JP2018106456A - Image processing device and image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device and an image processing method capable of detecting defects with high accuracy even in the case that the defects having different contrasts and sizes in a photographic image to be inspected exists in the photographic image.SOLUTION: The image processing device and the image processing method include: an initial stage step image processing process including a first defect component detection image generation step (St2) for generating a defect component detection image about a photographic image, and a first defect component removal image generation step (St3) for generating a defect component removal image about the photographic image; a post stage step image processing process including a second defect component detection image generation step (St4) for generating a defect component detection image about the defect component removal image, and a second defect component removal image generation step (St5) for generating a new defect component removal image obtained by removing a defect component from the defect component removal image; and a defect extraction process for generating a defect extraction image from the photographic image or the defect component removal image obtained in each step.SELECTED DRAWING: Figure 2

Description

  The present invention processes an image obtained by imaging an object to be inspected, so that a defect having a luminance difference (contrast) or a size different from that of a background such as a scratch, a foreign object, a stain, or unevenness of the object to be inspected is photographed. The present invention relates to an image processing apparatus and an image processing method capable of accurately detecting these defects even when they are mixed in an image.

  In the case of detecting a foreign matter or a flaw defect on the surface of the inspection object, automation is achieved by performing image processing on a captured image of the inspection object. For example, the technique disclosed in Patent Document 1 calculates a background pixel value (pixel luminance value) at a plurality of pixel positions in the vicinity of a predetermined pixel (target pixel) in a captured image of an inspection object. By performing statistical processing on each of the estimated background pixel values, a background image is generated by selecting a background pixel estimated value for each pixel of interest, and a defect is obtained by taking the difference between the captured image and the background pixel. Is extracted. Specifically, a plurality of pixels (tap) are set around the pixel of interest at a predetermined interval in a one-dimensional array or a two-dimensional array, and the luminance values of the set pixels are quadratic curve approximated or interpolated. Thus, the luminance change in these pixel ranges is approximated. Then, based on the obtained approximate curve, a value at the inspection target pixel position is estimated as a background pixel value, and a background image is generated. In this prior art, the process from the selection of taps around the pixel of interest to the estimation of the background pixel is the filtering process. In order to improve accuracy in this filtering process, a plurality of filtering processes with different tap intervals are applied. Then, for each filtering process, the difference between the pixel used for the approximation and the approximate curve is obtained, and the sum of squares of the differences or the sum of absolute values is obtained as the difference variance evaluation value, and the evaluation value is the minimum. Is selected to estimate the background pixel value.

JP 2014-135007 A

  As described above, in the filtering process according to the conventional technique, an error is detected between the estimated background pixel value and the actual background pixel value in the vicinity of the boundary between the defect and the background within the range of pixels to be approximated. This makes it difficult to detect defects accurately. In order to perform accurate detection, it is necessary to set the tap so as to include all the defects within the range between the target pixel and not to enter the defects. However, it is difficult to grasp what kind of defect is present in the image when setting the tap. For this reason, the conventional technology has a limit in the accuracy of defect detection. In particular, it has been difficult to accurately detect a low-contrast defect with a relatively low difference in luminance value from the background.

  The present invention has been made in view of such circumstances, and an object of the present invention is to accurately detect defects having different contrasts and sizes in a captured image of an inspection object even in the captured image. An object is to provide an image processing apparatus and an image processing method capable of detection.

The image processing apparatus of the present invention has been proposed in order to achieve the above object, and a first defect component detection image creation step for creating a defect component detection image for a photographed image obtained by imaging an inspection object. And first-stage image processing means for executing a first-stage step image processing step including a first defect component-removed image creating step for creating a defect component-removed image obtained by removing a defect component from the photographed image;
A second defect component detection image creation step for creating a defect component detection image for the defect component removal image, and a second defect component removal image creation for creating a new defect component removal image by removing the defect component from the defect component removal image A post-stage image processing means for executing a post-stage image processing process including the process;
Defect extraction means for creating a defect extraction image based on a difference in luminance value between arbitrary images of the captured image or the defect component removal image obtained in each step;
With
The first stage image processing means includes:
In the first defect component detection image creation step,
By setting the pixel distance indicating the number of pixels between the inspection target pixel and the comparison pixel to be compared with the inspection target pixel to the first distance,
For the captured image, a defect component detection image is created by detecting a defect component having a size corresponding to the first distance for each pixel constituting the image,
In the first defect component removal image creation step,
Create a defect component removal image by adding and subtracting a luminance value between the captured image and the defect component detection image created in the first defect component detection image creation step,
The latter stage image processing means includes:
In the second defect component detection image creation step,
By setting the pixel distance to a second distance that is longer than the pixel distance in the previous step,
For a defect component removal image having a pixel distance equal to or less than the pixel distance in the current step, a defect component corresponding to the second distance is detected for each pixel constituting the defect component detection image,
In the second defect component removal image creation step,
Creating a new defect component removal image by adding and subtracting luminance values of the defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step and the defect component detection image created in the second defect component detection image creation step;
It is characterized by that.

  According to the present invention, the defect component detection image creation step and the defect component removal image creation step are repeated while increasing the pixel distance for each step, and the defect components of the respective defect components are detected and removed in order from the smallest defect component. A component detection image and a defect component removal image are obtained. And since a defect is extracted by the difference between these defect component removal images, it is possible to extract a defect from a captured image of an inspection object with higher accuracy without using smoothing or an approximate expression as in the prior art. Is possible. In particular, even when a high-contrast defect and a low defect are mixed in the captured image, detection can be performed while reducing false detection. In addition, since defects of any size can be extracted using defect component-removed images at different pixel distances, the degree of freedom in defect detection is improved.

In the above configuration, the first-stage image processing means sets a plurality of comparison pixels at the first distance with respect to the inspection target pixel for the photographed image in the first defect component detection image creation step, and Create a defect component detection image based on the difference between the luminance value and the luminance value of the pixel to be inspected,
In the second defect component detection image creating step, the subsequent image processing means sets a plurality of comparison pixels at the second distance with respect to the inspection target pixel for a defect component removal image whose pixel distance is equal to or smaller than the pixel distance in the current step. It is desirable to adopt a configuration in which a defect component detection image is created based on the difference between the luminance value of each comparison pixel and the luminance value of the inspection target pixel.

  According to this configuration, for example, a relatively large size defect such as unevenness or a spot in the captured image includes a relatively small size defect such as a scratch, or Even when there are edges (parts where the contrast changes abruptly) such as structures (protrusions and recesses) of the object being inspected, it is detected and removed sequentially from the defect components with small sizes, thus suppressing false detection Is done.

  Further, in the above configuration, the first-stage image processing means and the second-stage image processing means respectively set the size of the defect to be detected in the first defect component detection image creation step and the second defect component detection image creation step, respectively. It is desirable to set the pixel distance accordingly.

  According to this configuration, it is possible to detect a defect of a target size in a captured image.

  In the above configuration, it is preferable that the post-stage image processing means adopts a configuration in which the post-step image processing process is executed in a plurality of steps by sequentially increasing the pixel distance for each step.

  According to this configuration, it is possible to further increase the detection accuracy of the defect component by sequentially increasing the pixel distance for each step and executing the subsequent step image processing step a plurality of steps.

The image processing method of the present invention also includes a first defect component detection image creating step for creating a defect component detection image for a photographed image obtained by imaging the inspection object, and removing the defect component from the photographed image. A first step image processing step including a first defect component removal image creation step of creating a defect component removal image;
A second defect component detection image creation step for creating a defect component detection image for the defect component removal image, and a second defect component removal image creation for creating a new defect component removal image by removing the defect component from the defect component removal image A post-step image processing process including a process;
A defect extraction step of creating a defect extraction image by a difference in luminance value between arbitrary images of the captured image or the defect component removal image obtained in each step;
Including
In the first defect component detection image creation step,
By setting the pixel distance indicating the number of pixels between the inspection target pixel and the comparison pixel to be compared with the inspection target pixel to the first distance,
For the captured image, a defect component detection image is created by detecting a defect component having a size corresponding to the first distance for each pixel constituting the image,
In the first defect component removal image creation step,
Create a defect component removal image by adding and subtracting a luminance value between the captured image and the defect component detection image created in the first defect component detection image creation step,
In the second defect component detection image creation step,
By setting the pixel distance indicating the number of pixels between the inspection target pixel and the comparison pixel to be compared with the inspection target pixel to a second distance longer than the pixel distance in the previous step,
For a defect component removal image having a pixel distance equal to or less than the pixel distance in the current step, a defect component corresponding to the second distance is detected for each pixel constituting the defect component detection image,
In the second defect component removal image creation step,
Creating a new defect component removal image by adding and subtracting the defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step and the defect component detection image created in the second defect component detection image creation step;
It is characterized by that.

  According to the present invention, the defect component detection image creation step and the defect component removal image creation step are repeated while increasing the pixel distance for each step, and the defect components of the respective defect components are detected and removed in order from the smallest defect component. A component detection image and a defect component removal image are obtained. And since a defect is extracted by the difference between these defect component removal images, it is possible to extract a defect from a captured image of an inspection object with higher accuracy without using smoothing or an approximate expression as in the prior art. Is possible. In particular, even when a defect having a relatively high contrast and a defect having a relatively low contrast are mixed in the captured image, detection can be performed while reducing false detection. In addition, since defects of any size can be extracted using defect component-removed images at different pixel distances, the degree of freedom in defect detection is improved.

In the above method, in the first defect component detection image creating step, a plurality of comparison pixels are set at the first distance with respect to the inspection target pixel in the captured image, and the luminance value of each comparison pixel and the luminance of the inspection target pixel are set. Create a defect component detection image based on the difference from the value,
In the second defect component detection image creation step, a plurality of comparison pixels are set at the second distance with respect to a pixel to be inspected for a defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step, and the luminance value of each comparison pixel It is desirable to create a defect component detection image based on the difference between the brightness value of the pixel to be inspected and the inspection target pixel.

  According to this, for example, a relatively large size defect such as unevenness or a stain in the captured image includes a relatively small size defect such as a scratch, or in the captured image. Even when there is an edge such as a structure of the object to be inspected, since the defect components having small sizes are sequentially detected and removed, erroneous detection is suppressed.

  In the first defect component detection image creation step and the second defect component detection image creation step, it is preferable that the pixel distance is set according to the size of a defect to be detected.

  According to this, it becomes possible to detect a defect of a target size in a captured image.

  Furthermore, it is preferable that the latter step image processing step is executed in a plurality of steps by sequentially increasing the pixel distance for each step.

  According to this, it is possible to further increase the detection accuracy of the defect component by sequentially increasing the pixel distance for each step and executing the subsequent step image processing step a plurality of steps.

It is a block diagram explaining the defect detection apparatus which is one form of the image processing apparatus of this invention. It is a flowchart explaining operation | movement of a defect detection apparatus. An example of a processing target image (captured image) is shown. It is a figure which shows an example of arrangement | positioning of a test object pixel and a comparison pixel in a defect component detection image creation process. It is a figure explaining the 1st comparison pixel group used at the defective component detection image creation process in a modification. It is a figure explaining the 2nd comparison pixel group used at the defective component detection image creation process in a modification. It is process drawing explaining the defect component detection image creation process in a modification. It is process drawing explaining the defect component detection image creation process in a modification. It is process drawing explaining the defect component detection image creation process in a modification. It is process drawing explaining the defect component detection image creation process in a modification. It is process drawing explaining the defect component detection image creation process in a modification. It is process drawing explaining the defect component detection image creation process in a modification. It is a figure which shows an example of the step 1 bright defect component detection image. It is a figure which shows an example of a step 1 dark defect component detection image. It is a figure which shows an example of a step 1 defect component removal image. It is a figure which shows an example of the step 2 bright defect component detection image. It is a figure which shows an example of the step 2 dark defect component detection image. It is a figure which shows an example of a step 2 defect component removal image. It is a figure which shows an example of the step 12 bright defect component detection image. It is a figure which shows an example of the step 12 dark defect component detection image. It is a figure which shows an example of the step 12 defect component removal image. It is a figure which shows an example of the step 100 bright defect component detection image. It is a figure which shows an example of the step 100 dark defect component detection image. It is a figure which shows an example of the step 100 defect component removal image. It is a figure which shows an example of the bright defect extraction image from step 2 to step 12. It is a figure which shows an example of the dark defect extraction image from step 2 to step 12. It is a figure which shows an example of the bright defect extraction image from step 1 to step 100. FIG. It is a figure which shows an example of the dark defect extraction image from step 1 to step 100. FIG. It is a figure which shows an example of the bright defect extraction image from step 13 to step 100. FIG. It is a figure which shows an example of the dark defect extraction image from step 13 to step 100. FIG. It is a figure which shows an example of the bright defect extraction image at the time of subtracting the defect component from Step 1 to Step 12 from the defect extraction image from Step 13 to Step 100. It is a figure which shows an example of the dark defect extraction image at the time of subtracting the defect component from Step 1 to Step 12 from the defect extraction image from Step 13 to Step 100. It is a figure which shows the pattern image as a process target image in 2nd Embodiment. It is a figure which shows an example of the bright defect extraction image from step 1 to step 9 in 2nd Embodiment. It is a figure which shows an example of the dark defect extraction image from step 1 to step 9 in 2nd Embodiment. It is a figure which shows an example of the bright defect extraction image from Step 1 to Step 65 in 2nd Embodiment. It is a figure which shows an example of the dark defect extraction image from step 1 to step 65 in 2nd Embodiment.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention will be described with reference to the accompanying drawings. In the embodiments described below, various limitations are made as preferred specific examples of the present invention. However, the scope of the present invention is not limited to the following description unless otherwise specified. However, the present invention is not limited to these embodiments.

  FIG. 1 is a block diagram showing a configuration of a defect detection apparatus 10 which is an embodiment of an image processing apparatus according to the present invention. The defect detection apparatus 10 in this embodiment includes an XY stage 2, an imaging apparatus 3, a computer 6, a display apparatus 7, an input apparatus 8, and the like, and is a flexible substrate, a liquid crystal panel (TFT panel), a semiconductor wafer, or a metal. Defects such as scratches and spots on the surface of the inspection object 1 such as a processed product are detected by image processing. The inspection object 1 is placed on an XY stage 2 and configured to be movable in a first direction (X direction) and a second direction (Y direction) orthogonal to the first direction.

  The imaging apparatus 3 includes an optical system including a lens or a microscope having a magnification sufficient to optically enlarge the inspection object 1 to detect defects, and an image of the inspection object 1 magnified by the optical system. CCD etc. which receive light and convert it into an electric signal are provided. The photographing apparatus 3 photographs the inspection object 1 on the XY stage 2, converts it into image data, and outputs it to the computer 6. The computer 6 is an image processing device that controls the imaging device 3 and performs processing for detecting defects in the inspection object 1 based on image data obtained from the imaging device 3. The display device 7 is a display device such as a liquid crystal display connected to the computer 6. The input device 8 includes, for example, a keyboard and a mouse, and outputs an operation signal to the computer 4 in response to a user operation.

  The computer 6 in this embodiment includes a calculation unit 11 and a storage unit 12. The storage unit 12 is composed of, for example, a hard disk drive, and the storage unit 12 stores an operation system, various application programs, image processing data, and the like. The calculation unit 11 includes a CPU, a ROM, a RAM, and the like (not shown), and performs various processes such as image processing according to an operation system stored in the storage unit 12. The calculation unit 11 includes an image input unit 13, an image processing unit 14, and a defect extraction unit 15. The image input means 13 is input with captured image data (hereinafter simply referred to as a captured image) of the inspection object 1 imaged by the imaging device 3. The captured image is stored in the storage unit 12 as a processing target image.

  The image processing means 14 creates a defect component detection image obtained by extracting a defect component within a predetermined range by performing filter processing described later on the processing target image (captured image), or detects the processing target image and the defect component. A defect component-removed image in which a defect component is removed from the processing target image is created by performing addition / subtraction of a luminance value (pixel value) with the image. The defect extraction unit 15 creates a defect extraction image based on a luminance value difference between arbitrary images of the processing target image or the defect component removed image created by the image processing unit 14. The defect component includes a bright defect having a high luminance value with respect to other pixel portions and a dark defect having a low luminance value.

Next, the operation of the defect detection apparatus 10 in this embodiment will be described.
FIG. 2 is a flowchart for explaining the operation of the defect detection apparatus 10. The operation shown in FIG. 2 is realized by a program executed on the computer 6. First, when the inspection object 1 is set on the XY stage 2, the image of the inspection object 1 is imaged by the imaging device 3, and the captured image data is taken into the image input means 13 of the computing unit 11 in the computer 6 ( Image acquisition step St1). At this time, the photographed image is taken into the image input means 13 as digital data of a predetermined gradation (for example, 4096 gradations (12 bits)) by an A / D converter (not shown) and temporarily stored in the storage unit 12 as a processing target image. Is done.

FIG. 3 shows an example of the processing target image (captured image) captured by the computer 6. In the present embodiment, inspection is performed for defects on the surface of a metal workpiece as the inspection object 1.
When the captured image of the inspection object 1 is captured by the computer 6 as a processing target image, the image processing unit 14 performs image processing on the processing target image, and defect components existing within a predetermined pixel distance range are detected. The detected defect component detection image and the defect component removed image obtained by removing the defect component from the processing target image are created while increasing the pixel distance for each step. The image processing process is roughly divided into two processes: a first step image processing process that is performed first, and a subsequent step image processing process that is performed at least once after the first step image processing process. In addition, the process of step 1 is a first step image processing process, and the process after step 2 is a latter step image processing process. Each of the first-stage step image processing process and the latter-stage step image processing process includes a filter process as described below. In the filter processing in the present embodiment, the comparison with the average value or median value of the comparison pixel group is not performed as in the conventional filter processing, and the defect component is extracted from the processing target image (the captured image or the defect component removal image described below). A defect component detection image creation step for creating a detected defect component detection image and a defect component removal image creation step for creating a defect component removal image obtained by removing the defect component from the processing target image are performed.

  The image processing means 14 functions as the first stage image processing means in the present invention, and performs the first stage step image processing process. In this first-stage step image processing step, first, a first defect component detection image creation step St2 for creating a defect component detection image in which a defect component is detected from a captured image as a processing target image is performed. In this step, the difference between the luminance value of the inspection target pixel selected in the processing target image and a plurality of comparison pixels (comparison pixel group) set at positions away from the inspection target pixel by a predetermined pixel distance is calculated. Detect defects by taking. In this process, when detecting a bright defect in which the luminance value of the defective portion is higher than the surrounding luminance value, the difference between the luminance value of the pixel to be inspected and the maximum luminance value of the comparison pixel group is taken. When detecting a defect and detecting a dark defect in which the luminance value of the defective portion is lower than the surrounding luminance value, the difference between the minimum luminance value of the comparison pixel group and the luminance value of the pixel to be inspected is taken. To detect defects.

  FIG. 4 is an example of an arrangement of inspection target pixels and comparison pixels set when a defect component detection image is created. In the figure, the inspection target pixel is shown as O1, and the comparison pixels are shown as S1 to S8. The pixel distance between the inspection target pixel O1 and the comparison pixel is set to 6. That is, six pixels are arranged between the comparison pixels S1 and S5 arranged in the vertical direction with respect to the inspection target pixel O1 and the inspection target pixel O1. Similarly, six pixels are arranged between the comparison pixels S3 and S7 aligned in the horizontal direction with respect to the inspection target pixel O1 and the inspection target pixel O1. Further, the comparison pixels S2, S4, S6, and S8 that are arranged obliquely with respect to the inspection target pixel O1 are on the circumference of a virtual circle that passes through the comparison pixels S1, S3, S5, and S7 with the inspection target pixel O1 as the center. Is arranged. That is, these comparison pixels S1 to S8 are arranged at different angles by 45 degrees on the circumference. In addition, regarding the comparison pixels S2, S4, S6, and S8 that are arranged in an oblique direction with respect to the inspection target pixel O1 because the pixel shape is a square (square) shape, the number of pixels between the inspection target pixel O1 Is smaller than the number of pixels between the inspection target pixel O1 and the comparison pixels S1, S3, S5, and S7 (four in the example of FIG. 4), but the center-to-center distance between the inspection target pixel O1 (one side) Is selected such that the distance between the center of the pixel and the center of the other pixel) is closest to the center-to-center distance between the pixel O1 to be inspected and the comparison pixels S1, S3, S5, and S7. In such an arrangement, the comparison pixels S2, S4, S6, and S8 that are arranged obliquely with respect to the inspection target pixel O1 are also comparison pixels having a pixel distance of 6. The pixel distance set in the first defect component detection image creation step St2 is set according to the size of the smallest defect among the defects to be detected, and corresponds to the first distance in the present invention. .

Such defect component detection image creation includes detection of a bright defect component in which the luminance of the defective portion is higher than the surrounding luminance, and detection of a dark defect component in which the luminance of the defective portion is lower than the surrounding luminance, The detection process is different. Specifically, in the detection of the bright defect component, when the luminance value of the inspection target pixel O1 is Io1 and the luminance values of the comparison pixels are Is1 to Is8, the luminance value Iw1 of the bright defect component is expressed by the following formula ( 1).
Iw1 = Io1-min (max (Is1, Is3, Is5, Is7)
, Max (Is2, Is4, Is6, Is8)) (1)
In the above formula (1), “min” means that the minimum luminance value among the luminance values in () that follows is selected, and “max” is the maximum among the luminance values in the subsequent (). Means that the luminance value of is selected. Therefore, the maximum luminance value of Is1, Is3, Is5, and Is7 is compared with the maximum luminance value of Is2, Is4, Is6, and Is8, and the smaller luminance value is the bright defect in the inspection target pixel O1. The luminance value Iw1 of the component is set. When a negative value is generated as a result of the expression (1), the negative value is replaced with 0.
Similarly, the luminance value Ib1 of the dark defect component in the inspection target pixel O1 is calculated using the following equation (2).
Ib1 = max (min (Is1, Is3, Is5, Is7)
, Min (Is2, Is4, Is6, Is8))-Io1 (2)
Also in the above formula (2), when a negative result is obtained, it is replaced with 0.
By applying the above processing to the entire processing target image (captured image), that is, the brightness value Iw1 of the bright defect component and the brightness value Ib1 of the dark defect component are obtained for all pixels constituting the processing target image. Thus, a bright defect component detection image in which a bright defect component existing in the first distance range is detected, and a dark defect component detection image in which a dark defect component existing in the first distance range is detected. Each is created and stored in the storage unit 12. If the comparison pixel cannot be set because the number of pixels is insufficient at the edge of the processing target image, padding processing is performed to virtually expand the pixel area so that the comparison pixel can be set (also in the following modified examples) The same).

Next, a modification of the defect component detection image creation process will be described.
FIG. 5 is a diagram for explaining the first comparison pixel group used in the defect component detection image creation process in the modification, and FIG. 6 is a diagram for explaining the second comparison pixel group used in the defect component detection image creation process in the modification. FIG. Also in this modified example, eight comparison pixels S1 to S8 are set for the inspection target pixel O1, but these comparison pixels are further divided into two groups to detect defective components. That is, as shown in FIG. 5, every other comparison pixel is selected in order from S1, and a first comparison pixel group composed of these comparison pixels S1, S3, S5, and S7 is set. Further, as shown in FIG. 6, every other comparison pixel is selected in the order of S2, and a second comparison pixel group composed of these comparison pixels S2, S4, S6, S8 is set. As described above, the pixel distance set in the first defect component detection image creation step St2 is the first distance. In this modified example, as in the above example, the processing is different between the detection of the bright defect component and the detection of the dark defect component. Therefore, an example in which the bright defect component is detected will be mainly described below.

  First, defect component detection is performed based on the first comparison pixel group. In performing this defect component detection, a processing result storage image in which the luminance value of all pixels is set to 0 as an initial value at the same resolution and size as the processing target image is stored in advance as an image for storing the processing result. 12 is secured. Two processing result storage images are prepared, one for bright defects and one for dark defects.

7 to 9 are diagrams for explaining a defect component detection process (defect enhancement process) using the first comparison pixel group. In these drawings, the upper part shows a part (left end part) of the processing target image, and the lower part is a processing result storage image corresponding to the processing target image. In this step, the luminance value If1 is calculated by the following equation (3) for the inspection target pixel O1 in the processing target image.
If1 = Io1-Max (IS1, IS3, IS5, IS7) (3)
Then, the calculated If1 is compared with the luminance value (initial value = 0) of the pixel corresponding to the inspection target pixel O1 in the processing result storage image, and as shown in the lower part of FIG. The value is stored as IR1 in the corresponding pixel position of the processing result storage image.

  Subsequently, as shown in the upper part of FIG. 8, the inspection target pixel O2 is set to the pixel immediately above the inspection target pixel O1 while the positions of the comparison pixels S1, S3, S5, and S7 are fixed. The brightness value If2 calculated by the equation (3) is compared with the brightness value at the corresponding pixel position in the processing result storage image, and as shown in the lower part of FIG. Stored in the corresponding pixel position. Similarly, processing is sequentially performed from the inspection target pixel O3 immediately above the inspection target pixel O2 to the inspection target pixel O7 before the comparison pixel S1, and the luminance values IR3 to IR7 are stored in the processing result storage image. In short, the processing is performed in order from the inspection target pixel close to the inspection target pixel O1. Then, as shown in FIG. 9, for the inspection target pixels O8 to O13 between the inspection target pixel O1 and the comparison pixel S3, the same processing is performed on the inspection target pixel between the inspection target pixel O1 and the comparison pixel S5. This is sequentially performed for O14 to O19 and for the inspection target pixels O20 to O25 between the inspection target pixel O1 and the comparison pixel S7. As a result, as shown in the lower part of FIG. 9, the luminance values IR2 to IR25 are stored in a cross shape with the luminance value IR1 corresponding to the inspection target pixel O1 as the center (intersection) in the processing result storage image.

Next, defect component detection is performed using the second comparison pixel group shown in FIG.
10 to 12 are diagrams for explaining a defect component detection process using the second comparison pixel group. As described above, the processing result storage image stores the brightness value that is the processing result based on the first comparison pixel group, and the brightness value that is the processing result of the second comparison pixel group is stored therein. . In this step, the luminance value If1 is calculated by the following equation (4) for the inspection target pixel O1 in the processing target image.
If1 = Io1-Max (IS2, IS4, IS6, IS8) (4)
This calculation result If1 is compared with IR1 of the processing result storage image, and the larger value is stored in the processing result storage image as IR1 again. As shown in FIG. 11, this process is performed for O26 to O29 between the inspection target pixel O1 and the comparison pixel S2 as in the first comparison pixel group, as O30 between the inspection target pixel O1 and the comparison pixel S4. ˜O33 are sequentially performed for O34 to O37 between the inspection target pixel O1 and the comparison pixel S6 and for O38 to O41 between the inspection target pixel O1 and the comparison pixel S8. In short, the processing is performed in order from the inspection target pixel close to the inspection target pixel O1. As a result, as shown in FIG. 12, in the processing result storage image, the luminance values IR2 to IR41 are stored in an asterisk shape with the luminance value IR1 corresponding to the inspection target pixel O1 as the center (intersection).

By applying the above processing to the entire processing target image (captured image), that is, the brightness values of the bright defect components are obtained for all the pixels constituting the processing target image. In the processing result storage image, when the luminance value is already stored in the pixel position corresponding to the newly obtained luminance value, the magnitudes of the two values are compared, and the larger value is renewed as the luminance value of the pixel position. Stored in the processing result storage image. Then, the processing result storage image in which the bright defect component existing within the first distance range is detected is set as the bright defect component detection image through the binarization process. Note that the following formula (5) is used instead of the above formula (3) for the calculation of the luminance value in the detection of the dark defect component existing in the range of the first distance, and the formula (4) Instead of the following equation (6) is used.
If1 = Min (IS1, IS3, IS5, IS7) -Io1 (5)
If1 = Min (IS2, IS4, IS6, IS8) -Io1 (6)
Similarly to the creation of the bright defect component detection image, the dark defect component existing within the first distance is detected by obtaining the luminance values of the dark defect components for all the pixels constituting the processing target image. The processed result storage image is subjected to binarization processing to be a dark defect component detection image, and is stored in the storage unit 12.

  FIG. 13 shows an example of the step 1 bright defect component detection image, and FIG. 14 shows an example of the step 1 dark defect component detection image. The bright defect component detection image and the dark defect component detection image created in the first defect component detection image creation step St2 are the step 1 bright defect component detection image and the step 1 dark defect component detection image, respectively. In the following, unless otherwise specified, they are also referred to as step 1 defect component detection images.

  When the step 1 defect component detection image is created as described above, the first defect component removal image creation step St3 is subsequently performed.

  FIG. 15 is a diagram illustrating an example of a step 1 defect component removal image. In the first defect component removal image creation step, a defect component removal image is created by adding and subtracting the processing target image and the defect component detection image. More specifically, the luminance value of the step 1 bright defect component detection image is subtracted from the luminance value of the processing target image, and the luminance value of the step 1 dark defect component detection image is added to FIG. As shown, a step 1 defect component-removed image in which the bright defect and dark defect of step 1 are removed from the processing target image is obtained. Here, regarding addition or subtraction of luminance values, it means that addition or subtraction is performed on luminance values in pixels corresponding to each other in two images (the same applies hereinafter).

  Note that such a defect component detection image creation step and defect component removal image creation step may be repeated a plurality of times within the same step while maintaining the pixel distance. For example, when a defective component exists in the comparison pixel, the detection accuracy decreases. However, since the defect component is sequentially detected and removed by repeating the above process a plurality of times, the possibility of the defect component being included in the comparison pixel is gradually reduced. As a result, the detection accuracy is improved. In the case of repeating a plurality of times in this manner, regarding the second and subsequent defect component detection images, the defect component removal image is subtracted from the processing target image to create a bright defect component detection image. The dark defect detection image is created by subtracting the processing target image from the defect component removal image. At this time, a negative value may be generated by the subtraction process. In this case, the value is replaced with 0, and a bright defect component detection image and a dark defect component detection image are created. Then, the finally obtained defect component removal image is used as the defect component removal image of the step.

  As described above, the first step image processing process is performed by the image processing means 14. In this first step image processing step, a defect component (for example, noise) having a size corresponding to the smallest first distance among the set pixel distances is detected and removed. In the following, since each defect component is detected from a processing target image that is a target in which a low contrast defect having a relatively low luminance value and a high contrast defect having a relatively high luminance value are mixed, image processing is performed. The means 14 functions as a post-stage image processing means, and the post-step image processing process, that is, the second defect component detection image creation process St4 and the second defect component removal image creation, while increasing the pixel distance by a predetermined interval in and after step 2. Step St5 is repeated a plurality of steps. The maximum value of the pixel distance is set according to the size of the largest defect component that is desired to be detected.

  In the second defect component detection image creation step St4, the pixel distance is expanded more than the pixel distance in the previous step. Here, when the subsequent step image processing process is repeated a plurality of times, pixels set longer (wider) than the pixel distance set in the step (step n-1) before the current step (step n) The distance corresponds to the second distance in the present invention. In the present embodiment, the pixel distance is longer by one pixel than the pixel distance in the previous step. The extent to which the second distance is increased from the pixel distance of the previous step in each step can be arbitrarily set. Then, the defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step (that is, the defect component removal image already created in the current step is also included) is set to the second distance for each constituent pixel. Corresponding defect components are detected and a defect component detection image is created. When steps are repeated without changing the pixel distance, those steps are the same as long as the pixel distance is the same.

  FIG. 16 shows an example of the step 2 bright defect component detection image, and FIG. 17 shows an example of the step 2 dark defect component detection image. In the second defect component detection image creation step St4 in the second step (step 2) performed first after the first step image processing step, the second pixel distance is longer than the first distance in the first step (step 1). Set to distance. Then, with respect to the step 1 defect component removal image (FIG. 15), similarly to the first defect component detection image creation step St2, the step 2 bright defect component detection image (FIG. 16) is filtered by the second distance filter. Step 2 Dark defect component detection images (FIG. 17) are created.

FIG. 18 is a diagram illustrating an example of a step 2 defect component removal image.
If the step 2 defect component detection image (step 2 bright defect component detection image and step 2 dark defect component detection image) is created through the second defect component detection image creation step St4, then the second defect component removal image is created. The creation process St5 is performed. In this process, the luminance value of the step 2 bright defect component detection image is subtracted from the luminance value of the defect component removal image whose pixel distance is equal to or smaller than the pixel distance in the current step, and the luminance value of the step 2 dark defect component detection image is added. Thus, a step 2 defect component removal image as shown in FIG. 18 is obtained.

  FIG. 19 shows an example of the step 12 bright defect component detection image, and FIG. 20 shows an example of the step 12 dark defect component detection image. FIG. 21 shows an example of the step 12 defect component removal image. The second defect component detection image creation step St4 and the second defect component removal image creation step St5 are repeated while increasing the pixel distance for each step, and a step n bright defect component detection image, a step n dark defect component detection image, and Step n Defect component removed images are respectively created. Thus, the defect components are detected and removed in order from the smallest defect component, and a defect component detection image and a defect component removal image of each defect component are obtained. For example, in the step 12 defect component removal image shown in FIG. 21, defect components such as scratches and spots having relatively small sizes are removed, and only relatively large defect components such as spots and unevenness remain. .

  The post-step image processing process is performed as described above. As described above, such a defect component detection image creation step and defect component removal image creation step may be repeated a plurality of times while maintaining the pixel distance (within the same step). In this case, the defect component removal image created this time is subtracted from the defect component removal image created last time in the same step to create a bright defect component detection image, and the defect created last time from the defect component removal image created this time The component-removed image is subtracted to create a dark defect component detection image. Then, the finally obtained defect component removal image is used as the defect component removal image of the step. Here, if a relatively small size defect is included in a relatively large size defect, there is a possibility of erroneous detection in the related art. In this regard, in the present embodiment, detection errors are suppressed by detecting and removing the defects while increasing the pixel distance in order from a defect having a smaller size, so that the detection accuracy is improved. For this reason, for example, a relatively small size defect such as a scratch is included in a relatively large size defect such as unevenness or a stain in the captured image, or a structure is included in the captured image. Even in the case where there is an edge such as a portion where the contrast changes abruptly, since a defect component having a small size is sequentially detected and removed, erroneous detection is suppressed. That is, for example, even when the above-described edge component exists in the photographed image, in the defect component-removed image obtained by removing the defect component having a relatively small size, the edge contrast is lowered and the boundary is blurred, etc. Is suppressed, and the edge component is maintained as it is. When detecting / removing a defect having a size larger than a scratch or the like, a defect component removal image that has already been detected / removed is used for a relatively small size defect such as a scratch that causes false detection. , Defects can be detected and removed with high accuracy.

  When the subsequent step image processing step of each step is completed, it is subsequently determined whether or not the pixel distance in the current step is a preset maximum distance (St6). If it is determined that the pixel distance in the current step is less than the preset maximum distance (No), the process returns to step St4, the pixel distance is widened, and the subsequent step image processing step, that is, the second defect The component detection image creation process St4 and the second defect component removal image creation process St5 are executed again.

  FIG. 22 is a diagram illustrating an example of a step 100 bright defect component detection image, and FIG. 23 is a diagram illustrating an example of a step 100 dark defect component detection image. FIG. 24 is a diagram illustrating an example of a step 100 defect component removal image. In the present embodiment, the second defect component detection image creation step St4 and the second defect component removal image creation step St5 are repeated while increasing the pixel distance by one up to step 100 corresponding to the maximum pixel distance, and FIG. Step 100 bright defect component detection image, Step 100 dark defect component detection image of FIG. 23, and Step 100 defect component removal image of FIG. 24 are respectively created. Of course, the number of steps corresponding to the maximum pixel distance is not necessarily 100.

  When it is determined that the pixel distance in the current step is the preset maximum distance (Yes in St6), the defect extraction unit 15 performs a defect extraction step St7. In the defect extraction step St7 in the present embodiment, the difference in luminance value between arbitrary defect component removal images among the step n defect component removal images obtained in each step, or the captured image and the arbitrary defect component removal image A defect extraction image is created based on the difference in luminance values. In this way, not only the difference between the captured image and the defect component removed image, but also the defect extraction image in which only fine defect components such as scratches are extracted by taking the difference between the defect component removed images at any step in the middle. Alternatively, it is possible to create a defect extraction image in which only large defect components such as spots and unevenness are extracted.

For example, based on the step 12 defect component removal image (FIG. 21) and the step 1 defect component removal image (FIG. 15), an image (bright defect extraction) from which the bright defect components and dark defect components from step 2 to step 12 are extracted. Image and dark defect extraction image) can be created. The bright defect extracted image is created by the following equation (7). The dark defect extracted image is created by the following equation (8).
Bright defect extraction image =
Step 1 Defect Component Removal Image-Step 12 Defect Component Removal Image (7)
Dark defect extraction image =
Step 12 Defect Component Removal Image-Step 1 Defect Component Removal Image (8)

  FIG. 25 is a diagram illustrating an example of a bright defect extraction image from step 2 to step 12, and FIG. 26 is a diagram illustrating an example of a dark defect extraction image from step 2 to step 12. In the above step 12 defect component-removed image, defect components such as scratches and dots are removed, and in the step 1 defect component-removed image, only very small defects such as noise are removed. The defect extracted image shown in FIG. 26 is an image from which defect components having a relatively small size such as scratches and dots are extracted.

When the defect extraction process St7 is more generalized, it is assumed that “step m> step n” and “n> 1”, a bright defect extraction image and a dark defect extraction image having a size corresponding to the pixel distance from step n to step m. Formulas to be created can be represented by the following formulas (9) and (10), respectively.
Step n to m Bright defect extraction image =
Step (n-1) Defect component removed image-Step m Defect component removed image (9)
Steps n to m dark defect extraction image =
Step m Defect component removed image-Step (n-1) Defect component removed image (10)
(N-1) in the above formulas (9) and (10) means the step immediately before step n, and if there is no step corresponding to (n-1), the captured image is Step (n-1) Use as a defect component removed image.

  FIG. 27 is a diagram showing an example of a bright defect extraction image from step 1 corresponding to the first pixel distance to step 100 corresponding to the maximum pixel distance in the present embodiment, and FIG. It is a figure which shows an example of a dark defect extraction image. These images are set to m = 100 and n = 1, and using the step 100 defect component removed image and the captured image as the step (n-1) defect component removed image, the above equations (9) and (10) are used. ), Respectively. These defect extracted images are images in which defect components of all sizes corresponding to the pixel distance from Step 1 to Step 100 are extracted, including minute defect components such as noise (defect components corresponding to the pixel distance 1). It is.

  As described above, according to the image processing device and the image processing method of the present invention, the defect component detection image creation step and the defect component removal image creation step are repeated while increasing the pixel distance for each step, so that the defect component with a small size is obtained. Are sequentially detected and removed, and a defect component detection image and a defect component removal image of each defect component are obtained. And since a defect is extracted by the difference between these defect component removal images, it is possible to extract a defect from a captured image of an inspection object with higher accuracy without using smoothing or an approximate expression as in the prior art. Is possible. In particular, even when a high-contrast defect and a low-defective defect are mixed in the captured image, it is possible to detect while reducing false detection.

  FIG. 29 is a diagram illustrating an example of a bright defect extraction image from step 13 to step 100, and FIG. 30 is a diagram illustrating an example of a dark defect extraction image from step 13 to step 100. In the defect extraction step St7, for example, using the step 100 defect component removed image and the step 12 defect component removed image, an image in which the bright defect component and the dark defect component from the pixel distance 13 to the pixel distance 100 are extracted is created. You can also. In this case as well, a bright defect extracted image and a dark defect extracted image are created based on the above equations (9) and (10). The step 12 defect component-removed image is an image obtained by removing defect components in the pixel distance range from step 1 to step 12 such as scratches and dots, and the step 100 defect component-removed image is, for example, This is an image from which defective components having a pixel distance in the range of 1 to 100 such as unevenness are removed. For this reason, the defect extraction image obtained using these defect component removal images is an image in which defect components such as unevenness are mainly extracted. That is, it is possible to extract only a defect component having a desired size by taking a difference between defect component-removed images in an arbitrary step. Thereby, the freedom degree of defect extraction improves.

Here, the defect extraction image obtained by the above formulas (9) and (10) is an image that does not include a defect component having a size corresponding to the pixel distance in the steps of (n-1) and subsequent steps. That is, for example, the defect extraction images from step 13 to step 100 shown in FIGS. 29 and 30 are created by the difference between the step 12 defect component removal image and the step 100 defect component removal image as described above. For this reason, the defect component having a relatively small size that has been removed up to step 12 is not included in the defect extraction images from step 13 to step 100, and the defect component is filled as if it did not originally exist. The image looks like this. Therefore, from the defect extraction images from step 13 to step 100, it is possible to grasp a relatively large defect such as unevenness, while the presence of a relatively small defect component removed by step 12 is present. Can not grasp. The defect component corresponding to Step 1 to Step 12 is not extracted as a defect on the defect extraction image having a size corresponding to the pixel distance from Step 13 to Step 100, but is removed from the defect extraction image in a form that is not filled. In this case, defect extraction may be performed by the following formulas (11) and (12). Note that the processing target image in the following expression is a captured image.
Bright defect extracted image = (processing target image−step 100 defect component removed image)
-(Processing target image-Step 12 defect component removed image) (11)
Dark defect extracted image = (Step 100 defect component removed image−processing target image)
-(Step 12 defect component removed image-image to be processed) (12)
In the above formulas (11) and (12), for example, when it is desired to obtain an extracted image from which noise has been removed, the step 1 defect component-removed image is used as the noise-removed image instead of the processing target image (captured image). That's fine. That is, defect extraction is performed by the following formulas (13) and (14).
Bright defect extracted image = (noise-removed image−step 100 defect component-removed image)
-(Noise-removed image-step 12 defect component-removed image) (13)
Dark defect extracted image = (Step 100 defect component removed image−noise removed image)
-(Step 12 defect component removed image-noise removed image) (14)

FIGS. 31 and 32 are diagrams showing an example of the defect extraction image when the defect components from step 1 to step 12 are subtracted from the defect extraction image from step 13 to step 100. FIG. 31 shows the bright defect extraction image. FIG. 32 is a diagram illustrating an example of a dark defect extraction image.
In this way, when the defect component removal image of the smaller (front) step is the defect component removal image of the intermediate step (step 2 or more), the defect component corresponding to the previous step is filled in. In the case of removing without, generally, “step m> step n”, “n> 1”, a bright defect extraction image and a dark defect extraction image having a size corresponding to the pixel distance from step n to step m. Equations for creating can be expressed by the following equations (15) and (16), respectively. Note that the processing target image in the following expression is a captured image.
Bright defect extracted image = (processing target image−step m defect component removed image)
-(Processing target image-Step (n-1) Defect component removed image) (15)
Dark defect extracted image = (step m defect component removed image−processing target image)
-(Step (n-1) Defect Component Removal Image-Processing Target Image) (16)
In the above formulas (15) and (16), when it is desired to obtain an extracted image from which noise has been removed as described above, the defect-removed image in step 1 is subjected to noise removal instead of the processing target image (captured image). What is necessary is just to use as an image. That is, defect extraction is performed by the following formulas (17) and (18).
Bright defect extraction image = (noise removal image−step m defect component removal image)
-(Noise-removed image-Step (n-1) Defect component-removed image) (17)
Dark defect extraction image = (step m defect component removal image−noise removal image)
-(Step (n-1) Defect component removed image-Noise removed image) (18)
As a result, it is possible to grasp a defect having a large size corresponding to the pixel distance between these steps from the defect extraction image from step n to step m, and the defect removed by the step before step n. It is also possible to grasp the presence of ingredients.

  In the above embodiment, the configuration in which the maximum value of the pixel distance is set to a value that covers the defect of the maximum size is exemplified. However, the maximum value of the pixel distance is a value that is larger than the defect of the maximum size. It is possible to further improve the defect detection accuracy by performing the filter processing with setting to.

  FIG. 33 is a diagram illustrating a pattern image as a processing target image in the second embodiment. FIG. 34 is a diagram illustrating an example of a bright defect extraction image from step 1 to step 9, and FIG. 35 is a diagram illustrating an example of a dark defect extraction image from step 1 to step 9. In the pattern image shown in FIG. 33, there are a square frame-shaped pattern, a horizontal line pattern arranged in total in the vertical direction inside the frame-shaped pattern, a horizontally long rectangular pattern on the lower left side, and a circular pattern on the lower side. Each of them is assumed to be a defect (bright defect). Then, in the pattern image shown in FIG. 33, from step 1 to step 9, the pixel distance is increased by 1 and the second defect component detection image creation step St4 and the second defect component removal image creation step St5 are repeated. A component-removed image is created, and m = 9, n = 1, and the bright defect extraction image from step 1 to step 9 is obtained as shown in FIG. Similarly, as shown in FIG. 35, the dark defect extraction image from step 1 to step 9 is obtained by the above equation (10) with m = 9 and n = 1.

  When the pattern image shown in FIG. 33 is compared with the bright defect extraction image shown in FIG. 34, the pixel distance filter processing corresponding to step 1 to step 9 covers only the defect size of the frame or horizontal line pattern. A defect having a size larger than this, that is, a rectangular pattern and a circular pattern are not detected as bright defects. As for the horizontal line pattern, the portion where the pattern exists is extracted as a bright defect, but the contrast of the bright defect extracted image is lower than that of the pattern image. Further, when comparing the pattern image shown in FIG. 33 and the dark defect extraction image shown in FIG. 35, the portions between both ends of each line are extracted as dark defects for the six horizontal line patterns. This is a false detection.

  FIG. 36 is a diagram illustrating an example of a bright defect extraction image from step 1 to step 65, and FIG. 37 is a diagram illustrating an example of a dark defect extraction image from step 1 to step 65. When the subsequent step image processing process is repeated up to step 65 in order to increase the pixel distance, the filter of the size can cover patterns of all sizes. For this reason, in the bright defect extraction image from step 1 to step 65 shown in FIG. 36, a rectangular pattern and a circular pattern that were not detected until step 9 are detected. Further, the horizontal line pattern is detected as a bright defect, and there is no problem with the contrast of the bright defect extracted image. Further, in the dark defect extraction image from Step 1 to Step 65 shown in FIG. 37, no erroneous detection is observed between both ends of the horizontal line that was erroneously detected as a dark defect until Step 9. Thus, it can be seen that the detection accuracy improves as the pixel distance increases.

  Note that when the processing target image includes, for example, only a bright defect, if the bright defect component and the dark defect component are detected and removed as described above, the defect detection accuracy may decrease. In particular, if the distance between the defects is short, erroneous detection such as detection as a dark defect despite the fact that it is a bright defect is likely to occur. Therefore, in this case, such a problem can be suppressed by detecting and removing only the bright defect component. Similarly, when only dark defect components exist, defect detection accuracy can be improved by performing processing by detecting and removing only dark defect components.

  In the above embodiment, an example in which a defect having an arbitrary size is extracted using a difference between defect component removal images created in different steps and having different pixel distances. However, the defect component removal image is not used. It is also possible to extract defects using a defect component detection image (bright defect component detection image, dark defect component detection image). For example, assuming that the size of a defect component to be extracted as a defect corresponds to the pixel distance from step 21 to step 33, the defect component detection images obtained in each step are added together and subjected to binarization processing. A defect component extraction image obtained by extracting the defect component can be obtained. In this case, the result obtained by the difference between the defect component removed image in step 20 and the defect component removed image in step 33 is the same.

  Moreover, although the said embodiment illustrated the structure which produces a defect extraction image by the difference of the two defect component removal images produced at a different step, it is not restricted to this. For example, a defect extraction image can be created using a plurality (three or more) of defect component removal images created in different steps.

  Furthermore, it is also possible to detect a defect from another defect component removed image by other image processing. For example, it is possible to extract defects by image processing using a median filter or a smoothing process for the step 100 defect component-removed image shown in FIG. That is, the defect component-removed image at a predetermined step is an image from which a defect component having a size corresponding to the pixel distance to the previous step has been removed. Components can be extracted by combining other image processing.

The processing target image in the first defect component detection image creation step (St2) is not limited to the captured image obtained in the image acquisition step. For example, a general smoothing process or a median filter is applied to generate noise. An image other than the captured image, such as a removed image or an image obtained as a result of applying various other image processing such as edge detection processing, may be used as the processing target image.
Furthermore, instead of the noise-removed image obtained by applying the present technology, an image obtained as a result of applying various other image processing may be used as the processing target image.

  The present invention also detects foreign matter defects in a flexible substrate as an inspection object, detection of scratches and dirt on the surface of the inspection object, and luminance spot defects and color spot defects of various display devices as the inspection object. It can be widely applied to detection and the like.

  DESCRIPTION OF SYMBOLS 1 ... Test object, 2 ... XY stage, 3 ... Imaging apparatus, 6 ... Computer, 7 ... Display apparatus, 8 ... Input device, 10 ... Defect detection apparatus, 11 ... Calculation part, 12 ... Memory | storage part, 13 ... Image input Means, 14 ... image processing means, 15 ... defect extraction means

Claims (8)

A first defect component detection image creating step for creating a defect component detection image for a photographed image obtained by imaging the inspection object, and a first defect for creating a defect component removal image obtained by removing the defect component from the photographed image First-stage image processing means for performing a first-stage step image processing process including a component removal image creation process;
A second defect component detection image creation step for creating a defect component detection image for the defect component removal image, and a second defect component removal image creation for creating a new defect component removal image by removing the defect component from the defect component removal image A post-stage image processing means for executing a post-stage image processing process including the process;
Defect extraction means for creating a defect extraction image based on a difference in luminance value between arbitrary images of the captured image or the defect component removal image obtained in each step;
With
The first stage image processing means includes:
In the first defect component detection image creation step,
By setting the pixel distance indicating the number of pixels between the inspection target pixel and the comparison pixel to be compared with the inspection target pixel to the first distance,
For the captured image, a defect component detection image is created by detecting a defect component having a size corresponding to the first distance for each pixel constituting the image,
In the first defect component removal image creation step,
Create a defect component removal image by adding and subtracting a luminance value between the captured image and the defect component detection image created in the first defect component detection image creation step,
The latter stage image processing means includes:
In the second defect component detection image creation step,
By setting the pixel distance to a second distance that is longer than the pixel distance in the previous step,
For a defect component removal image having a pixel distance equal to or less than the pixel distance in the current step, a defect component corresponding to the second distance is detected for each pixel constituting the defect component detection image,
In the second defect component removal image creation step,
Creating a new defect component removal image by adding and subtracting luminance values of the defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step and the defect component detection image created in the second defect component detection image creation step;
An image processing apparatus. In the first defect component detection image creation step, the first-stage image processing means sets a plurality of comparison pixels at the first distance with respect to the inspection target pixel for the photographed image, and the luminance value of each comparison pixel and the inspection target Create a defect component detection image based on the difference from the luminance value of the pixel,
In the second defect component detection image creating step, the subsequent image processing means sets a plurality of comparison pixels at the second distance with respect to the inspection target pixel for a defect component removal image whose pixel distance is equal to or smaller than the pixel distance in the current step. The image processing apparatus according to claim 1, wherein a defect component detection image is created based on a difference between a luminance value of each comparison pixel and a luminance value of the inspection target pixel.   In the first defect component detection image creation step and the second defect component detection image creation step, the first-stage image processing means and the subsequent-stage image processing means respectively set the pixel distance according to the size of a defect to be detected. The image processing apparatus according to claim 2, wherein the image processing apparatus is set.   4. The image processing apparatus according to claim 1, wherein the post-stage image processing unit sequentially increases the pixel distance for each step and executes the post-stage image processing step a plurality of steps. 5. . A first defect component detection image creating step for creating a defect component detection image for a photographed image obtained by imaging the inspection object, and a first defect for creating a defect component removal image obtained by removing the defect component from the photographed image A first step image processing step including a component removal image creation step;
A second defect component detection image creation step for creating a defect component detection image for the defect component removal image, and a second defect component removal image creation for creating a new defect component removal image by removing the defect component from the defect component removal image A post-step image processing process including a process;
A defect extraction step of creating a defect extraction image by a difference in luminance value between arbitrary images of the captured image or the defect component removal image obtained in each step;
Including
In the first defect component detection image creation step,
By setting the pixel distance indicating the number of pixels between the inspection target pixel and the comparison pixel to be compared with the inspection target pixel to the first distance,
For the captured image, a defect component detection image is created by detecting a defect component having a size corresponding to the first distance for each pixel constituting the image,
In the first defect component removal image creation step,
Create a defect component removal image by adding and subtracting a luminance value between the captured image and the defect component detection image created in the first defect component detection image creation step,
In the second defect component detection image creation step,
By setting the pixel distance indicating the number of pixels between the inspection target pixel and the comparison pixel to be compared with the inspection target pixel to a second distance longer than the pixel distance in the previous step,
For a defect component removal image having a pixel distance equal to or less than the pixel distance in the current step, a defect component corresponding to the second distance is detected for each pixel constituting the defect component detection image,
In the second defect component removal image creation step,
Creating a new defect component removal image by adding and subtracting the defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step and the defect component detection image created in the second defect component detection image creation step;
An image processing method. In the first defect component detection image creation step, a plurality of comparison pixels are set at the first distance with respect to the inspection target pixel in the captured image, and the difference between the luminance value of each comparison pixel and the luminance value of the inspection target pixel Create a defect component detection image based on
In the second defect component detection image creation step, a plurality of comparison pixels are set at the second distance with respect to a pixel to be inspected for a defect component removal image whose pixel distance is equal to or less than the pixel distance in the current step, and the luminance value of each comparison pixel The image processing method according to claim 5, wherein a defect component detection image is created based on a difference between a luminance value of the inspection target pixel and the luminance value of the inspection target pixel.   7. The image according to claim 6, wherein in the first defect component detection image creation step and the second defect component detection image creation step, the pixel distance is set according to the size of a defect to be detected. Processing method.   The image processing method according to any one of claims 5 to 7, wherein the subsequent step image processing step is executed in a plurality of steps by sequentially increasing the pixel distance for each step.