be defective in advance is set and stored as a background image. In the background difference step (ST120), the background image is subtracted from the image data stored in the inspection image acquisition unit 310. Then, for example, noise due to dirt on the screen 134 is removed from the image data, and image data reflecting only display defects due to the liquid crystal light valve 112 is obtained.

  Next, in ST130, the area extraction unit 330 performs an area extraction process for extracting only the screen portion to be inspected from the image data. That is, the image projected on the screen 134 is projected not on the entire surface of the screen 134 but on, for example, the central area excluding the edge portion. In the area extraction step (ST130), an unnecessary portion is trimmed from the image data and displayed. Extract only the image of the area.

  In ST140, the luminance value calculation unit 340 performs the luminance value calculation process on the image extracted in this way. An image picked up by the CCD camera 140 is image data for each image pickup element of the CCD camera 140, and one display pixel is picked up by a plurality of image pickup elements. Therefore, a value obtained by averaging the luminance values of the image sensor for each display pixel is calculated as the luminance value of each display pixel.

After the pretreatment process (ST100), a point / stain defect repairing step 400 is performed by the point / stain defect repairing means 400 in ST200.
FIG. 5 is a flowchart showing the processing procedure of the spot / stain defect repairing step.
In the point / spot defect repairing step (ST200), first, in ST210, the point / stain defect detecting step (non-evaluation target defect type detecting step) is performed by the point / stain defect detecting means 410. In the point / stain defect detection step (ST210), defect detection is performed by applying a defect enhancement filter 431 (see FIG. 7) of a predetermined size to the image data, and the point to be repaired in the point / stain defect repair step (ST200). While reducing the image size in several stages so that the defect 710 and the spot defect 720 can be detected by the defect enhancement filter 431 (ST211), the reduced image data is subjected to the defect enhancement filter 431 (ST212), and the defect (point defect 710) is detected. , Spot defect 720) is detected.

For example, when the defect enhancement filter 431 is applied in a state where the image size is halved, the size of the point defect 710 matches the size of the defect enhancement filter 431, and the point defect 710 is detected. When the defect enhancement filter 431 is applied in the state of / 8, the size of the stain defect 720 matches the size of the defect enhancement filter 431, and the stain defect 720 is detected.
Hereinafter, each step (ST211 to ST213) of the point / stain defect detection step (ST210) will be described.

  In the point / stain defect detection step (ST210), first, an image reduction step is performed by the image reduction means 420 in ST211. In the image reduction means 420, several stages of image reduction ratios are set in advance. For example, image reduction ratios of 1/2, 1/4, 1/8, etc. are set in the vertical and horizontal sizes. For example, the image is reduced to the original size shown in FIG. 6A by combining four pixels into one pixel. When combining four pixels, the average of the four pixels is taken as the luminance value. In this way, a reduced image (B in FIG. 6) obtained by reducing the original size to 1/2 in length and width and the number of pixels to 1/4 is obtained.

  Each cell in the image data thus reduced (B in FIG. 6) is a collection of display pixels of the original image data, and is different from the display pixels of the original image data. In the following description, the cells in the reduced image data are also referred to as pixels.

Next, in ST212, the filtering unit 430 performs a filtering step of performing defect detection by applying the defect enhancement filter 431 (FIG. 7) to the reduced image data. The filtering unit 430 stores and stores a 5 × 5 defect enhancement filter 431 shown in FIG. The defect enhancement filter 431 subtracts the average luminance data of the luminance values (luminance data) of the surrounding four pixels A, B, C, and D from the luminance value (luminance data) of the central pixel of interest E to obtain the pixel of interest. This is a filter for the enhanced luminance value of (E) (see FIG. 7B). By applying the defect enhancement filter 431, the luminance difference between the center pixel of interest (E) and the surrounding pixels (A to D) is emphasized, and in the defective pixels with abnormal luminance included in the point defect 710 and the spot defect 720, The enhanced luminance value as the filter output value is a very large value.
An example of the result of applying the defect enhancement filter 431 to the image data is shown in FIG.

Next, in ST213, the binarization unit 440 performs a binarization step of binarizing the enhanced luminance value of the image data (FIG. 8) whose luminance difference is enhanced by the defect enhancement filter 431 with a predetermined threshold.
In the binarizing means 440, a threshold value (for example, 250) is set for distinguishing normal pixels having normal luminance and defective pixels having abnormal luminance based on the emphasized luminance value.
In the binarizing means 440, a binarized array space 441 (see FIG. 9) having the same size as the image data is created and stored. In the binarization step (ST213), the emphasized luminance value of each pixel in the image data (FIG. 8) is compared with the threshold value (250). If the emphasized luminance value is smaller than the threshold value, the binarized array space 441 ( “0” is assigned to the corresponding cell in FIG. 9, and “1” is assigned to the corresponding cell in the binarized array space 441 if the emphasized luminance value is greater than or equal to the threshold value.
In this way, in the binarized array space 441, “0” is arranged in the cell corresponding to the normal pixel, “1” is arranged in the cell corresponding to the defective pixel, and the binarized array shown in FIG. A space 441 is generated.

When a defective pixel located in the point defect 710 or the spot defect 720 is detected in the point / stain defect detection step (ST210), next, the emphasized luminance value of each pixel of the detected display defect is determined for the surrounding pixels. An interpolation process (ST220) is performed by the interpolation means 450 to repair the defect by replacing it with an interpolation value based on the emphasized luminance value.
Hereinafter, each step of the interpolation step (ST220) will be described.

In the interpolation step (ST220), first, a repair target pixel search unit (460) performs a repair target pixel search step (ST221) for sequentially searching for pixels to be repaired.
In the repair target pixel search means 460, matching filters Nos. 1 to 25 shown in FIGS. 10, 11, 12, and 13 are preset and stored.
Here, the matching filter has a size of 5 × 5 and is a filter in which “0” or “1” is arranged in each cell.
The matching filter is composed of four groups. The first group matching filter from No. 1 to No. 8 shown in FIG. 10 and the second group matching filter from No. 9 to No. 16 shown in FIG. The third group matching filter from No. 17 to No. 24 shown in FIG. 12 and the fourth group matching filter of No. 25 shown in FIG. 13 are configured.

The first group matching filter (No. 1 to No. 8, FIG. 10) arranges three “1” s around “1” arranged in the center target cell and has five directions (a, b, c, d, and e) cover two consecutive “0” patterns.
The second group matching filter (No. 9 to No. 16, FIG. 11) arranges two “1” s around “1” arranged in the center target cell and has six directions (a˜) as viewed from the target cell. f) covers a pattern of two consecutive “0” s.
The third group matching filter (No. 17 to No. 24, FIG. 12) arranges one “1” around “1” arranged in the center target cell and has seven directions (a˜) as viewed from the target cell. g) covers the pattern of two consecutive “0” s.
In the 25th matching filter (FIG. 13), which is the fourth group of matching filters, “1” is arranged in the center target cell, and “0” in all eight directions (a to h) as viewed from the target cell. "Is two consecutive.

  10 to 13, the cell in which “*” is arranged may be “0” or “1”.

In the repair target pixel search step (ST221), the first through the 25th matching filters start from, for example, the upper left of the binarized array space 441 (FIG. 9) and go to all the cells to the lower right.
At this time, it is applied to the binarized array space 441 in order from the first matching filter of the first group, and it is determined whether or not “0” and “1” arrays that match the matching filter exist in the binarized array space 441. View. Then, since cells where “0” is arranged around the cell of interest are sequentially found, pixels located on the outer edge among the display defects are found in order.

  For example, when the binarized array space 441 shown in FIG. 9 is applied in order from the first matching filter, as shown in FIG. 14, the array that matches the first matching filter is the binarized array space. As can be seen from FIG. When an array that matches the matching filter is found in this way, the pixel corresponding to the center target cell becomes the repair target pixel.

Interpolation value calculation step of calculating the interpolation value of the luminance at the position of the repair target pixel in ST222 based on the emphasized luminance value of the surrounding pixels in ST222 for the repair target pixel searched in the repair target pixel search step of ST221 Is performed by the interpolation value calculation means 470.
In the interpolation value calculation means 470, an interpolation value calculation formula for calculating the interpolation value of the repair target pixel from the emphasized luminance value of the surrounding pixels is set with the target cell in the center of the matching filter as the repair target pixel.
In addition, since the arrangement patterns of “0” and “1” are different in the matching filters No. 1 to No. 25, the interpolation value calculation formula is set for each of the No. 1 to No. 25 matching patterns.

The interpolation value calculation formula will be briefly described.
Each matching filter is an array in which two “0” s are arranged in a plurality of directions with respect to the central target cell, and the interpolation value of the central target cell (the pixel to be repaired) is the cell in which “0” is arranged. It is calculated from the change rate of the emphasized luminance value of the corresponding pixel.
That is, in the direction in which two “0” s are arranged, the degree of change in the emphasized luminance value from the far side to the near side of the target cell is obtained, and the emphasized luminance value on the near side is added to interpolate the repair target pixel. A value can be obtained.
For example, in the arrangement of Figure 15, if the pixel exceeds the threshold value as emphasized luminance value L _M of the pixel corresponding to the center of the target cell M shown in FIG. 16 is a defective pixel, adjacent to the target cell Then, an interpolation basic value _{IM (AG)} based on the directions of A and G (two directions “a” in FIG. 15) in which two “0” s are arranged is obtained by the following equation.

Interpolated baseline _{I M (AG) = L G} + (L G -L A)

  For example, in the first matching filter, since two “0” s are consecutive in five directions (a direction to e direction) around the target cell, the interpolation value of the pixel (repair target pixel) corresponding to the target cell is used. Takes an average value of the basic interpolation values based on each of these five directions (a direction to e direction).

  Specifically, an example of calculating the interpolation value of the repair target pixel in the numerical example shown in FIG. 17 is as follows.

  In this way, the interpolation value calculation means 470 calculates each interpolation basic value in the direction in which “0” continues for the matching filters from No. 1 to No. 25, and averages the interpolation basic value in each direction. Thus, an interpolation value calculation formula for calculating the interpolation value is set in advance. For each matching filter, the direction in which “0” for calculating the interpolation value continues is shown in FIG. 18, FIG. 19, FIG. 20, and FIG.

In ST221, the binarized array space 441 (FIG. 9) is applied with a matching filter to find a matching portion, and the interpolation value of the pixel (repair target pixel) corresponding to the target cell according to the interpolation value calculation formula of the matched matching filter. After calculating (ST222), the calculated interpolation value is replaced with the enhanced luminance value of the pixel to be repaired in the image data.
In this way, when the emphasized luminance value of the repair target pixel is replaced with the interpolation value, the cell corresponding to the repair target pixel in the binarized array space 441 is rewritten from “1” to “0” and replaced with the interpolation value. A mark is used (ST223).

  In addition, even if the interpolation value of the repair target pixel is calculated based on the emphasized luminance value of two normal pixels around the repair target pixel, this interpolation value may still exceed the threshold value, Even if the calculated interpolation value exceeds the threshold value, the emphasized luminance value of the pixel to be repaired is replaced with the interpolation value, and the target cell in the binarized array space 441 is rewritten to “0”.

In this manner, the matching filter is applied (ST221). When matching is performed, the emphasized luminance value of the pixel to be repaired is replaced with an interpolation value, and the cell of interest in the binarized array space 441 is set to “0” (ST222). , ST223), this process is repeated until all the cells in the binarized array space 441 reach “0”.
Then, the emphasized luminance value of the defective pixel existing in the area of the point defect 710 / stain defect 720 is sequentially replaced with the interpolated value. For example, as shown in FIG. 22, in the point / stain defect detection step (ST210). When the spot defect 720 has been detected, the highlight luminance value of the defective pixel existing in the spot defect 720 is sequentially replaced with an interpolated value based on the surrounding highlight brightness value, and the spot defect that has protruded from the surroundings before repairing. The enhanced luminance value 720 is smoothly corrected according to the surroundings. Note that FIG. 22 shows a case where the spot defect 720 overlaps the unevenness defect 730.

Then, when all the cells in the binarized array space 441 become “0” (ST224: YES), repair of defective pixels existing in the area corresponding to the point defect 710 and the spot defect 720 is performed next. It is determined whether it is completed (ST230). That is,
The area of the defect emphasized (detected) by the defect enhancement filter 431 (FIG. 7) in the filtering step (ST212) varies depending on the size of the image data reduced in the image reduction step (ST211). It is determined whether or not defects (point defects 710 and spot defects 720) in an area smaller than the uneven defect 730 to be repaired are repaired.

  Specifically, when the image reduction unit 420 in the image reduction process ST211 has 1/2, 1/4, and 1/8 stages as image reduction ratios, 1/2, 1/4, and 1/8. It is confirmed whether or not the image reduction at the stage has been performed. If the image reduction at all the stages 1/2, 1/4, and 1/8 has not been performed (ST230: NO), the process returns to the image reduction process at ST211. Then, the processing of ST211 to ST230 is repeated.

  Here, when returning to ST211 and repeating the processing, the image data that has already repaired the point defect 710 at a reduction ratio of 1/2 is further reduced to 1/2 to generate a reduced image of 1/4 size of the original size. (Image reduction step ST211), and the steps subsequent to the filtering step ST212 (ST212 to ST260) are performed on the 1 / 4-size image data. Then, the point defect 710 is repaired in the 1/2 size reduced image data, and the spot defect 720 is repaired in the 1/4 size reduced image data. In this way, by repairing the point defect 710 and the spot defect 720 while reducing the image data to a predetermined size, only the unevenness defect 730 that is the target of rank determination remains.

When the repair of the defective pixel existing in the area corresponding to the point defect 710 and the spot defect 720 is completed (ST230: YES), the mura defect detection step of detecting the mura defect 730 in ST300 is performed by the mura defect detection means 500. Is called.
In the mura defect detection means 500, a 7 × 7 defect enhancement filter shown in FIG. 23 is set and stored as the mura defect detection filter 510.

The unevenness defect detection filter 510 is a filter that subtracts the average of the enhancement luminance values of the surrounding ten pixels from the enhancement luminance value of the central nine pixels to obtain the enhancement luminance value of the target pixel.
When this mura defect detection filter 510 is applied to the image data in which the point defect 710 / stain defect 720 has been repaired in the point / stain defect repair process ST200, the defect of the mura defect size is emphasized.
Then, the enhanced luminance value obtained by the mura defect enhancement filter is compared with a predetermined threshold value to detect defective pixels that become the mura defect 730. If there are a predetermined number or more of defective pixels detected in this way, the region is detected as a mura defect 730 (ST300).

  Here, a defect detection method is constituted by the mura defect detection of ST300 from the pretreatment process of ST100.

When the mura defect 730 is detected in the mura defect detection step (ST300), a rank determination unit 600 performs a rank determination step based on a quantitative evaluation value of the mura defect region in ST400. .
In the rank determination means 600, a determination threshold curve is set and stored as shown in FIG.
In the rank determination step (ST400), the defect area and the defect contrast are calculated as a quantitative evaluation value of the mura defect 730 detected in the mura defect detection step (ST300), and the calculated defect area and defect contrast are calculated. Plotting on the graph of the threshold curve determines whether the liquid crystal light valve 112 to be inspected is good or bad, and further, if it is in the non-defective range, it is determined which rank from A rank to D rank corresponds.

According to 1st Embodiment provided with such a structure, there can exist the following effects.
(1) When it is desired to evaluate the liquid crystal light valve 112 based on the mura defect 730, a display defect (point defect 710 / stain defect 720) having an area smaller than that of the mura defect 730 is present in a state where it overlaps with the mura defect 730. However, the point defect 710 and the spot defect 720 are repaired in advance in the point / stain defect repairing step (ST200), and the unevenness defect 730 is detected from the image data in which the point defect 710 / stain defect 720 is repaired. Therefore, even if the point defect 710 / stain defect 720 and the mura defect 730 initially overlap in the image data, the already repaired point defect 710 / stain defect 720 is not detected, and only the mura defect 730 is detected. Is properly detected. As a result, the liquid crystal light valve 112 can be appropriately evaluated based on the unevenness defect 730.

(2) In the point / spot defect repairing step (ST200), the brightness of each pixel of the point defect 710 / stain defect 720 is replaced with an interpolated value obtained by extending the surrounding brightness change. When the 710 and spot defect 720 overlap, the average luminance calculated based only on the uneven defect 730 portion and the point defect 710 and spot defect 720 repaired after the point defect 710 and spot defect 720 are repaired. It is assumed that the average brightness of the entire area of the mura defect 730 including this area is an approximate value, and further, the mura defect 730 continues in that area instead of the point defect 710 and the spot defect 720. The average brightness can be obtained.
Therefore, when the uneven defect 730 is detected from the image data after the point defect 710 and the spot defect 720 are repaired and the average luminance of the uneven defect 730 is simply calculated as a quantitative evaluation value of the uneven defect 730, A value corresponding to the average luminance recognized by the eyes can be obtained. Therefore, in the rank determination step (ST400), an appropriate evaluation based on the mura defect 730 can be performed, and an evaluation that approximates the human impression can be obtained.

(3) In repairing the point defect 710 and the spot defect 720, the luminance data (emphasized brightness value) of each pixel of the point defect 710 and the spot defect 720 is replaced with an interpolation value, so that the area changes before and after the repair. The area of the mura defect 730 does not change even if the point defect 710 and the spot defect 720 are repaired at the portion where the mura defect 730 and the point defect 710 and the spot defect 720 overlap. Therefore, the mura defect 730 is detected from the image data after repairing the point defect 710 / stain defect 720 in the point / stain defect repairing step (ST200) (unevenness defect detecting step ST300), and the detected mura defect 730 is detected. When the defect area is calculated as a quantitative evaluation value, the defect area does not change before and after the repair of the point defect 710 and the spot defect 720. Therefore, if the mura defect 730 is detected in the image data after repairing the point defect 710 and the spot defect 720 and the area of the detected mura defect 730 is simply calculated, the defect area recognized by the human eye can be obtained. The corresponding value can be obtained. As a result, in the rank determination step (ST400), an appropriate evaluation based only on the mura defect 730 can be performed, and an evaluation approximating the impression of human appearance can be obtained.

(4) In calculating the interpolation value in the interpolation step (ST220), the interpolation value is calculated in order from the outside to the inside of the detected point defect 710 / stain defect 720, thereby the outside of the point defect 710 / stain defect 720. Interpolation values reflecting changes in luminance data of normal pixels can be calculated in order from the outside to the inside, and the luminance data (emphasized luminance value) of each pixel of the point defect 710 and the spot defect 720 is replaced with the interpolation value thus calculated. The point defect 710 and the spot defect 720 can be repaired. The interpolated value of the pixel located inside the point defect 710 / stain defect 720 can also be calculated using the interpolation value already calculated for the pixel located outside the point defect 710 / stain defect 720. Interpolation values can be calculated and replaced for all 720 pixels.

(5) In the point / stain defect detection step (ST210), the image data is reduced to a predetermined size in the image reduction step (ST211), and then a defect enhancement filter for defect detection is applied to the reduced image data. The point defect 710 and spot defect 720 are detected over time, and the image data is reduced in several stages and a defect enhancement filter is applied to the reduced image data to thereby apply point defects 710 of various sizes in the image data. All the spot defects 720 can be detected.

(6) When a display defect of a predetermined size is detected by applying a defect enhancement filter to the image data reduced to a predetermined size, and the detected display defect is repaired in the interpolation step (ST220), at this stage, Where display defects (point defects 710 and spot defects 720) that do not apply to the defect enhancement filter at this reduction rate are not repaired, image reduction is further performed on the image data with the display defects repaired at a predetermined reduction rate. When a defect enhancement filter is applied to the reduced image data, display defects that have not been repaired are applied to the defect detection filter, so that the detected display defects can be repaired in an interpolation process. In this way, the image reduction process, the filtering process, and the interpolation process are repeated, and the reduction rate of the image data is set so that all display defects of the size of the point defect 710 and the spot defect 720 are applied to the defect detection filter. All of the spot defects 720 can be repaired.

(7) The matching filter is applied to the binarized array space 441 to search for the pixel to be repaired. Further, since the direction in which 0 continues is known from the array of the matching filter, the brightness data of the pixel in the direction in which 0 continues is obtained. Based on this, the interpolation value of the pixel to be repaired can be calculated. Since the matching filter in which the array of 1 and 0 is determined is applied to the binarized array space 441, the array direction of 0 immediately turns around at the matched location, and interpolation of the pixel to be repaired is easily and immediately performed by the interpolation value calculation formula A value can be calculated. Then, the pixel to be repaired can be repaired by replacing the luminance data of the pixel to be repaired with the calculated interpolation value.

(8) Since the pixel corrected by the interpolation value is replaced with “0” in the binarized array space 441, initially, the periphery is surrounded by defective pixels and there are a predetermined number or more of normal pixels. For the defective pixels that did not exist, the pixels around the defective pixels are corrected with the interpolated values and replaced with “0” in the binarized array space 441. Then, 0s are gradually arranged around, and finally, the search with the matching filter starts. Therefore, all the defective pixels initially detected are sequentially searched by the matching filter. As a result, all the pixels of the point defect 710 and the spot defect 720 in the image data can be repaired with the interpolation value.

It should be noted that the present invention is not limited to the above-described embodiment, and modifications, improvements, and the like within the scope that can achieve the object of the present invention are included in the present invention.
The inspection object may be not only a display image by an optical panel such as a liquid crystal light valve (liquid crystal panel), but also a coating unevenness of a sheet metal or a printing paper surface. Of course, the optical panel is not limited to the liquid crystal panel, and may be an organic EL panel or the like.

  The evaluation target defect type that is the target of rank determination in the rank determination step (ST400) is the mura defect 730, and the non-evaluation target defect type that is not the target of rank determination and that is to be repaired in the present invention is the point defect 710 and Although the case where the spot defect 720 is used has been described, for example, when the spot defect 720 is the target of rank determination, the defect type to be evaluated is the spot defect 720, and the non-evaluation target defect type to be repaired is a point defect. It may be 710. That is, when a defect type having a predetermined area or more is set as the evaluation target defect type and a defect less than the predetermined area is set as the non-evaluation target defect type, the “predetermined area” can be changed according to the purpose.

  In detecting the dot / spot defect of various sizes in the point / stain defect detection step (ST210), the defect enhancement filter 431 has one size, while the image reduction process (ST211) reduces the image data. The example of detecting various size points and spot defects by using multiple reduction ratios has been explained. However, the defect enhancement filter size can be detected using various size points and spot defects without image reduction. You may prepare it in multiple sizes so that you can. That is, first, a point defect is detected by applying a defect enhancement filter of a small size to the image data, the detected point defect is repaired by an interpolation process, and the size of the next stage is added to the image data in which the point defect is repaired. The defect enhancement filter may be applied to detect a spot defect, and the spot defect may be repaired by an interpolation process.

  In the case of replacing the point defect / stain defect pixel with the interpolated value, the case where the repair target pixel is searched by matching using the matching filter (FIGS. 10 to 13) has been described as an example. As long as the pixels near the outer edge of the point defect / stain defect can be found sequentially, the method of searching for the repair target pixel is not particularly limited.

In the interpolation value calculation step ST222, the basic interpolation value is calculated in each direction based on the change rate of the luminance data (emphasized luminance value) of the pixel in the direction where “0” continues around the pixel to be repaired. Although the example in which the average of the values is used as the interpolation value of the repair target pixel has been described, the method of calculating the interpolation value may be variously changed depending on the purpose.
For example, the maximum value may be selected from the interpolation basic values calculated in each direction to be an interpolation value, the minimum value may be selected to be an interpolation value, or an intermediate value may be selected to be an interpolation value. Good.

  When the interpolation value is calculated for the pixel to be repaired, it is also effective if the interpolation value exceeds the threshold value, and the luminance data (emphasized luminance value) of the pixel to be repaired is replaced with this interpolation value and the binarized array The case of rewriting from 1 to 0 in the space has been described as an example. For example, when the calculated interpolation value exceeds the threshold value, the interpolation luminance value of the pixel to be repaired is replaced with this interpolation value while binary. It may be left as 1 in the chemical array space. In this way, if the interpolated value is equal to or greater than the threshold value, if it is left as 1 in the binarized array space, the interpolated value equal to or greater than the threshold value is not used as the basis for the interpolated value calculation in the interpolated value calculation for other pixels. This is also because the search is performed again as a pixel to be repaired by the search using the matching filter and the interpolation value is calculated again. In this case, all cells in the binarized array space are not necessarily 0, but the 1 and 0 arrays in the binarized array space match any of the matching filters (FIGS. 10 to 13). When no more, the process returns to the image reduction step ST211 and the processing (ST211 to ST230) may be repeated.

  Also, in the binarized array space, 1 is assigned to a pixel with an emphasized luminance value greater than or equal to the threshold value, and 0 is assigned to a pixel with an emphasized luminance value less than the threshold value. is there.

  In the image reduction process (ST211), the case where four pixels are combined into one pixel has been described as an example (FIG. 6). For example, six pixels are combined into one as shown in FIG. Of course, the reduction ratio of one image in the image reduction process is not particularly limited.

  In the evaluation process, the defect area and the average luminance of the detected mura defect are calculated, and the case where the determination of good / bad and the ranking from A to D is performed based on these values is described as an example. The method of evaluation is not particularly limited, and may be changed variously according to the purpose.

  The present invention can be applied to a defect detection method for detecting a display defect in an image displayed on an optical panel, and can also be applied to, for example, detection of uneven coating on a coated surface of a sheet metal or a printing paper surface.

The figure which shows the whole structure of the optical panel inspection apparatus in 1st Embodiment. The figure which shows the structure of the arithmetic processing part in 1st Embodiment. 5 is a flowchart illustrating a processing procedure of an optical panel inspection method in the first embodiment. The flowchart which shows the process sequence of a pre-processing process in 1st Embodiment. The flowchart which shows the process sequence of a point and a spot defect repair process in 1st Embodiment. The figure which shows the original image data (A) before image reduction and the image data (B) after reduction in 1st Embodiment. It is a figure which shows a defect emphasis filter in 1st Embodiment. The figure which shows an example of the result of having applied the defect emphasis filter to image data in 1st Embodiment. The figure which shows the example of the binarization arrangement | sequence space in 1st Embodiment. The figure which shows the 1st group of matching filters in 1st Embodiment. The figure which shows the 2nd group of matching filters in 1st Embodiment. The figure which shows the 3rd group of matching filters in 1st Embodiment. The figure which shows the 4th group matching filter in 1st Embodiment. The figure which shows a mode that the arrangement | sequence matched with the 1st matching filter was searched in the binarization arrangement | sequence space in 1st Embodiment. The figure for demonstrating the arrangement | sequence of a matching filter in 1st Embodiment. The figure which shows the interpolation basic value together with the graph of a change of luminance data in 1st Embodiment. The figure which shows the numerical example which calculates the interpolation value of a repair object pixel in 1st Embodiment. The figure which shows the direction where "0" continues in each matching filter in 1st Embodiment. The figure which shows the direction where "0" continues in each matching filter in 1st Embodiment. The figure which shows the direction where "0" continues in each matching filter in 1st Embodiment. The figure which shows the direction where "0" continues in each matching filter in 1st Embodiment. The figure which shows a mode that a spot defect is repaired in 1st Embodiment. The figure which shows the nonuniformity defect detection filter in 1st Embodiment. The figure which shows the modification in the case of image reduction. The figure which shows the example of a point defect, a spot defect, and a nonuniformity defect. The figure which shows an example of the threshold curve set from the relationship between the area of a defect, and contrast. The figure which shows the example which a spot defect and a point defect overlap with a nonuniformity defect. The conceptual diagram which shows a mode that the spot defect which exists in a nonuniformity defect was removed. The conceptual diagram which shows a mode that the contrast of the stain defect which exists in a nonuniformity defect was substituted with the whole average contrast.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Optical panel inspection apparatus, 110 ... Image projection means, 111 ... Light source, 112 ... Liquid crystal light valve, 113 ... Projection lens, 114 ... Mirror, 120 ... Pattern generator, 130 ... Projection part, 131 ... Dark box, 132 ... Opening , 133 ... opening, 134 ... screen, 140 ... CCD camera, 200 ... arithmetic processing part, 300 ... pre-processing means, 310 ... inspection image acquisition means, 320 ... background difference means, 330 ... area extraction means, 340 ... luminance Value calculation means, 400... Point / stain defect repair means, 410... Point / stain defect detection means, 420... Image reduction means, 430 ... filtering means, 431... Defect emphasis filter, 440 ... binarization means, 441. ,... Interpolation means, 460... Repair target pixel search means, 470... Interpolation value calculation means, 500. Defect detection means, 510... Mura defect detection filter, 600. Rank determination means, 610 ... threshold curve graph, 710 ... point defect, 711 ... bright spot defect, 720 ... spot defect, 721 ... white spot defect, 722 ... black spot Defect, 730 ... unevenness defect, 731 ... white unevenness defect, 732 ... black unevenness defect.

Claims (13)

An inspection image acquisition step of imaging the inspection object and acquiring the image data;
In the image data acquired in the inspection image acquisition step, the luminance data of each pixel of the non-evaluation target defect type which is a defect type whose area is less than a predetermined size is replaced with an interpolation value based on the luminance data of surrounding pixels. Non-evaluation target defect type repair process to repair non-evaluation target defect type,
An evaluation target defect type detection step of detecting an evaluation target defect type that is a defect type having an area of a predetermined size or more in the image data in which the non-evaluation target defect type is repaired in the non-evaluation target defect type repair step. A defect detection method characterized by the above. The defect detection method according to claim 1,
The non-evaluation target defect type repair process,
The interpolation value is calculated based on a tendency that luminance data of a pixel satisfying a predetermined condition around each pixel of the non-evaluation target defect type changes toward each pixel of the non-evaluation target defect type. Defect detection method. In the defect detection method according to claim 1 or 2,
The non-evaluated defect type repair process is:
A non-evaluation target defect type detection step for detecting the non-evaluation target defect type in the image data acquired in the image data acquisition step;
An interpolation step of calculating the interpolation value for each pixel of the non-evaluation target defect type and replacing luminance data of each pixel with the interpolation value, and
The interpolation step includes
Starting from the pixel located at the outer edge of the detected non-evaluation target defect type, sequentially calculate the interpolation value of the pixels closer to the inside,
In determining the interpolation value of the pixel located at the outer edge, the interpolation value is calculated based on luminance data of a normal pixel adjacent to the pixel and having normal luminance,
In calculating the interpolation value of the pixel closer to the inner side, the pixel for which the interpolation value is to be calculated out of the luminance data of the normal pixel and the interpolation value already calculated as the interpolation value of the pixel closer to the outer edge A defect detection method characterized in that the interpolated value is calculated on the basis of what is adjacent to the object. The defect detection method according to claim 3,
The non-evaluation target defect type detection step includes:
An image reduction process for reducing the image data to be inspected;
A filtering step of detecting a display defect by applying a defect enhancement filter that emphasizes a difference in luminance data between the pixel of interest and pixels around the pixel of interest to the image data reduced in the image reduction step, and
The non-evaluation target defect type repair process,
After correcting the non-evaluation target defect type by calculating the interpolation value of the pixel of the non-evaluation target defect type detected in the filtering step for the image data reduced to a predetermined size in the image reduction step Further, the defect detection is characterized in that the image reduction process, the filtering process and the interpolation process are repeated for the repaired image data to repair the non-evaluation target defect types of the repair target area sequentially. Method. The defect detection method according to claim 4,
The defect type to be evaluated is a mura defect,
The non-evaluation target defect type is a spot defect having a smaller area than the uneven defect and a point defect having a smaller area than the spot defect,
The non-evaluation target defect type repair process,
After detecting the point defect in the filtering step for the image data reduced in the first image reduction step and calculating the interpolation value of the point defect in the interpolation step,
Further, after reducing the image data in which the point defect is corrected in the image reduction process, the spot defect is detected in the filtering process, and an interpolation value of the spot defect is calculated in the interpolation process. A defect detection method characterized by repairing a defect. In the defect detection method according to any one of claims 3 to 5,
The non-evaluation target defect type detection step compares the luminance data of each pixel of the image data with a predetermined threshold value to detect defective pixels whose luminance data is equal to or higher than the threshold value and normal pixels whose luminance data is lower than the threshold value. Including a binarization step for generating a binarized array space in which a defective pixel is 1 and a normal pixel is 0 and binarized;
The pre-interpolation process is
A repair target pixel search step for sequentially searching for a pixel to be repaired among defective pixels of the non-evaluation target defect type detected in the non-evaluation target defect type detection step;
An interpolation value calculation step of calculating an interpolation value of a pixel to be repaired searched in the repair target pixel search step,
A plurality of matching filters covering a pattern in which 1 is arranged in a central cell of interest in a matrix arrangement of a predetermined size and two or more 0s are continuously arranged in at least 5 or more of the surrounding 8 directions. Prepared,
The repair target pixel search step applies the matching filter to the binarized array space in order and searches the binarized array space for an array location of 1 and 0 that matches one of the matching filters, and performs matching. Recognize in the image data the pixel corresponding to the target cell of the matching filter at the array location that matches the filter as the pixel to be repaired,
The interpolation value calculation step calculates an interpolation value of the repair target pixel based on a change rate of luminance data of the pixel in a direction in which 0 or more continues in the matching filter.
For the pixel in which the interpolation value is calculated in this interpolation value calculation step and the luminance data is replaced with the interpolation value, the corresponding cell is rewritten from 1 to 0 in the binarization array space, and the binarization array space A defect detection method comprising repeating the repair target pixel search step and the interpolation value calculation step until all the cells are zero. The defect detection method according to claim 6,
The defect detection method characterized in that the repair target pixel search step sequentially goes to the binarized array space from a matching filter having a small direction in which two or more zeros continue around the target cell. In the defect detection method according to claim 6 or 7,
The interpolation value calculation step includes
The luminance data of the outer pixel is subtracted from twice the luminance data of the pixel adjacent to the pixel to be repaired in each direction in which two or more 0s are adjacent to the pixel to be repaired in the binarized array space. An average value obtained by averaging the basic interpolation values calculated in each direction after calculating the basic interpolation value obtained in this manner is used as the interpolation value of the repair target pixel. A defect detection step of detecting a defect type to be evaluated in an inspection object by the defect detection method according to claim 1;
An evaluation step of evaluating the inspection object based on a quantitative evaluation value of the defect type to be evaluated detected in the defect detection step. Inspection image acquisition means for imaging the inspection object and acquiring the image data;
In the image data acquired in the inspection image acquisition step, the luminance data of each pixel of the non-evaluation target defect type which is a defect type whose area is less than a predetermined size is replaced with an interpolation value based on the luminance data of surrounding pixels. Non-evaluation target defect type repairing means for repairing non-evaluation target defect types,
Evaluation target defect type detection means for detecting an evaluation target defect type whose area is a defect type having a predetermined size or more in the image data in which the non-evaluation target defect type repairing means is repaired by the non-evaluation target defect type repairing means. The defect detection apparatus characterized by the above-mentioned. The defect detection apparatus according to claim 10;
An evaluation unit that evaluates the inspection object based on a quantitative evaluation value of the defect type to be evaluated detected by the defect detection apparatus. Incorporating a computer into a defect detection apparatus for detecting a defect type to be evaluated, which is a defect type having a predetermined area or more in the inspection target,
Inspection image acquisition means for imaging the inspection object and acquiring the image data;
In the image data acquired in the inspection image acquisition step, the luminance data of each pixel of the non-evaluation target defect type which is a defect type whose area is less than a predetermined size is replaced with an interpolation value based on the luminance data of surrounding pixels. Non-evaluation target defect type repairing means for repairing non-evaluation target defect types,
As an evaluation target defect type detecting means for detecting an evaluation target defect type whose area is a defect type having a predetermined size or more in the image data in which the non-evaluation target defect type repairing means is repaired by the non-evaluation target defect type repairing means, A computer-readable defect detection program characterized by being made to function.   A recording medium in which the defect detection program according to claim 12 is recorded so as to be readable by a computer.

Images