JP2001174417A - Apparatus and method for detecting defect - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an apparatus and a method capable of detecting a defect at each part of a reticle with severity corresponding to the change degree of a brightness value due to the pixel position of each part of the reticle and capable of detecting only a defect becoming a problem at a time of the exposure of a wafer at a high speed with a simple constitution. SOLUTION: The defect detection apparatus has a difference image forming part 9 for forming difference image data wherein the absolute value of the difference between the multivalue digital image data of the inspection region of a reticle 1 and the multivalue digital image data compared with the digital image data in order to detect a defect is set as the brightness value of each pixel, a region classifying part 11 for classifying the inspection region into a plurality of regions corresponding to the change degree of the brightness value due to a pixel position with respect to at least one of the multivalue digital image data of the inspection region and the comparing multivalue digital image data and a defect judging part 13 calculating the integrated value of the brightness value in the region of a predetermined size around each pixel of the difference image data to judge a defect based on an inspection parameter preset at every region wherein each pixel is classified.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for detecting a defect, and more particularly to an apparatus and a method for detecting a defect in a reticle (photomask) used in, for example, semiconductor manufacturing.

[0002]

2. Description of the Related Art A conventional defect detection apparatus of this kind is used for detecting minute defects of a photomask used in semiconductor manufacturing as disclosed in Japanese Patent Publication No. Hei 8-10463. Have been.

In this conventional defect detection apparatus, a left detector and a right detector located apart from each other respectively detect reflected light of light scanning two portions of a photomask to be compared, and detect the reflected light of the photomask. The data of each pixel of those parts is generated, the corresponding digitizer converts the data into multi-valued pixel data, and the corrector calculates the skew difference, enlargement difference, distortion difference and critical dimension between these data outputs by one pixel. The aligner calculates the alignment error between the left image and the right image with an accuracy of 1 pixel or less, stores the left and right pixel data, and moves one side by the alignment error. The sum of the four pixels after the subtraction was subtracted from the sum of the four pixels on the other side, and the absolute value of the difference was compared with a threshold to detect a defect.

[0004]

The defects to be detected by the defect inspection apparatus are only those which have an effect during wafer exposure and have a problem as a product in terms of throughput and yield. However, in the conventional defect inspection apparatus, if the detection level is made strict in order to detect a minute defect, even a level having no influence upon exposure is detected. In particular, in the case of a reticle, the luminance value of the surrounding pixels is almost constant over a wide range, such as a portion where the luminance value changes sharply depending on the position, such as the edge portion of the light shielding pattern, and a portion other than the edge portion of the light shielding pattern of the reticle. In the part that is
The defects to be detected are different, and the required detection accuracy is very strict except for the edge part, especially in the transmission part. However, in the conventional inspection method, since the entire reticle is inspected for defects at a uniform detection level, a level that should not be a defect at the edge portion of the light shielding pattern is detected.
In some cases, a defect to be detected cannot be detected in a portion other than the edge portion of the light-shielding pattern.

According to the present invention, in each portion of the reticle, a defect in each portion of the reticle can be detected with a severity corresponding to the degree of change in the luminance value depending on the pixel position of that portion, and only a defect having a problem during wafer exposure can be detected in a simple configuration. It is an object of the present invention to provide a defect detection device and method capable of detecting defects at high speed.

[0006]

According to the present invention, there is provided a defect detection apparatus comprising: a first multi-valued digital image data of an inspection area to be detected; and a second digital image data for comparing the first digital image data for defect detection. A difference image generating unit that generates difference image data in which the absolute value of the difference between the multi-value digital image data is the luminance value of each pixel, the first multi-value digital image data and the second multi-value digital image data An area classifying unit that classifies the inspection area into a plurality of areas according to a change in luminance value depending on a pixel position for at least one of
Determine the integral value of the luminance value in a region of a predetermined size around each pixel of the difference image data, and determine whether or not there is a defect based on an inspection parameter preset for each region into which each pixel is classified. A defect determining unit for determining whether

[0007] According to this configuration, the inspection area is classified into a plurality of types according to the change of the luminance value depending on the pixel position.
The size of the defect to be detected for each is changed, and the inspection is performed at an appropriate inspection level. Therefore, it is possible to detect only a defect that has an effect during wafer exposure and has a problem as a product.

In the above structure, the area classifying section sequentially obtains the amount of change in the luminance value in a predetermined first range area around each pixel by filter operation processing, and obtains the luminance value according to the pixel position. And the defect determination unit calculates the integral value of the luminance value in the predetermined second range of the difference image data for each region classified around each pixel by the filter calculation process. May be requested.

According to this configuration, the process of classifying the inspection area and determining the defect uses the filtering process. Therefore, according to the present invention, a defect can be detected at a high speed with a simple configuration.

Also, upon receiving a classification number designation input indicating how many types of inspection regions are classified at the time of inspection, the region classification unit is provided with a classification filter of a number corresponding to the classification number designation input and a threshold for region classification. The apparatus may further include a classification number setting unit configured to set and set a threshold value for defect determination to the defect determination unit in accordance with the classification number designation input. According to this configuration, the defect inspection can be performed with an accuracy corresponding to the required defect detection level, and the defect inspection can be performed efficiently or with high accuracy.

[0011]

Next, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of an embodiment of a defect detection apparatus according to the present invention.

The defect detection apparatus according to the present invention includes an inspection stage 2 for holding and moving a reticle 1, an illumination unit 3 for illuminating the reticle 1, and an imaging unit 4 for capturing an image of an inspection area of the reticle 1 and outputting an image signal. An A / D converter 5 that converts an image signal from the imaging unit 4 into multi-valued digital image data;
Inspection image memory 6 for storing multi-valued digital image data of the inspection area of reticle 1, and comparison for storing multi-valued digital image data to be compared with image data stored in inspection image memory 6 for defect detection. The image memory 7 and the two images stored in the inspection image memory 6 and the comparison image memory 7 are aligned in pixel units or less, and the inspection image memory 6
And a registration unit 8 stored in the comparison image memory 7, and a difference using the absolute value of the difference between the data of the pixels at the same position in the two images stored in the post-alignment inspection image memory 6 and the comparison image memory 7 as the luminance value. A difference image generation unit 9 for generating image data, a difference image memory 10 for storing the difference image data generated by the difference image generation unit 9, and an inspection image memory 6
And an area for performing a process of sequentially obtaining an amount of change in the luminance value in an area of a predetermined size of the image data stored in the comparison image memory 7 and classifying the inspection area into a plurality of areas according to the amount of change. A classification unit 11, a classification result memory 12 for storing the classification result, and an integrated value of the luminance value in a region of a predetermined size of the difference image stored in the difference image memory 10, and for each region to be classified A defect determination unit 13 for determining a defect based on inspection parameters set in advance, and a result output unit 14 for outputting an inspection result to a screen
And

Further, as shown in FIG. 1, an image of an inspection object actually taken from a reticle is compared with a pseudo image generated from corresponding design data to detect a defect. And a pseudo image generation unit 15 for generating digital pseudo image data of the above and storing the digital pseudo image data in the comparison image memory 7.

In addition to the control of the movement of the inspection stage 2 and the control of the irradiation light of the illumination unit 3, an imaging control unit 16 for controlling the imaging of the imaging unit 4 in synchronization with these controls.
Is provided.

Also, as shown in FIG. 1, an input for designating the number of classifications into which inspection areas are to be classified at the time of inspection is given to the area classification unit 11 for the number of classification filters and the number of area classifications corresponding to the input. And a classification number setting unit 17 for setting a threshold value for the defect determination unit 13 and setting an integration filter and a threshold value for the defect determination for the defect determination unit 13 and performing defect inspection by classifying the number of inspection areas according to the input. It may be.

The illumination unit 3 irradiates a laser beam from the upper surface of the reticle, for example, and the imaging unit 4 receives a transmitted light of the reticle and outputs an image signal of a reticle inspection area, for example. Good.

The area classifying unit 11 classifies the classification filters F1, F2,... Fn of n different sizes (n is 2 or more) based on the two images stored in the inspection image memory 6 and the comparison image memory 7, respectively. Is used to determine the amount of change in the luminance value within the range of the classification filter around each pixel, and if the amount of change is equal to or less than a predetermined threshold value for each filter, it is classified into a corresponding region, and the remaining pixels are assigned to the (n + 1) th pixel. The inspection area is classified into any of the (n + 1) areas by classifying the inspection area. It is assumed that the first classification filter F1 has the largest size and the size decreases in order up to the classification filter Fn. The minimum n-th classification filter Fn is 3 pixels × 3 pixels.
Pixels or larger size. Further, in the first area corresponding to the first classification filter F1, pixels whose luminance values of surrounding pixels have a gradual change are classified, and the luminance value of surrounding pixels rapidly changes in order up to the (n + 1) th area. Pixels are classified.

In this embodiment, first, scanning is performed by the classification filter F1 to determine whether each pixel belongs to the first area, and then the classification filter F1 is applied to each pixel that does not belong to the first area.
2 to determine whether each pixel belongs to the second area, and repeat this until the determination by the n-th classification filter to determine whether the pixel belongs to the n-th area.
Pixels that do not belong to any of the regions from n to n are classified into the (n + 1) th region.

The classification of the k-th (k = 1 to n) region will be described in detail. Pixels which are not classified in any of the first to (k-1) -th regions are sequentially sorted by the k-th classification filter. Fk is Wk with the pixel of interest (x, y) as the center.
A luminance value d (x + x) of an area of pixels × Wk pixels (W1 is an odd number)
i, y + j) (where i, j = − (Wk−1) / 2− (W
k-1) / 2) is input, and the variation ak (x, y) of the luminance value in this area is calculated and output based on the following equation.

Ak (x, y) = max ｛d (x + i, y
+ J) {-min {d (x + i, y + j)} (where i, j =-(Wk-1) / 2 to (Wk-1) /
2) max {d (x + i, y + j)} indicates the maximum value of the luminance value d (x + i, y + j) in the k-th classification filter Fk, and min {d (x + i, y + j)} indicates the k-th classification filter Fk. Luminance value d (x + i, y) in the classification filter Fk of
+ J).

The area classifying unit 11 classifies the target pixel (x, y) of the image data stored in the inspection image memory 6 and the comparison image memory 7 with the classification filter F
The amount of change in the luminance value due to k is calculated. If one of the two amounts of change is smaller than a predetermined threshold Ak corresponding to the k-th region, the pixel of interest (x, y) is set to the k-th region. Classify into The k-th threshold Ak is set to be smaller than the k + 1-th threshold or equal to the k + 1-th threshold.

This classification method will be described in detail using a specific example of image data. FIG. 2 shows the inspection image memory 6.
FIG. 3 shows an example of image data of the inspection area stored in the comparison image memory 7. FIG. 3 shows a classification result generated by the area classification unit 11 and stored in the classification result memory 12 for the image data example of FIG. FIG. Note that the first classification filter F1 has a size of 5 pixels × 5 pixels, the second classification filter F2 has a size of 3 pixels × 3 pixels, and an example in which all pixels are classified into the first to third regions will be described. I do.

As shown in FIG. 2, the image data is, for example, 8 bits, that is, digital image data of 256 gradations. A numerical value representing the luminance value of each pixel corresponding to the actual position on the reticle 1 is mapped. And stored in the inspection image memory 6 and the comparison image memory 7. In the classification result, a numerical value representing the classification result at each pixel is mapped to correspond to the actual position on the reticle 1, and the classification result memory 12
It is recorded in.

For example, as shown in FIG. 3, a numerical value "1" is recorded for a pixel belonging to the first area, "2" is recorded for a pixel belonging to the second area, and a third A numerical value “3” is recorded for the pixels belonging to the area.

When classifying the pixels 21 shown in FIG. 2,
First, the maximum value max {d (x + i, y + j)} in the input range 22 of the operation by the classification filter F1 is 255 and the minimum value min {d (x + i, y + j)} is 200, so that the classification filter F1 Fluctuation amount of luminance value a1
(X, y) becomes 55. If the threshold value A1 corresponding to the first area is, for example, 60, the threshold value is smaller than the threshold value. Therefore, as shown in FIG. 3, the classification result 26 of the pixel 21 is classified into the first area in which the fluctuation of the luminance value is the slowest. It becomes "1" which indicates that.

When the pixels 23 in FIG. 2 are classified, the maximum value max {d (x + i, y + j)} within the input range 24 of the operation by the classification filter F1 is 255, and the minimum value min {d ( x + i, y + j)} is 32, so that the variation amount a1 of the luminance value due to the classification filter F1
(X, y) becomes 223, and since the change in the surrounding luminance value is large, the pixel 21 is not classified into the first area because it is larger than the threshold value A1, and is subjected to the classification processing using the classification filter F2. . The classification filter F2 is smaller in size than the classification filter F1, and has a threshold A2 corresponding to the second area.
Is set to 100, for example, and in the second area,
Those in which the change in the luminance value of the surrounding pixels is sharper than the first area are classified.

The classification filter F2 for the pixel 23 in FIG.
The maximum value max ｛d (x +
i, y + j)} is 255, and the minimum value min {d (x
+ I, y + j)} is 128, so the classification filter F
The variation amount a1 (x, y) of the luminance value due to 2 is 127, which is larger than 100, which is the threshold value A2 corresponding to the second area, and is therefore not classified into the second area. Therefore,
The classification result 27 of the pixel 23 becomes “3” indicating that the change of the luminance value in the surrounding pixels is classified into the third region which is steeper than the second region as shown in FIG.

As shown in FIGS. 2 and 3, the portion where the luminance value changes depending on the position, for example, the edge portion of the light-shielding pattern is classified into a region "3", and the luminance value of the surrounding pixels is substantially constant over a wide range. , For example, a light-shielding portion or a transmissive portion other than the vicinity of the edge portion of the light-shielding pattern of the reticle is classified into the area “1”.

Next, the defect judging unit 13 shown in FIG. 1 will be specifically described. The defect determination unit 13 uses an integration filter that is one size smaller than the classification filter used when classifying each pixel, and calculates the luminance of the difference image stored in the difference image memory 10 for an area in the filter centered on each pixel. Integrate the values. Further, it is determined whether or not the pixel is defective based on the inspection parameters set for each of the previously classified areas.

For example, for the pixels classified into the k-th area, if the classification filter Fk is Wk pixels × Wk pixels, the defect determination unit 13 determines that the size is smaller than the classification filter Fk by a size Uk pixel × Uk pixel (Uk pixel). The integral value of the difference image data is output using the integral filter Gk in the region of <Wk and Uk are odd numbers). For the pixels classified into the n-th and (n + 1) -th regions, the integration value of the difference image data is calculated using an integration filter Gn having a smaller size than the smallest classification filter Fn used for determining the classification at the end. Output.

The integration filter Gk calculates the pixel of interest (x, y)
, The luminance value sub (x + i, y + j) of the difference image in the area of Uk pixels × Uk pixels (where i, j = − (Uk
−1) / 2 to (Uk−1) / 2) as inputs, calculate the integral value of the input Uk × Uk pixels, and calculate the pixel of interest (x,
It is output as an integrated value sk (x, y) for y).

When the integration filter Gn performs integration of one pixel, the integration filter F4 outputs the luminance value sub of the difference image of the pixel of interest (x, y) belonging to the nth and (n + 1) th regions. (X, y) is used as the integral value sn
(X, y) and an integrated value sn + 1 (x, y).

The defect judging section 13 determines whether the integration value sk (x, y) of the pixel of interest (x, y) belonging to the k-th region is predetermined by the integration filter Gk corresponding to the k-th region. If it is larger than the parameter Bk, the pixel of interest (x, y) is determined to be defective.

The inspection parameter Bk is set to a small value in a region where the luminance value changes gradually, and is set to a large value in a region where the luminance value changes rapidly. This makes the defect detection level stricter in a portion where the luminance value of surrounding pixels is almost constant over a wide range, such as the central portion of the light-shielding pattern of the reticle or the central portion of the transmissive portion, and makes the inspection image and the comparison image Only a small difference in luminance value is allowed, and a required severe defect detection level can be satisfied. In areas where the luminance value changes abruptly depending on the position, such as the edge part of the light-shielding pattern, the defect detection level is relaxed to allow a large difference between the luminance value of the inspection image and the comparative image, and that there is no problem during exposure. It can be prevented from being detected as a defect.

Next, the operation of defect detection according to this embodiment will be described in detail.

FIG. 4 is a flowchart showing the operation of capturing an image of the inspection target portion and the comparison target portion of the reticle of FIG. 1 and generating a difference image thereof.

First, the reticle 1 set on the inspection stage 2 for inspecting the inspection target portion in the inspection area of the reticle is positioned (step S1). Next, the lighting unit 3
Irradiates the inspection target portion of the inspection region of the reticle with laser light from the upper surface of the reticle, receives the transmitted light by the light receiving element which is the imaging section 4, and images the inspection target portion, and the A / D converter 5 The data is converted into multi-value digital image data (step S2). At this time, the inspection stage 2 is controlled so that the image to be captured is not blurred. Next, the inspection stage 2 is moved to an imaging position for capturing an image of a comparison target portion to be compared with the image captured in step S2 (step S3). Next, an image of the comparison target portion to be compared with the image captured in step S2 is captured at the position after the movement (step S4). The image captured in step S2 is stored in the inspection image memory 6, and the image captured in step S4 is stored in the comparison image memory 7 (step S5). In the case of the die-to-die method in which the same pattern on the reticle is compared with each other, the image acquisition as described above is performed. However, the image actually obtained from the reticle and the corresponding image are obtained as in the die-to-database method. When comparing with the pseudo image generated from the design data, step S3 is not performed after step S2, and in step S4, a pseudo image is generated from the design data, and in step S5, the pseudo image data is compared with the comparison image. It is stored in the memory 7.

The two images stored in the inspection image memory 6 and the comparison image memory 7 are registered in the unit of a pixel or less and stored in the inspection image memory 6 and the comparison image memory 7 (step S6). Then, a difference image is generated using the absolute value of the difference between the two images stored in the comparison image memory 7 as the luminance value, and stored in the difference image memory 10 (step S7).

Next, the operation of classifying the images stored in the inspection image memory 6 and the comparison image memory 7 will be described. FIG. 5 is a flowchart showing the operation of classifying the inspection target portion in FIG.

First, an area to be classified is set as a first area (k = 1) (step S10). Next, a k-th classification filter Fk and a threshold Ak are set (step S1).
1). Next, the position of the k-th classification filter Fk is set at the position of the first pixel in a range that is not classified into any of the first to k-th regions (step S12). Next, with respect to the images stored in the inspection image memory 6 and the comparison image memory 7, the amount of change in the luminance value, that is, the difference between the maximum luminance value and the minimum luminance value is obtained by the calculation using the k-th classification filter Fk (step S13). ). It is determined whether any one of the fluctuation amounts is smaller than a k-th threshold Ak predetermined corresponding to the k-th classification filter Fk (step S1).
4), the fluctuation amount in one of the images is the k-th threshold Ak
If it is smaller, the portion is classified into the k-th region in both images (step S15), and the amount of change in both is k-th threshold A
If it is larger than k, the process proceeds to the next step without assigning a classification.

In step S13, for the image stored in one of the inspection image memory 6 and the comparison image memory 7, the amount of change in the luminance value is determined in step S14, and the amount of change is smaller than the k-th threshold Ak. May be determined. In this case, the classification processing speed is doubled, and the classification processing can be performed efficiently.

Next, it is determined whether the scanning by the k-th classification filter Fk has been completed (step S16). If the scanning has not been completed, the range not classified into any of the first to (k-1) -th areas is determined. The position of the classification filter Fk is moved to the position of the next pixel (step S17), and the process returns to step S13. When scanning by the k-th classification filter Fk is completed,
It is determined whether the scanning of the n-th classification filter Fn has been completed (k = n) (step S18). If the scanning of the n-th classification filter Fn has not been completed, the area to be classified is set to the (k + 1) -th area (step S18). S19), returning to step s11. When the scanning of the n-th classification filter Fn is completed, the pixels in the range where the classification is not determined are classified into the (n + 1) -th area (step s20). Finally, the classification result indicating the range of the area classified into n + 1 areas is stored in the classification result memory 12 (step S21).

The above steps S10 to S21
Is a case of the die-to-die system, and in the case of the die-to-database system, these processes need only be performed on the pseudo image stored in the comparative image memory 7. That is, step S13
In step (1), only for the image stored in the comparison image memory 7, the amount of change in the luminance value in the input area of the classification filter Fk is determined. It may be determined whether the amount is smaller than a k-th threshold Ak predetermined for the k-th classification filter Fk.

Next, the operation of detecting a defect based on the classification result will be described. FIG. 6 is a flowchart showing the operation of the defect detection of FIG.

With respect to the difference image stored in the difference image memory 10, the first pixel to be subjected to defect judgment is set (step S30), and the area to which the pixel belongs is checked with reference to the classification result memory 12 (step S30). Step S31),
According to the classification result, if the region is the k-th region, an integral filter Gk smaller than the classification filter Fk used at the time of region determination by one turn is set at the position of the pixel (step S32).
The integral value sk is calculated by integrating the luminance value of the difference image stored in the difference image memory 10 in the integration filter Gk (step S33). Next, an inspection parameter Bk corresponding to the classification result is selected (step S34), and it is determined whether the integration value sk calculated in step S33 is larger than the inspection parameter Bk selected in step S34 (step S35). The pixel is determined to be defective (step S36), and if smaller, the pixel is determined to be normal (step S37). It is determined whether the defect determination has been completed for all the pixels of the difference image (step S38). If the determination has not been completed, the next pixel to be subjected to defect determination is set (step S3).
9) Return to step S31. When the determination of the defect is completed for all the pixels of the difference image, the inspection result, for example, the detected defective portion is output to the result output unit 14 (Step S).
40).

The processes in steps S30 to S40 can be performed in exactly the same manner whether in the die-to-die system or the die-to-database system.

[0048]

As described above, according to the defect detection method and apparatus of the present invention, the inspection areas are classified into a plurality of types according to the degree of change in the luminance value depending on the pixel position, and each of the inspection areas has a defect to be detected. Change the size and perform the inspection at an appropriate inspection level. Therefore, it is possible to detect only a defect that has an effect during wafer exposure and has a problem as a product.

Further, since the processing of the inspection area classification and the defect inspection is only the filter processing, simple and high-speed processing is possible.

Also, regarding the number of automatic classification of the inspection area,
The operator can also make a selection, and can perform the defect inspection with an accuracy that matches the required defect detection level, and can perform the defect inspection efficiently or with high accuracy.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an embodiment of a defect detection device according to the present invention.

FIG. 2 is a diagram showing an example of image data of an inspection area according to the present invention.

FIG. 3 is a diagram showing a classification result generated for the image data example of FIG. 2;

FIG. 4 is a flowchart showing an operation of capturing an image of a portion to be inspected and a portion to be compared of the reticle of FIG. 1 and generating a difference image thereof.

FIG. 5 is a flowchart illustrating an operation of classifying the inspection target portion in FIG. 1;

FIG. 6 is a flowchart showing an operation of the defect detection of FIG. 1;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Reticle 2 Inspection stage 3 Illumination part 4 Imaging part 5 A / D conversion part 6 Inspection image memory 7 Comparative image memory 8 Alignment part 9 Difference image generation part 10 Difference image memory 11 Area classification part 12 Classification result memory 13 Defect judgment part 14 Result output unit 15 Pseudo image generation unit 16 Imaging control unit 17 Classification number setting unit

Continued on the front page (51) Int.Cl. ⁷ Identification code FI Theme coat II (reference) 9A001 F term (reference) 2F065 AA11 AA49 BB27 CC18 DD00 DD02 DD06 FF04 GG04 JJ03 JJ26 PP12 QQ03 QQ08 QQ13 QQ14 QQ24 QQ25 QQ29 QQ33 Q38 SS11 UU05 2G051 AA56 AB02 BA10 DA05 EA08 EA09 EA11 EA14 EA30 EB01 EB02 EC01 EC05 4M106 AA09 CA39 DB04 DB08 DB19 DJ13 DJ14 DJ18 DJ21 DJ26 5B057 AA03 BA02 BA29 CA12 CB12 CC02 CE55 DA03 DC22 DC33 5G09 BG03 GA03 9G HH27 JJ49 KK37 KK54

Claims (8)

[Claims] An absolute value of a difference between first multi-valued digital image data of an inspection area to be defect-detected and second multi-valued digital image data to be compared with the first digital image data for defect detection. A difference image generating unit that generates difference image data having a brightness value of each pixel; and a change in brightness value according to a pixel position for at least one of the first multi-value digital image data and the second multi-value digital image data. An area classification unit that classifies the inspection area into a plurality of areas according to speed, obtains an integral value of a luminance value within a predetermined range around each pixel of the difference image data, and the integral value is classified. A defect determining unit that determines that each of the pixels is defective when the inspection parameter is smaller than a predetermined inspection parameter for each of the regions. 2. The area classifying section sequentially obtains a variation amount of a luminance value within a predetermined first range around each pixel by a filter operation process, and obtains a change of the luminance value depending on a pixel position. 2. The method according to claim 1, wherein the defect determination unit obtains an integrated value of luminance values in a predetermined second range around each pixel by a filter operation process.
The defect detection device according to the above. 3. The defect detection apparatus according to claim 1, wherein the inspection parameter is set to be smallest in an area where a change in luminance value is gentlest. 4. The area classifying unit decreases the size in the order of first to n-th (n is a positive integer) and uses the order to calculate the variation of the luminance value around each pixel. And a second filter for classifying each pixel into the first to n-th regions when the amount of change is smaller than the amount of change calculated using the first to n-th classifying filters. 2. The defect detection method according to claim 1, further comprising a first to an n-th threshold, wherein a k-th threshold (k is an integer equal to or less than n-1) is smaller than or equal to a k + 1-th threshold. 5. A classification number designation input indicating how many types of inspection regions are classified at the time of inspection, and a classification filter and a threshold value for region classification corresponding to the classification number designation input are provided to the region classification unit. 2. The defect detection apparatus according to claim 1, further comprising a classification number setting unit configured to set and set a threshold value for defect judgment to the defect determination unit in accordance with the classification number designation input. apparatus. 6. An absolute difference between the first multi-value digital image data of the inspection area to be detected and the second multi-value digital image data to be compared with the first multi-value digital image data for defect detection. A first step of generating difference image data whose value is a luminance value of each pixel; and a luminance around each pixel for at least one of the first multi-value digital image data and the second multi-value digital image data. A second step of classifying the inspection area into a plurality of areas in accordance with the rate of change of the value, obtaining an integral value of a luminance value in a predetermined range around each pixel of the difference image data, A third step of determining that each pixel is defective if the inspection parameter is smaller than a predetermined inspection parameter for each area. 7. The method according to claim 1, wherein the second step is sequentially performed from a first area to an n-th area (n is a positive integer).
A fourth step of obtaining a variation amount of a luminance value in a predetermined k-th (k is a positive integer equal to or less than n) classification filter around each pixel for at least one of the digital image data of If the amount of change is smaller than a k-th preset value corresponding to the k-th classification filter, repeating the fifth step of classifying into the k-th area; 7. The defect detection method according to claim 6, further comprising the step of classifying the pixels in the range in which is not determined into the (n + 1) th region. 8. The method according to claim 1, wherein the third step includes the first and second steps.
Integrating, with respect to the difference image data of the digital image data, a luminance value within a range of a size smaller than the predetermined k-th classification filter corresponding to the k-th region where each pixel is classified; 7. The method according to claim 6, further comprising the step of: if the integrated value is larger than a predetermined k-th inspection parameter corresponding to the k-th area where each pixel is classified, making the pixel defective. Detection method.