JP2007172397A - Edge gradient detection method, stain defect detection method, edge gradient detection device and stain defect detection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an edge gradient detection method capable of detecting an edge gradient with high accuracy. <P>SOLUTION: This edge gradient detection method has: a first edge gradient calculation process S92 for finding a luminance change rate of a target pixel selected in an imaged image and each first comparison pixel disposed adjacently to the target pixel around the target pixel; a second edge gradient calculation process S93 for finding a luminance change rate of the target pixel selected in the imaged image and each second comparison pixel disposed apart from the target pixel as compared with the first comparison pixel around the target pixel; and an edge gradient composition process S96 for setting the luminance change rate having a maximum absolute value among the respective luminance change rates calculated in the first and second edge gradient calculation processes as an edge gradient value of the target pixel. Because finding and composing the edge gradient of the first comparison pixel adjacent to the target pixel and the edge gradient of the second comparison pixel, edge components in various kinds of directions each including even high frequency component can be easily detected with high accuracy. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention provides an edge gradient detection method and apparatus that automatically and accurately detects an edge gradient in an inspection process of various products such as an inspection process in the manufacture of a display device such as a liquid crystal panel and its application product, such as a projector, The present invention relates to a spot defect detection method and apparatus using the edge gradient detection method and apparatus.

In the inspection of LCD panels such as TFT panels, surface defects called spots and unevenness are difficult to detect automatically with an inspection apparatus because of their indefinite shape and low contrast. For this reason, the inspection is still carried out by the inspector's visual observation, and there is an urgent need to automate the inspection in order to reduce manufacturing costs.
Note that a spot or uneven defect is a state in which an area on a display screen has a difference in luminance from other areas, and in a certain range, there is a brighter or darker part than the surrounding area. Usually, a case where the defect area is relatively small is often referred to as a spot defect, and a case where the defect area is relatively large is often referred to as a mura defect. However, since there is no strict definition, in the present invention, surface system defects such as a spot defect and a mura defect are collectively referred to as a spot defect.

  When such a spot defect is visually inspected, the defect non-defective product level determination has a great influence not only on the defect area and contrast feature amount but also on the luminance change rate (edge gradient) of the edge portion of the defect. Accordingly, even when defect inspection is automated by performing image processing on a spot defect, it is required to accurately detect the edge gradient of the defect.

  By the way, as a method for obtaining an edge gradient in image processing, a method using a Sobel filter or a Prebitt filter has been conventionally known (for example, see Non-Patent Document 1).

  Each of these filters often uses a 3 × 3 filter. For example, as in the prebit filter shown in FIGS. 19A and 19B, an fx filter that obtains an edge gradient in the X direction, Using the fy filter for obtaining the edge gradient, the difference between each pixel one pixel before and one pixel behind the pixel to be inspected (center coordinates of each filter) is obtained to obtain the edge gradient of the pixel to be inspected.

Supervised by Mikio Takagi and Yoshihisa Shimoda, "Image Analysis Handbook", First Edition, The University of Tokyo Press, January 17, 1991, p.550-555

However, each of the filters obtains the edge gradient of the pixel to be inspected by obtaining the difference between each pixel one pixel before and one pixel behind the pixel to be inspected, and therefore, as shown in FIGS. 20 (A) and 20 (B). Even when the luminance change is different, detection may be performed with the same luminance change rate (edge gradient), and the accuracy is not necessarily high.
That is, the luminance values of the pixel B to be inspected, the pixel A before the pixel A, and the pixel B after the pixel are gradually increased to “0, 1, 2” as shown in FIG. Even if each filter is applied in the case where the filter is applied and when it is abruptly changed to “0, 2, 2” as shown in FIG. 20B, both results are the edge gradient (gradient line ga in FIG. 20). , Gb) have the same inclination. As described above, the conventional filter has a problem in that it may not be accurately detected even when the luminance change state of the edge portion is different.

  Furthermore, since a 3 × 3 filter is used, the edge gradient direction that can be detected is limited to only eight directions, and there is a problem that the edge gradient cannot be detected with higher accuracy.

  In addition, since the edge gradient cannot be detected with high accuracy, there is also a problem that when the stain defect is automatically detected, the stain defect cannot be detected with high accuracy.

  The present invention has been made in view of the problems as described above, and provides an edge gradient detection method and apparatus capable of accurately detecting an edge gradient, and a stain defect detection method and apparatus capable of accurately detecting a spot defect. With the goal.

  The edge gradient detection method of the present invention is a first method for obtaining a luminance change rate between a target pixel selected in a captured image and each first comparison pixel arranged around the target pixel and adjacent to the target pixel. Luminance change rates of the edge gradient calculation step, the target pixel selected in the captured image, and the second comparison pixels arranged around the target pixel and further away from the first comparison pixel than the target pixel, respectively A second edge gradient calculation step to be obtained; an edge gradient synthesis step in which the luminance change rate calculated in the first and second edge gradient calculation steps has the maximum absolute value as the edge gradient value of the target pixel; It is characterized by providing.

In the present invention, the luminance change rate (edge gradient) between the target pixel and the first comparison pixel adjacent to the target pixel is calculated in the first edge gradient calculating step, and the target pixel is calculated in the second edge gradient calculating step. Since the luminance change rate (edge gradient) between the pixel and the second comparison pixel arranged away from the target pixel is calculated, an edge portion where the luminance changes abruptly, an edge portion where the luminance changes gently to some extent, etc. It is possible to accurately detect the edge gradients of various edge portions.
In addition, since edge gradients having different distances between the target pixel and each comparison pixel are calculated and the results are synthesized, edge components including high frequency components can be detected with high accuracy.
Furthermore, the first comparison pixel adjacent to the target pixel can only detect edge gradients in a maximum of 8 directions, but the second comparison pixel can detect edge gradients in more directions such as 12 directions and 16 directions. Detection accuracy can be further improved, and edge components in various directions including high-frequency components can be detected and evaluated easily and accurately.

Here, in the edge gradient detection method according to claim 1, in each of the edge gradient calculation steps, the luminance of the target pixel is O, the luminance of the nth comparison pixel is Sn, and the coordinates of the target pixel are (X ₀ , Y ₀ ), the coordinate of the comparison pixel is (Xn, Yn), and the distance between the target pixel and the nth comparison pixel is D (O, Sn), the edge gradient (luminance change rate) gn and the distance D (O, Sn) ) Is preferably obtained by the following formulas 1 and 2.

  With such a configuration, the edge gradient gn can be easily calculated, and the edge detection process can be executed in a short time. In addition, since the distance between the pixels is obtained from the coordinates of the target pixel and the comparison pixel, the edge gradient can be obtained with high accuracy.

  Further, a third edge for respectively obtaining a luminance change rate between the target pixel selected in the captured image and each third comparison pixel arranged around the target pixel and further away from the second comparison pixel than the target pixel. A gradient calculating step, wherein the edge gradient synthesizing step uses the luminance change rate calculated in each of the first to third edge gradient calculating steps as the edge gradient value of the target pixel having the maximum absolute value. Is preferred.

  With such a configuration, since the edge gradients having different distances between the target pixel and each comparison pixel are further calculated and combined, the edge component including the high frequency component can be detected with higher accuracy.

At this time, it is preferable that the second comparison pixel is disposed adjacent to the first comparison pixel, and the third comparison pixel is disposed adjacent to the periphery of the second comparison pixel.
With this configuration, the first comparison pixel can detect the edge gradient for one pixel from the target pixel, the second comparison pixel can detect the edge gradient for two pixels from the target pixel, and the third comparison pixel An edge gradient corresponding to a distance of 3 pixels from the target pixel can be detected. For this reason, it is possible to accurately detect the luminance change of the edge portion of the spot defect including the high frequency component.

  The spot defect detection method of the present invention is a spot defect area extraction step for extracting a spot defect area from an image obtained by imaging a spot defect inspection target, and any one of claims 1 to 4 for each pixel of the spot defect area. An edge gradient detection step in which the edge gradient value of each target pixel having the maximum absolute value among the edge gradient values of each target pixel is set as the edge gradient value of the stain defect region, and the area of the stain defect region An area calculating step for calculating the contrast, a contrast calculating step for obtaining the contrast of the spot defect area, and a defect rank evaluation for evaluating the defect rank of the spot defect area based on the three types of feature values of the edge gradient, area, and contrast And a process.

  In the present invention, a spot defect can be evaluated and ranked based on three types of feature values of edge gradient, area, and contrast, and therefore can be detected with high accuracy. In particular, since the edge gradient can be detected accurately and automatically, a spot defect can be detected accurately and automatically, and the efficiency of inspection work for a liquid crystal panel or the like can be realized.

  The edge gradient detection device according to the present invention is a first method for obtaining a luminance change rate between a target pixel selected in a captured image and each first comparison pixel arranged around the target pixel and adjacent to the target pixel. Luminance change rates of the edge gradient calculating means, the target pixel selected in the captured image, and the second comparison pixels arranged around the target pixel and further away from the first comparison pixel than the target pixel, respectively Second edge gradient calculating means to be obtained; and edge gradient synthesizing means that uses an edge gradient value of the target pixel that has a maximum absolute value among the luminance change rates calculated by the first and second edge gradient calculating means; It is characterized by providing.

  In the present invention, as in the edge gradient detection method, it is possible to accurately detect the gradients of the edge portions of various spot defects such as the edge portion where the luminance changes abruptly and the edge portion where the luminance changes gently to some extent. Can do. In addition, since edge gradients with different distances between the target pixel and each comparison pixel are calculated and synthesized, the edge components in various directions including high-frequency components can be detected and evaluated easily and accurately. Can do.

  The spot defect detection apparatus according to the present invention includes a spot defect area extracting unit that extracts a spot defect area from an image obtained by imaging a spot defect inspection target, a target pixel selected in the spot defect area, and a target pixel around the target pixel. A first edge gradient calculating means for obtaining a luminance change rate with each of the first comparison pixels arranged adjacent to the pixel, a target pixel selected in the spot defect area, and surrounding the target pixel and from the target pixel A second edge gradient calculating means for obtaining a luminance change rate with each of the second comparison pixels arranged apart from the first comparison pixel; and each of the luminance changes calculated by the first and second edge gradient calculating means. An edge gradient synthesis unit that uses an edge rate having a maximum absolute value among the ratios as an edge gradient value of the spot defect region; an area calculation unit that calculates an area of the spot defect region; Contrast calculation means for obtaining a contrast, and defect rank evaluation means for evaluating a defect rank of the spot defect region based on the three types of feature values of the edge gradient, area, and contrast.

  In the present invention, the spot defect can be evaluated and ranked based on three kinds of feature values of edge gradient, area, and contrast, as in the spot defect detection method. It can detect well and automatically, and can realize the efficiency of inspection work such as liquid crystal panel.

  In the edge gradient detection device and the spot defect detection device, the target pixel selected in the captured image and each third comparison arranged around the target pixel and further away from the target pixel than the second comparison pixel. Third edge gradient calculating means for obtaining a luminance change rate with respect to each pixel is provided, and the edge gradient combining means has a maximum absolute value among the luminance change rates calculated by the first to third edge gradient calculating means. May be used as the edge gradient value of the target pixel.

FIG. 1 is a block diagram showing the configuration of a spot defect detection apparatus according to an embodiment of the present invention.
In FIG. 1, reference numeral 1 denotes a liquid crystal panel (liquid crystal light valve) to be inspected, and reference numeral 2 denotes a projector which is an image projection device, which can set the liquid crystal panel 1 from the outside. Reference numeral 3 denotes a pattern generator that is a pattern generation device that outputs various patterns to the liquid crystal panel 1. Reference numeral 4 denotes a screen. Reference numeral 5 denotes a CCD camera that is an image pickup unit that captures an image projected on the screen 4. It has a CCD with a resolution of. Reference numeral 6 denotes a computer device which is an image processing means for controlling the pattern generator 3 and the CCD camera 5 to detect a spot defect in the liquid crystal panel 1, and 7 is a display device connected to the computer device 6.

  The computer device 6 includes an image input unit 60, a background image difference processing unit 61, a display area extraction unit 62, a reduced image creation unit 63, a spot defect enhancement processing unit 64, a noise removal unit 65, and a spot defect extraction. The processing unit 66, the blob processing unit 67, the edge gradient detection unit 69, and the defect rank classification processing unit 68 are configured.

Image data of the captured image captured by the CCD camera 5 is input to the image input means 60 of the computer device 6. The captured image is stored in a storage means (not shown). Accordingly, an imaging process (image input process) is performed in which the inspection target is imaged using the CCD camera 5 by the image input means 60.
The background image difference processing means 61 carries out a background image difference processing step for obtaining a background difference image in which a difference between an input image and a background image created in advance is removed and defective luminance changes caused by things other than the inspection target are removed. To do. The display area extracting means 62 performs a display area extracting step of extracting only the image portion of the part to be inspected from the background difference image. When an image of a TFT panel to be inspected is picked up by the CCD camera 5, in order to pick up an image of the entire TFT panel, the surroundings of the panel must be somewhat copied. For this reason, there is data unrelated to the panel around the panel projection image in the captured image, and the projection image is not rectangular but distorted. Therefore, the display area extraction unit 62 performs a process of cutting out only the portion of the TFT panel projection image, and creates an inspection target image.

  The reduced image creating unit 63 performs a reduced image creating step of creating a reduced image from the image extracted by the display area extracting unit 62. For example, the reduced image creating means 63 performs a process of reducing the captured image to 1/8 size in order to detect a relatively large spot defect in the captured image. As will be described later, the size of the stain that can be emphasized is determined to some extent in the stain defect enhancement filter used in the stain defect enhancement processing means 64. For this reason, even if the same spot defect emphasis filter is used, a relatively large spot defect is detected by reducing the size of the spot defect by reducing the size of the image itself as compared with the case where the image is not reduced. be able to. Note that when a small spot defect is detected, the reduced image creating step by the reduced image creating means 63 may not be performed.

The spot defect enhancement processing unit 64 performs a spot defect enhancement process step of enhancing and detecting a spot defect by applying a spot defect enhancement filter (stain defect detection filter) to the image.
The noise removing unit 65 performs a noise removing step of removing noise by applying a median filter to the result obtained by the spot defect enhancement processing unit 64.

The spot defect extraction processing unit 66 extracts a spot defect candidate by comparing the result processed by the noise removing unit 65 with a predetermined threshold. Note that the spot defect includes a white spot defect having a high luminance value with respect to other pixel portions and a black spot defect having a low luminance value. For this reason, the white spot defect threshold value and the black spot defect threshold value are set as the threshold values, and the white spot defect candidate is extracted by comparing with the white spot defect threshold value, and the black spot defect threshold value is compared with the black spot defect threshold value. A defect candidate is extracted.
The blob processing means 67 performs a blob processing step for obtaining the area of the region extracted as the defect candidate and the average luminance.

  Edge gradient detection means (edge gradient detection device) 69 applies the edge gradient detection method of the present invention to the image extracted by the display area extraction means 62 and detects the maximum edge gradient in the spot defect candidate region. A gradient detection step is performed. The edge gradient means the rate of change in luminance at the edge portion. Specifically, the difference in luminance between two pixels of the image is divided by the distance between the pixels.

The defect rank classification processing means 68 evaluates the rank of the spot defect based on the area and average luminance obtained by the blob processing means 67 and the edge gradient obtained by the edge gradient detection means 69, and which defect is the object to be inspected this time. A defect rank classification process is performed for classifying whether the rank is satisfied.
Therefore, in the present embodiment, the spot defect detection apparatus is configured to include the spot defect extraction processing unit 66, the blob processing unit 67, the edge gradient detection unit 69, and the defect rank classification processing unit 68.

Next, the operation of the spot defect detection apparatus according to the embodiment of the present invention will be described.
FIG. 2 is a flowchart for explaining the operation of the spot defect detection apparatus of this embodiment. The operation shown in FIG. 2 is realized by a program executed on the computer device 6.
First, the liquid crystal panel 1 to be inspected is set in the projector 2, the pattern generator 3 is controlled by the computer device 6 to display a specific brightness pattern on the liquid crystal panel 1, and this is projected onto the screen 4 by the projector 2. To do. Then, the image projected on the screen 4 is photographed by the CCD camera 5, the image of the photographed data is output to the computer device 6, and a spot defect detection process is performed by the computer device 6 to detect a spot defect in the liquid crystal panel 1. The result is displayed on the display device 7.

Here, the operation of spot defect detection by the computer device 6 will be described based on the flowchart of FIG.
First, an image projected on the screen 4 is photographed by the CCD camera 5, and an image of the photographed data is taken into the image input means 60 of the computer device 6, and an image input process (imaging process) is performed (step S1). . At this time, the photographing data is taken into the computer device 6 as digital data of, for example, 4096 gradations (12 bits) by an A / D converter (not shown). Note that the imaging data is not limited to 4096 gradations, and may be 1024 gradations (10 bits) or 256 gradations (8 bits). The higher the gradation, the higher the accuracy of processing becomes possible. However, since an expensive CCD camera 5 capable of acquiring high gradation data is required, the gradation of the imaged data depends on the defect inspection target or the like. What is necessary is just to set to a required gradation.
Next, the background image difference processing means 61 performs a background image difference processing step for removing defective luminance changes caused by things other than the liquid crystal panel such as illumination and lenses from the captured image data (step) S2).

  In this background image difference processing step, the background image shown in FIG. 3C is created by subtracting the background image shown in FIG. 3B from the input image (projected image) shown in FIG. The background image is an image of luminance change of the optical system excluding the liquid crystal panel 1. Since the luminance change on the defect due to the projection lamp and the projection lens occurs in both the input image and the background image, if the background image is subtracted from the input image, the background difference image is not a liquid crystal panel such as a projection lamp or a projection lens. The luminance change component on the defect caused by the object is removed.

Subsequently, the display area extracting means 62 performs a display area extracting process for extracting only the screen portion of the part to be inspected (step S3).
In the display area extraction processing step, the coordinates of the four corners of the inspected part image (background difference image) shown in FIG. 4A are subjected to pattern matching processing (four small regions of several tens of pixels by several tens of pixels near the four corners of the image data). The four corner reference images prepared in advance are subjected to pattern matching processing and the coordinates of the four corners are specified), and the display is performed by affine transformation so that the positional relationship of the coordinates of the four corners is rectangular. Extract area. As a result, as shown in FIG. 4B, the peripheral edge on the screen 4 is removed, and only the screen portion having an accurate rectangle is extracted.

Next, the reduced image creating means 63 performs a reduced image creating process for reducing the inspection target image that has undergone the display area extraction process (step S4).
In the reduced image creation processing step, an image reduced to a predetermined size, for example, 1/4 size, is created from the inspection target image. Specifically, when a 1/4 size reduced image is created, a 1/2 size reduced image is created by setting the average value of 4 pixels of the inspection target image to 1 pixel. A reduced image of ¼ size is created by setting the average value of four pixels of the reduced image to one pixel.

Therefore, if the flattened original image is 1.3 million pixels of 1300 × 1000, the reduced image of ½ is 325,000 pixels of 650 × 500, and the reduced image of ¼ is 80,000 of 325 × 250. It will have 1250 pixels.
In this way, a reduced image of a predetermined size is created because a stain defect enhancement filter, which will be described later, is designed to emphasize a spot defect of a predetermined size, and by reducing the image size, This is to make it possible to emphasize a relatively large spot that cannot be emphasized by a spot defect enhancement filter, which will be described later, by reducing the image size.

Next, the spot defect emphasis processing means 64 performs a spot defect emphasis process step for emphasizing a spot defect on the created reduced image (step S5). This spot defect enhancement processing step emphasizes only a spot defect in an image because it is difficult to detect a minute level of white spot defect / black spot defect as it is.
In addition, when detecting only a spot defect of a predetermined size, it suffices to apply one kind of spot defect enhancement filter to the created reduced image, but when detecting a spot defect having a different size, A plurality of types of spot defect enhancement filters having different sizes of spot defects to be detected may be applied to detect a spot defect of a predetermined size.

As shown in FIG. 5, the spot defect enhancement filter has a central pixel in a set detection region as a target pixel, is arranged in a substantially circular shape around the target pixel, and is arranged a predetermined length away from the target pixel. A luminance comparison pixel is provided.
Then, two luminance comparison pixels arranged at symmetrical positions with the target pixel in between are set as one set, and all luminance comparison pixels are divided into sets. Then, the difference F is obtained by subtracting the average luminance value of each set of luminance comparison pixels from the luminance value of the target pixel. That is, when the luminance value of the target pixel is “O” and the luminance values of the luminance comparison pixels are “S1 to S32”, the differences F1 to F16 are calculated using the following equations 3 to 18.

F1 = O- (S1 + S17) / 2 (3)
F2 = O- (S2 + S18) / 2 (4)
F3 = O- (S3 + S19) / 2 (5)
F4 = O- (S4 + S20) / 2 ... (6)
F5 = O- (S5 + S21) / 2 (7)
F6 = O- (S6 + S22) / 2 (8)
F7 = O- (S7 + S23) / 2 (9)
F8 = O- (S8 + S24) / 2 (10)
F9 = O- (S9 + S25) / 2 (11)
F10 = O− (S10 + S26) / 2 (12)
F11 = O- (S11 + S27) / 2 (13)
F12 = O- (S12 + S28) / 2 (14)
F13 = O- (S13 + S29) / 2 (15)
F14 = O− (S14 + S30) / 2 (16)
F15 = O− (S15 + S31) / 2 (17)
F16 = O- (S16 + S32) / 2 (18)

After performing the above calculation, the spot defect enhancement filter selects a value having the minimum absolute value from the values F1 to F16, and uses the value as the result of the spot enhancement.
For example, when there is no defect in the area surrounded by the luminance comparison pixels, there is almost no difference between the luminance value of the target pixel and the luminance value of each luminance comparison pixel. Accordingly, F1 to F16 are all small values.
On the other hand, as shown in FIG. 6, when a white spot defect 70 brighter than other parts is present in the area surrounded by the luminance comparison pixels and the target pixel is a part thereof, each of the pixels is compared with the target pixel. The luminance value of the luminance comparison pixel becomes low. Accordingly, the above F1 to F16 are larger than the case where there is no spot defect 70.

In addition, as shown in FIG. 7, when some luminance comparison pixels are included in the spot defect 70 together with the target pixel, the average luminance value of the set having the luminance comparison pixels included in the spot defect 70 is It becomes larger than the luminance average value of the set having no luminance comparison pixel included in the spot defect 70, and the difference from the luminance value of the target pixel is reduced accordingly.
Similarly, as shown in FIG. 8, when the stripe defect 71 exists, the luminance average value = (S2 + S18) / 2 of the set having the luminance comparison pixels included in the stripe defect 71 is a stripe shape. Since the luminance value O of the target pixel included in the defect 71 is substantially the same, the difference therebetween becomes small.

  Therefore, as shown in FIG. 6, when the spot defect 70 exists in the area surrounded by the luminance comparison pixels, F1 to F16 are all relatively large values. On the other hand, when there is no spot defect, when there is a streak-like defect 71 as shown in FIG. 8, or when the spot defect 70 does not fit in the area surrounded by the luminance comparison pixels as shown in FIG. , At least one of F1 to F16 is a small value. Therefore, if a value having the minimum absolute value is selected from the values F1 to F16, it is determined whether or not there is a spot defect 70 that includes the target pixel and falls within the area surrounded by the luminance comparison pixels. Since it can be detected, the minimum value of the absolute value may be stored as the spot defect enhancement value of each target pixel.

  In the spot defect enhancement filter shown in FIG. 5, the distance from the target pixel to the vertical and horizontal luminance comparison pixels is six pixels, and the distance between the target pixel and the other luminance comparison pixels is also approximately six pixels. It is. The spot defect enhancement filter of the present embodiment is suitable for extracting a spot defect having an area that is slightly smaller than the area surrounded by the luminance comparison pixels. That is, when the area of the measurement target stain is larger than the area within the area surrounded by each luminance comparison pixel, the stain defect enhancement filter emphasizes only the outline portion and cannot emphasize the entire stain portion. . Accordingly, the size of the spot defect that can be detected differs depending on the distance between the target pixel and the luminance comparison pixel. For this reason, in the following description, in each spot defect emphasis filter, a pixel having a distance between the target pixel and the luminance comparison pixel of n is called a distance n pixel filter. For example, the spot defect enhancement filter shown in FIG. 5 is a 6-pixel distance filter.

In the filter with a distance of 6 pixels, the secondary difference processing is performed with 16 sets, but this number changes as the filter distance n changes.
By applying this process to the entire screen, it is possible to obtain an image in which a spot defect is emphasized. However, as described above, it is not possible to deal with various defect sizes with only one type of filter having a distance between the target pixel and the luminance comparison pixel. For example, the distance 6 pixel filter shown in FIG. 5 can detect only a spot defect having a diameter up to 11 diameters. For this reason, in the present embodiment, processing is performed by changing the distance between the target pixel and the luminance comparison pixel in several stages, and enhanced images corresponding to spot defects of various sizes are obtained.
Specifically, the spot defect emphasis process is performed using three filters of distance 3, distance 6, and distance 12 pixels. In other words, if a 12-pixel filter is used, it is possible to detect a spot defect having a diameter of 23. However, if the defect is sufficiently small compared to the filter, the distance from the pixel being compared with the target is long and an accurate result is obtained. Does not come out. For this reason, the distance 3 and 6 pixel filters are used in combination to improve the spot defect detection accuracy.
Since three stain-enhanced images can be obtained by this processing, the brightness of the processing result pixels at the same point in each image is finally compared, and the value having the maximum absolute value is used as the stain defect enhancement result. The image is synthesized as a single spot-enhanced image.

Here, the process of the spot defect emphasis process step S5 of the present embodiment will be described. In this embodiment, three types of spot defect enhancement filters are applied to an image reduced to ¼. Specifically, each spot defect emphasis filter of distance 3, 6, 12 pixel filter is applied. Then, the result of synthesis processing is obtained by synthesizing the application result of the spot defect enhancement filter of the distance 3, 6, 12 pixel filter.
Here, the plurality of spot defect emphasis filters having different distances are used for detecting spot defects of various sizes as described above. That is, if only the stain defect enhancement filter whose distance from the processing target pixel is 6 pixels is used, the size of the stain defect that can be detected is limited (only a stain defect in the vicinity of about 12 pixel size can be detected).
Therefore, in the present embodiment, in order to deal with spot defects of various sizes existing in the screen, the enhancement processing is performed by changing the distance in three steps, that is, three pixels, six pixels, and twelve pixels. Emphasizes spot defects.

  The reason why the enhancement results are combined is to improve detection accuracy. That is, when spot defects are extracted from each enhancement result and evaluated individually, one spot defect may be detected by being divided into enhancement results of 3 pixels and 6 pixels. May not match. Therefore, by combining a plurality of enhancement results to match the results, the separated spot defects are combined into one to improve detection accuracy.

Here, the synthesis processing is to compare the enhancement result values of the pixels at the same position in the enhancement results of each filter, and use the value having the maximum absolute value as the synthesis result.
When the distance of the spot defect emphasis filter is very large compared to the size of the spot defect, if the brightness comparison pixel is located in another spot area, the spot cannot be correctly detected due to the spot defect. For this reason, it is preferable to use a stain defect enhancement filter having a size slightly larger than the size of the stain defect to be detected, for example, a size of about 1 to 3 pixels.
Therefore, if the enhancement results of the filters of the distances 3, 6, and 12 pixels are synthesized, the spot defect corresponding to each size of the distances 3, 6, and 12 can be enhanced and detected.

  Next, the noise removing means 65 applies a median filter to the result of the spot defect enhancement processing step to connect and smooth the defect components separated by noise, and removes noise other than the spot defect. A removal process step (smoothing process: 3 × 3 median process) is performed (S6). As the median filter, for example, a median filter may be used in which the median value (median value) of each luminance value of 9 pixels of 3 × 3 is used as the luminance value of the target pixel located at the center of 3 × 3.

Next, the spot defect extraction processing unit 66 sets a threshold value for cutting out a white spot defect and a threshold value for cutting out a black spot defect with respect to the spot defect enhanced image from which noise has been removed, and cuts out areas for each spot defect candidate. A spot defect extraction process is performed (S7). Here, each threshold value may be set to an optimum value according to the situation of the image. For example, the average value of the spot emphasis value (luminance value) of the spot defect emphasis image (composite image) and its standard deviation are obtained, and the threshold value is set by the following equation.
White spot defect threshold wslevel = avr + α1 ・ σ + β1
Black spot defect threshold bslevel = avr−α2 ・ σ + β2
Here, avr is an average value of the composite image, σ is a standard deviation of the composite image, α1, α2, β1, and β2 are arbitrarily determined depending on the situation of the image to be inspected.

  Next, the blob processing means 67 carries out a blob processing step for obtaining the area of the region cut out as a defect candidate and the average luminance (contrast) for the white spot defect candidate image and the black spot defect candidate image (S8). .

On the other hand, the edge gradient detection means 69 performs an edge gradient detection processing step for detecting the edge gradient on the image extracted in the display area extraction step S3 (step S9).
Specifically, the edge gradient detection unit 69 includes a first edge gradient calculation unit 691, a second edge gradient calculation unit 692, a third edge gradient calculation unit 693, and an edge gradient synthesis unit 694, as shown in FIG. ing.
In the edge gradient detection processing step S9, the edge gradient detection unit 69 first selects a target pixel (S91). Specifically, the edge gradient detecting unit 69 sequentially selects target pixels one by one from all the pixels of the image extracted in the display area extracting step S3. At this time, only the pixel in the spot defect region extracted in the spot defect extraction processing step S7 may be set as the target pixel. However, as in this embodiment, when all the pixels of the image extracted in the display area extraction step S3 are targeted, the defect portion that cannot be detected in the stain defect enhancement processing step S5 or the stain defect extraction processing step S7 is also evaluated. This is preferable in that the accuracy of defective product detection can be improved.

When the target pixel is selected, the first edge gradient calculating unit 691 calculates an edge gradient value of one pixel using the edge gradient detection filter F1 shown in FIG. 11 (S92).
The edge gradient detection filter F1 uses the center pixel of the set detection region as a target pixel, and the first pixel is arranged in a substantially circular shape around the target pixel and adjacent to the target pixel, that is, separated by one pixel. As a comparison pixel, an edge gradient (luminance change rate) between the target pixel and each first comparison pixel is calculated, and the one having the maximum absolute value is set as the edge gradient value of one pixel distance in the target pixel. .

Next, the second edge gradient calculating means 692 calculates the edge gradient value of the distance 2 pixels using the edge gradient detection filter F2 shown in FIG. 12 (S93).
The edge gradient detection filter F2 is the same as the edge gradient detection filter F1, and the center pixel of the set detection area is set as a target pixel, is arranged in a substantially circular shape around the target pixel, and the first comparison pixel includes An adjacent pixel, that is, a pixel arranged approximately two pixels away from the target pixel is used as the second comparison pixel, and an edge gradient (luminance change rate) between the target pixel and each second comparison pixel is calculated, and the absolute value is the maximum. This is set as an edge gradient value of a distance of 2 pixels in the target pixel.

Next, the third edge gradient calculating unit 693 calculates an edge gradient value of a distance of 3 pixels using the edge gradient detection filter F3 shown in FIG. 13 (S94).
The edge gradient detection filter F3 is the same as the edge gradient detection filters F1 and F2, and the center pixel of the set detection region is set as a target pixel, is arranged in a substantially circular shape around the target pixel, and the second comparison A pixel adjacent to the pixel, that is, a pixel arranged approximately three pixels away from the target pixel is set as a third comparison pixel, and an edge gradient (luminance change rate) between the target pixel and each of the third comparison pixels is calculated, and an absolute value thereof is calculated. The maximum value is set as the edge gradient value of a distance of 3 pixels in the target pixel.

Here, the specific calculation procedure of the edge gradient value by each filter F1-F3 is demonstrated.
Each filter F1 to F3 subtracts the luminance value of each comparison pixel from the luminance value of the target pixel, divides the luminance difference by the length (distance) between the target pixel and the comparison pixel, and Edge gradient is calculated.
For example, in the edge gradient detection filter F2 of FIG. 12, the luminance value of the target pixel is “O”, the luminance values of the comparison pixels are “S1 to S12”, and the distance between the target pixel and each comparison pixel is D (O, S1). ˜D (O, S12), the edge gradients g1 to g12 are calculated using the following formulas 19 to 30.

g1 = (O−S1) / D (O, S1) (19)
g2 = (O−S2) / D (O, S2) (20)
g3 = (O−S3) / D (O, S3) (21)
g4 = (O−S4) / D (O, S4) (22)
g5 = (O−S5) / D (O, S5) (23)
g6 = (O−S6) / D (O, S6) (24)
g7 = (O−S7) / D (O, S7) (25)
g8 = (O−S8) / D (O, S8) (26)
g9 = (O−S9) / D (O, S9) (27)
g10 = (O−S10) / D (O, S10) (28)
g11 = (O−S11) / D (O, S11) (29)
g12 = (O−S12) / D (O, S12) (30)

Then, the calculated edge gradients g1 to g12 having the maximum absolute value are selected and set as the edge gradient value of the target pixel detected by the edge gradient detection filter F2. The other filters F1 and F3 perform the same processing to calculate each edge gradient value.
Here, the distance between the target pixel and the luminance comparison pixel can be calculated using an expression for obtaining a distance between two points on the XY coordinates. For example, when obtaining the distance D (O, S1), assuming that the coordinates of the target pixel are (X _O , Y _O ) and the coordinates of the comparison pixel are (X ₁ , Y ₁ ), the distance D (O, S1) is an expression. 31.

  When the third edge gradient calculating step S94 is completed, the edge gradient detecting unit 69 determines whether the selection of the inspection target pixel is completed (S95). Here, when the inspection has not been completed for the pixels of the entire image, each of the edge gradient calculation steps S91 to S93 is repeatedly executed.

On the other hand, when the inspection has been completed for the pixels of the entire image, the edge gradient synthesis unit 694 performs an edge gradient synthesis step (S96).
The edge gradient synthesis unit 694 synthesizes the detection results obtained by the edge gradient detection filters F1 to F3. Specifically, the maximum value of the edge gradient values of the edge gradient detection filters F1 to F3 in each target pixel is the combined result of the target pixel.
Thus, the edge gradient detection processing step S9 is completed.

Next, the defect rank classification processing means 68 performs a defect rank classification step of classifying the rank of the spot defect in the image based on the calculated area, average luminance, and edge gradient of the defect candidate area (S10).
Specifically, determination threshold curves TR1 to TR3 shown in FIG. 15 are defined in the defect rank classification processing unit 68, and are specified by the value of the area of the defect candidate region and the value of the average luminance (contrast). It is determined which area the point enters in the graph of the threshold curves TR1 to TR3 in FIG. 15, and it is determined as the non-defective product ranks 1 to 3 and NG (defective product). Specifically, the lower the contrast, the higher it is determined as rank 1 (the best product), and the higher the contrast, the higher it is determined as rank 2, rank 3, and defective products.

  In addition, the edge gradient detecting unit 69 adjusts the values of the determination threshold curves TR1 to TR3 from the gradient of the defect candidate. FIG. 15 is a determination threshold curve created using a defect having a certain inclination. As the inclination of the defect candidate becomes smaller, the determination threshold curves TR1 to TR3 become smaller in human judgment. Therefore, the determination threshold curve is adjusted by applying a coefficient (contrast threshold magnification) to the determination threshold curve of FIG. FIG. 14 is a value for adjusting the determination threshold curve TR3, and is a graph showing the contrast threshold magnification corresponding to the inclination of the defect candidate. The contrast threshold magnification is obtained from the inclination of the defect candidate, and the determination threshold curve is recalculated by applying a coefficient to the determination threshold curve TR3 in FIG. Concerning the determination threshold curves TR1 and TR2, the contrast threshold magnification is similarly obtained, the contrast threshold magnification is obtained from the inclination of the defect candidate, and the determination threshold curves TR1 and TR2 are recalculated. By determining the defective defective product rank from the newly obtained determination threshold curves TR1 to TR3, it is possible to determine the defective defective product rank in consideration of the inclination of the defect.

The defect rank obtained by the defect rank classification processing means 68 is displayed on the display device 7, and the inspector can easily grasp the defect rank of the liquid crystal panel 1 to be inspected.
Therefore, in the present embodiment, the spot defect detection process is configured by the spot defect extraction process S7, the blob process S8, the edge gradient detection process S9, and the defect rank classification process S10.

16 to 18 show specific examples of edge gradient detection processing according to the present invention and the conventional example.
FIG. 16 is a pseudo spot defect image to be processed, and a white spot defect is created in a pseudo manner. FIG. 17 is an enlarged view of the captured image of the spot defect portion in an unprocessed state.
FIG. 18 is a combination of the result of applying the edge gradient detection filter F1 having a distance of 1 pixel between the target pixel and the comparison pixel and the result of applying the edge gradient detection filter F2 having a distance of 2 pixels.

According to this embodiment, there are the following effects.
(1) Since the edge gradient detection means 69 uses the edge gradient detection filters F1 to F3, and the target pixel and the comparison pixel change the distance in three stages of 1 to 3 pixels, the edge gradient is calculated. It is possible to accurately detect the gradient of the edge portion of various spot defects such as an edge portion where the luminance changes rapidly and an edge portion where the luminance changes gently to some extent.
In particular, since the first edge gradient calculation means 691 for calculating the edge gradient value using the edge gradient detection filter F1 with a distance of 1 pixel is provided, edge portions having different gradients as shown in FIG. The gradient can be detected with high accuracy.
In addition, the first edge gradient calculation unit 691 can detect only the edge gradient in eight directions, but the second edge gradient calculation unit 692 that can detect the edge gradient in 12 directions and the third edge gradient calculation that can detect the edge gradient in 16 directions. Since the means 693 is also received, the detection accuracy of the edge gradient can be further improved.
Further, since the edge gradient synthesis means 694 synthesizes the calculation results of the edge gradient calculation means 691 to 693, it is possible to easily and accurately detect and evaluate edge gradients in various directions including high frequency components. .

(2) Since each of the edge gradient calculating units 691 to 693 obtains the distance between the pixels from the coordinates of the target pixel and the comparison pixel, the edge gradient can be detected with higher accuracy. That is, since each of the filters F1 to F3 is arranged in a matrix, a “variation” occurs in the distance between the center coordinates of each comparison pixel and the center coordinates of the target pixel. However, in the present embodiment, since the length between the coordinates is calculated, an accurate distance can be obtained, and the edge gradient calculated using this distance can be highly accurate.

(3) The defect rank classification processing unit 68 uses the three types of feature values of the defect candidate area and average luminance (contrast) obtained by the blob processing unit 67 and the edge gradient obtained by the edge gradient detection unit 69. Based on this, the rank of the liquid crystal panel 1 to be inspected is evaluated, so that rank evaluation close to the visual inspection result of the inspector can be automatically performed.

(4) The spot defect enhancement processing means 64 uses a spot defect enhancement filter that emphasizes a spot defect based on the luminance value of a luminance comparison pixel arranged at a predetermined distance from the target pixel and around the target image. Therefore, for example, as compared with the case where the luminance comparison pixels are arranged only in the upper, lower, left, and right directions with respect to the target pixel, it is possible to reliably detect a spot spreading in a planar shape.
In addition, the stain defect enhancement filter divides each luminance comparison pixel into a set of two luminance comparison pixels arranged at symmetrical positions across the target pixel, and the luminance value of the target pixel and 2 of each set. A difference value from the average luminance value of the two luminance comparison pixels is obtained, and among the difference values of each set, the one having the smallest absolute value is selected and used as the spot defect enhancement value. For this reason, when there is a streak-like defect or when a part of the brightness comparison pixel is included in the defect, the spot defect enhancement value becomes small and the spot defect is contained in the area surrounded by the brightness comparison pixel. The spot defect emphasis value becomes large only when it is, and by comparing the spot defect emphasis value with a predetermined threshold, streaky defects and the like can be eliminated, and only the spot defect can be detected with high accuracy.

(5) In this embodiment, the spot enhancement filter obtains an average value of the luminance values of two luminance comparison pixels arranged at point-symmetric positions with the target pixel in between and compares it with the target pixel. The effect of shading can be reduced, and spot defects can be detected with higher accuracy. That is, one of the two luminance comparison pixels arranged at a point-symmetrical position across the target pixel has a high luminance value due to the influence of background shading, and the other is often small. Therefore, the influence of shading can be smoothed by obtaining the average luminance value of these two luminance comparison pixels, and a spot defect can be detected with higher accuracy.

(6) Furthermore, since the spot defect enhancement filter can be set appropriately according to the distance between the target pixel and the luminance comparison pixel, the size of the spot defect that can be detected by the filter can be set. Can be easily detected.
Moreover, even when a plurality of spot defect emphasis filters corresponding to the size of each spot defect are applied, it is possible to easily detect a spot defect of each size by evaluating only the synthesis result of the enhancement results.

(7) In the present embodiment, the spot defect enhancement value is calculated from the luminance value of the target pixel and the average luminance value of the luminance comparison pixel, and the luminance value of the pixel between the target pixel and the luminance comparison pixel is used. Absent. For this reason, as long as the spot defect is smaller than the area surrounded by the luminance comparison pixels and includes the target pixel, even if the spot defect size changes to some extent, it is the same as the conventional visual inspection of the inspector. Spot defects can be detected. For this reason, it is not necessary to set the filter distance size very finely, and accordingly, the number of filters can be reduced and the inspection time can be shortened.

(8) Both the white spot defect and the black spot defect can be emphasized by the same spot defect enhancement filter, and the white spot defect processing unit 66 can use the white spot defect threshold and the black spot defect threshold separately. Both spot defects and black spot defects can be easily detected. For this reason, various stain defects can be efficiently detected by a simple process.

The present invention is not limited to the above embodiment.
For example, in the embodiment, the edge gradient is obtained using the three types of edge gradient detection filters F1 to F3. However, the edge gradient detection filter F1 having a distance of 1 pixel between the target pixel and the comparison pixel and the distance of 2 pixels are described. Alternatively, the detection may be performed using two types of filters, the edge gradient detection filter F2 and the edge gradient detection filter F3 having a distance of 3 pixels. In short, the edge gradient detection filter F1 at the distance 1 can be used to accurately detect the edge portion where the luminance change is abrupt, and the edge gradients in more directions can be obtained by using the edge gradient detection filters F2 and F3. The edge gradients in various directions including high frequency components can be detected with high accuracy by combining the results of the filters.
Note that an edge gradient detection filter having a distance of 4 pixels or more may be further combined, but usually the luminance gradient of the edge portion can be detected by a filter having a distance of 3 pixels or less. If the edge gradient detection filters F1 to F3 of three pixels are used, the edge gradient can be detected with high accuracy.

  In the embodiment, the edge gradient detection unit 69 calculates the edge gradient in the edge gradient detection processing step S9 using the entire image extracted in the display area extraction step S3 as the target pixel. The edge gradient may be calculated using a pixel included in the spot defect candidate extracted in 66 as a target pixel. In this way, the number of edge gradient calculation objects is reduced, and processing can be performed efficiently.

The defect rank classification processing unit 68 of the above embodiment performs the determination using the determination curve graph of FIG. 15, but the determination is performed using a three-dimensional determination graph having three axes of edge gradient, contrast, and area. You may go.
Further, the specific method for obtaining the contrast and the defect area is not limited to that of the above embodiment.

  In the embodiment, the edge gradient calculating means 691 to 693 obtains the edge gradient having the maximum absolute value for each of the edge gradient detection filters F1 to F3, and the edge gradient synthesizing means 694 calculates the absolute value among these three edge gradients. Although the maximum one is selected (synthesized), it is also possible to directly select the one having the maximum absolute value from all the edge gradients obtained by the filters F1 to F3 and obtain the result of synthesis.

  The calculation method of the edge gradient (luminance change rate) is not limited to the method of obtaining the luminance difference of each pixel and dividing the luminance difference by the distance between the pixels, as in the above-described embodiment. It may be used.

The detection target of the spot defect is not limited to the liquid crystal light valve using the TFT element as described above, but a liquid crystal panel, plasma display, organic EL display, DMD (direct mirror mirror) using other diode elements. It can be used for inspection of display body parts such as devices) and display devices and products using them, and it goes without saying that even if used for these, they are not excluded from the scope of the present invention. Absent.
Furthermore, the present invention is not limited to the inspection of various display devices. For example, when there are spot-like scratches on a printed matter, a case of a household electrical appliance, a car body, or the like, an image with a spot-like defect is picked up. If it is obtained, the stain can be detected, so that it can also be applied to stain inspection of various products such as surface coating and printed matter.
In short, the present invention can be used to detect a brightness spot defect or a color spot defect because it can detect any spot having a brightness difference from the surroundings.

  Furthermore, the edge gradient detection method and apparatus of the present invention can be widely used not only for detection of spot defects but also for the purpose of detecting edge portions having a relatively large luminance change rate in image processing and the like.

The block block diagram of the spot defect detection apparatus of the screen by embodiment of this invention. The flowchart for demonstrating operation | movement of the said spot defect detection apparatus. Explanatory drawing which shows the background image difference process of the same spot defect detection apparatus. Explanatory drawing which shows the display area extraction process of the same spot defect detection apparatus. The figure which shows the example of a spot defect emphasis filter. The figure which shows the example of the spot defect detected with a spot defect emphasis filter. The figure which shows the example of the spot defect which is not detected with a spot defect emphasis filter. The figure which shows the example of the stripe-shaped defect which is not detected with a spot defect emphasis filter. The block diagram which shows the structure of an edge gradient detection means. The flowchart explaining an edge gradient detection process process. The figure which shows the example of an edge gradient detection filter. The figure which shows the example of an edge gradient detection filter. The figure which shows the example of an edge gradient detection filter. The graph which shows the judgment threshold curve of a spot defect. The graph which shows the judgment threshold curve of a spot defect. The figure which shows the image which provided the pseudo stain defect. The enlarged view of the non-processed image of a spot defect. The figure which shows the state which synthesize | combined the detection defect of the edge gradient detection filter. The figure which shows the example of the conventional prebit filter. The figure which shows the example of calculation of the edge gradient at the time of using the conventional prebit filter.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Liquid crystal panel, 2 ... Projector, 3 ... Pattern generator, 4 ... Screen, 5 ... CCD camera, 6 ... Computer apparatus, 7 ... Display apparatus, 60 ... Image input means, 61 ... Background image difference processing means, 62 ... Display Area extraction means 63 ... Reduced image creation means 64 ... Blemish defect enhancement processing means 65 ... Noise removal means 66 ... Blemish defect extraction processing means 67 ... Blob processing means 68 68 Defect rank classification processing means 69 69 Edge Gradient detection means, 70 ... spot defects, 691 ... first edge gradient calculation means, 692 ... second edge gradient calculation means, 693 ... third edge gradient calculation means, 694 ... edge gradient synthesis means.

Claims (7)

A first edge gradient calculating step for obtaining a luminance change rate between the target pixel selected in the captured image and each of the first comparison pixels arranged around the target pixel and adjacent to the target pixel;
Second edge gradient calculation for obtaining a luminance change rate between the target pixel selected in the captured image and each second comparison pixel arranged around the target pixel and further away from the first comparison pixel than the target pixel Process,
An edge gradient synthesis step in which the absolute value of the luminance change rates calculated in the first and second edge gradient calculation steps is the edge gradient value of the target pixel;
An edge gradient detection method comprising: The edge gradient detection method according to claim 1,
In each of the edge gradient calculating steps, the brightness of the target pixel is O, the brightness of the nth comparison pixel is Sn, the coordinates of the target pixel are (X ₀ , Y ₀ ), and the coordinates of the comparison pixel are (X _n , Y _n ). When the distance between the target pixel and the nth comparison pixel is D (O, Sn),
The luminance change rate gn

The distance D (O, Sn) is

An edge gradient detection method characterized by: The edge gradient detection method according to claim 1 or 2,
Third edge gradient calculation for determining the luminance change rate between the target pixel selected in the captured image and each third comparison pixel arranged around the target pixel and further away from the second comparison pixel than the target pixel With a process,
The edge gradient synthesizing step uses, as the edge gradient value of the target pixel, the one having the maximum absolute value among the luminance change rates calculated in the first to third edge gradient calculation steps. Detection method. The edge gradient detection method according to claim 3,
The edge gradient detection method, wherein the second comparison pixel is disposed adjacent to the first comparison pixel, and the third comparison pixel is disposed adjacent to the periphery of the second comparison pixel. A spot defect area extracting step for extracting a spot defect area from an image obtained by imaging a spot defect inspection target;
The edge gradient detection method according to any one of claims 1 to 4 is executed for each pixel in the stain defect area, and an edge gradient value having a maximum absolute value among the edge gradient values of each target pixel is stained. An edge gradient detection step for setting an edge gradient value of the region;
An area calculating step for calculating an area of the spot defect region;
A contrast calculating step for obtaining a contrast of the spot defect region;
A defect rank evaluation step for evaluating the defect rank of the spot defect region based on the three types of feature values of the edge gradient, area, and contrast;
A spot defect detection method comprising: First edge gradient calculating means for obtaining a luminance change rate between the target pixel selected in the captured image and each of the first comparison pixels arranged around the target pixel and adjacent to the target pixel;
Second edge gradient calculation for obtaining a luminance change rate between the target pixel selected in the captured image and each second comparison pixel arranged around the target pixel and further away from the first comparison pixel than the target pixel Means,
An edge gradient synthesizing unit that sets an edge gradient value of the target pixel as the edge gradient value of the target pixel among the luminance change rates calculated by the first and second edge gradient calculating units;
An edge gradient detection apparatus comprising: A spot defect area extracting means for extracting a spot defect area from an image obtained by imaging a spot defect inspection object;
First edge gradient calculating means for respectively determining a luminance change rate between the target pixel selected in the spot defect area and each first comparison pixel arranged around the target pixel and adjacent to the target pixel;
A second edge gradient for determining a luminance change rate between the target pixel selected in the spot defect area and each second comparison pixel arranged around the target pixel and away from the first comparison pixel from the target pixel A calculation means;
An edge gradient synthesizing unit that uses an edge gradient value of a spot defect region as the edge gradient value of the absolute value among the luminance change rates calculated by the first and second edge gradient calculating units;
Area calculating means for calculating the area of the spot defect region;
Contrast calculation means for obtaining the contrast of the spot defect area;
Defect rank evaluation means for evaluating the defect rank of the spot defect area based on the three types of feature values of the edge gradient, area, and contrast;
A spot defect detection apparatus comprising:

Images