JP2001099625A - Device and method for pattern inspection - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and a method for pattern inspection which can efficiently perform accurate defect detection. SOLUTION: Of the pattern inspection device 1, a comparing inspection part 10 generates a defect candidate area D in an inspected image S by comparing the inspected image S with a reference image R. Further, a mask 21 generates a masked image MS wherein the inspected image S is masked with a defect candidate image Q, and a feature extraction part 31 extracts features from the masked image MS. Similarly, a mask 22 generates a masked image MR wherein the reference image R is masked with the defective candidate image Q and a feature extraction part 32 extracts features from the masked image MR. A decision part 40 decides a defect according to the feature quantities CS and CR extracted by the feature extraction parts 31 and 32. As those feature quantities CS and CR, mean density, a process result of an autocorrelation function, etc., are used.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a pattern inspection apparatus and a pattern inspection method applicable to pattern defect inspection on a color filter, a shadow mask, a printed wiring board, a semiconductor wafer, and the like.

[0002]

2. Description of the Related Art As a technique for performing a pattern defect inspection on an inspection object such as a semiconductor wafer having a predetermined pattern, for example, there is a technique described in Japanese Patent Publication No. Hei 6-21769. According to this technique, when comparing both the binarized image to be inspected and the reference image,
An exclusive OR of the two images is used to determine a non-coincidence portion between the two images. As a result, if the non-coincidence portion is detected as a lump having a predetermined size or more, it is determined as a defect. When both the image to be inspected and the reference image are multi-valued images, for example, when the absolute value of the difference between the image to be inspected and the reference image is larger than a predetermined threshold, the portion is determined as a defect. .

[0003]

However, in the above-described technique, the detection accuracy is a threshold for binarizing the original multi-valued image in the case of a binary image, and a defect in the case of a multi-valued image. It largely depends on how much the threshold value for detecting the portion is set. Here, when the threshold value is lowered so as to increase the defect detection rate, the rate of detecting so-called pseudo defects other than true defects also increases, and conversely, when the threshold value is increased so as to reduce the detection of pseudo defects, the true defect More likely to miss. As described above, in the above technique, the degree of dependence on the binarization based on the threshold value is large, and it is extremely difficult to separate and extract only true defect portions, that is, to perform accurate defect detection. There's a problem.

Further, in order to perform accurate defect detection, it is conceivable to conduct a detailed study using multi-valued gradation values. However, such processing generally requires a very large processing amount. There is a problem that it is inefficient.

In view of the above problems, an object of the present invention is to provide a pattern inspection apparatus and a pattern inspection method capable of efficiently performing accurate defect detection.

[0006]

According to a first aspect of the present invention, there is provided a pattern inspection apparatus for performing a defect inspection of a pattern by image processing. A defect candidate image generating means for generating a defect candidate image for specifying a defect candidate area in the inspection image by comparing with a reference image, and masking the defect image other than the defect candidate area in the inspection image using the defect candidate image A first masking unit, a first feature extracting unit for performing feature extraction on the masked image to be inspected, and a pattern defect of the inspected image based on the first feature amount extracted by the first feature extracting unit. Determining means for performing the determination.

According to a second aspect of the present invention, there is provided the pattern inspection apparatus according to the first aspect, wherein the defect candidate image generating means is configured to: (a) correspond to each other in the inspected image and the reference image; Pixel value matching means for determining the matching of the pixel values according to a predetermined criterion, (b) Among the pixels whose pixel values do not match each other according to the criterion,
A defect candidate area specifying unit that specifies a defect candidate area by grouping spatially continuous ones and specifies the defect candidate image as a set of the defect candidate areas;
A feature value is extracted from the masked image to be inspected for each of the defect candidate areas, and the determining unit determines the defect by comparing the first feature value with a predetermined threshold value.

According to a third aspect of the present invention, there is provided the pattern inspection apparatus according to the first or second aspect, wherein the reference image uses the defect candidate image to mask an area other than the defect candidate area. 2 mask means;
A second feature extraction in the masked reference image
And a feature extracting unit, wherein the determining unit determines a pattern defect of the image to be inspected based on the second feature amount extracted by the second feature extracting unit.

According to a fourth aspect of the present invention, in the pattern inspecting apparatus according to the third aspect, the defect candidate image generating means includes: (a) the defect candidate image generating means corresponding to each other in the inspected image and the reference image; Pixel value matching means for determining the matching of the pixel values according to a predetermined criterion, (b) Among the pixels whose pixel values do not match each other according to the criterion,
A defect candidate area specifying unit that specifies a defect candidate area by grouping spatially continuous ones and specifies the defect candidate image as a set of the defect candidate areas;
The feature amount is extracted from the masked image to be inspected for each of the defect candidate areas, the second feature amount is extracted from the reference image for each of the defect candidate areas, and the determining unit includes: The amount and the second feature amount are compared with each other to determine a defect.

According to a fifth aspect of the present invention, in the pattern inspection apparatus according to any one of the first to fourth aspects, the pattern inspection apparatus further includes a third feature extracting means for extracting a feature from the defect candidate image. , The determining means comprises:
It is characterized in that the pattern defect of the image to be inspected is determined on the basis of the third feature quantity extracted by the third feature extracting means.

According to a sixth aspect of the present invention, in the pattern inspection apparatus according to the fifth aspect, the defect candidate image generating means includes: (a) a plurality of defect candidate images corresponding to each other in the inspection image and the reference image; Pixel value matching means for determining the matching of the pixel values according to a predetermined criterion, (b) Among the pixels whose pixel values do not match each other according to the criterion,
A defect candidate area specifying unit that specifies a defect candidate area by grouping spatially continuous ones and specifies the defect candidate image as a set of the defect candidate areas;
The feature amount is extracted from the masked inspection image for each of the defect candidate areas, the second feature amount is extracted from the reference image for each of the defect candidate areas, and the third feature amount is Classifying means calculated for each candidate area, the determining means comparing the third feature value with a predetermined threshold, and classifying the masked image to be inspected into an image to be inspected and an image to be inspected, Selection of comparing the first feature amount and the second feature amount with respect to an area of the inspection target image only for the inspection target image among the inspection target image and the non-inspection target image, and selecting a defect. Characteristic determination means.

A pattern inspection method according to a seventh aspect is characterized in that an inspection image including a pattern is inspected by a feature extraction method only in a range selected by screening by a comparative inspection.

In the pattern inspection method according to the present invention, a comparison inspection for comparing each pixel value of an image to be inspected including a pattern with each pixel value of a reference image is performed. From the image to be inspected, a first feature amount of each selected region of the image to be inspected selected while excluding an area whose pixel value matches the reference image by the comparative inspection is obtained, and based on the first feature amount. And determining a pattern defect of the image to be inspected.

According to a ninth aspect of the present invention, in the pattern inspection method according to the eighth aspect, a second feature amount of the reference image is obtained for each of the selected areas, and the first feature is obtained for each of the selected areas. A pattern defect of the image to be inspected is determined by comparing the amount with the second characteristic amount.

According to a tenth aspect of the present invention, there is provided a pattern inspection method comprising:
10. The pattern inspection method according to claim 8, wherein a third feature amount is obtained from each of the selected regions, and a region where the third feature amount does not satisfy a predetermined criterion is excluded for each of the selected regions. And comparing the first feature value with the second feature value to determine a pattern defect in the image to be inspected.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First Embodiment FIG. 1 is a conceptual diagram showing a hardware configuration of a pattern inspection apparatus 1A according to a first embodiment of the present invention. Pattern inspection device 1A
A computer system including a CPU 2, a storage unit 3 including a semiconductor memory and a hard disk, a media drive 4 for reading information from various recording media 7, an input unit 5 including a keyboard and a mouse, and a display unit 6 including a monitor and the like. It is.

The CPU 2 is connected to a storage unit 3, a media drive 4, an input unit 5, a display unit 6, and the like via a bus line BL and an input / output interface IF.
The media drive 4 is a CD-ROM, DVD
(Digital Versatile Disk), a flexible disk or the like, and reads the information recorded therein from a portable recording medium 7. This computer system has a pattern inspection function described later by reading the program from the portable recording medium 7 on which the program is recorded. Further, the storage unit 3 includes a program storage unit 3a that stores all or a part of the read program, and an image storage unit 3 that stores various processed images.
b etc.

Further, the computer system includes a two-dimensional CCD camera 8 for reading a digital image and an object 9b.
XY table 9 on which (semiconductor wafer W here) is placed
a is also provided. The inspection image S representing the whole or a part of the surface state of the object 9b is captured by the CCD camera 8, input to the computer system via the bus line BL or the like, and transferred to the image storage unit 3b or the like. You. Thereafter, a defect inspection is performed on the inspection target image S by performing appropriate processing as described later.

FIG. 2 is a functional block diagram showing a schematic configuration of the pattern inspection apparatus 1A. As shown in FIG. 2, the pattern inspection apparatus 1A functionally compares the inspected image S with the reference image R, and detects a defect candidate area D in the inspected image S.
(D1 to D5, see FIG. 3) Comparative inspection unit 10
A mask unit 21 for masking the area other than the defect candidate area D in the inspected image S, a feature extracting unit 31 for extracting a characteristic from the masked inspected image MS (hereinafter also referred to as “masked image MS”), A mask unit 22 for masking a portion other than the defect candidate region D in the reference image R; a feature extraction unit 32 for performing feature extraction on a masked reference image MR (hereinafter also referred to as “masked image MR”); and a feature extraction unit 31 And a determining unit 40 for determining a defect based on the feature amounts CS and CR extracted by the feature extracting unit 32.

Next, the operation of the pattern inspection apparatus 1A will be described with reference to FIGS.

As shown in FIG. 3, the comparison inspection unit 10
The inspected image S is compared with the reference image R to identify defect candidate areas D (D1 to D5) in the inspected image S. Specifically, a difference between the gradation values (multi-valued) of the corresponding pixels of the inspection image S and the reference image R is calculated, and the absolute value of the difference is calculated by a predetermined threshold value TH (not shown). The binarized defect candidate image Q is binarized (“1” when it is larger than the threshold TH, and “0” when it is smaller than the threshold TH).
Generate Then, a pixel region whose pixel value is “1” in the defect candidate image Q is specified as the defect candidate region D. FIG. 3 shows a case where regions D1 to D5 are extracted as defect candidate regions D in the defect candidate image Q. Here, areas in which the value “1” is spatially continuous in the defect candidate image Q are labeled as a single area, and the respective areas are denoted by reference numerals and are referred to as areas D1 to D5. That is, the matching between the values of the pixels corresponding to each other in the inspection image S and the reference image R is determined based on a predetermined criterion, and among the pixels whose pixel values do not match each other based on the criterion, spatially continuous pixels are determined. That is, the defect candidate areas D1 to D5 are specified collectively, and the defect candidate image Q is defined as a set of these defect candidate areas D1 to D5.

Here, in this binarization, the threshold value TH
It is preferable to use a small value. The reason is that an area that is not extracted as the defect candidate area D is determined as not being a defect without any inspection in a later step, while an area that is extracted as the defect candidate area D is described later. As described above, it is because whether or not a defect is detected by repeating feature extraction and the like. That is, by using a small value as the threshold value TH, as many defect candidate areas as possible are extracted, so that the ratio of overlooking true defects is kept low, and a feature extraction operation (described later) is further performed on the extracted defect candidate areas. This is because the addition makes it possible to accurately detect a true defect and to reduce the rate of detecting a so-called pseudo defect other than the true defect.

The mask unit 21 uses the defect candidate image Q to mask an area other than the defect candidate area D in the image S to be inspected, and generates a masked image MS. That is, for each pixel of the inspection image S, when the value of the corresponding pixel in the defect candidate image Q is “1”, the gradation value of the pixel of the inspection image S is set to the same value as the original gradation value. Conversely, when the value of the corresponding pixel in the defect candidate image Q is “0”, the pixel of the image to be inspected S is masked and its gradation value is set to “0”, so that the masked inspection An image, that is, a masked image MS can be obtained. In FIG. 3, each of the regions D <b> 1 to D <b> 1 shown in white
Each pixel of D5 has the original tone value (of the image S to be inspected),
Other pixels have the value “0”.

The feature extracting section 31 performs feature extraction on the masked image MS. Here, first, a case where an average density (average tone value) is obtained as the feature amount CS in the feature extraction will be described.

FIG. 4 is a diagram showing a part of the masked image MS. Each numeral in the matrix in FIG. 4 indicates a defect candidate area (shown by a thick line) D1 to D3 in the masked image MS.
This represents the density value (tone value) of each pixel in the vicinity. In this case, the average density (A1, A2, A3) for each of the defect candidate areas D1 to D3 is (75, 80, 20). In this way, the other defect candidate areas D4, D5
Also, the average densities A4 and A5 are determined. Here, the operation of calculating the feature amount may be performed only on the area where the density value of the masked image MS is not “0”, and it is necessary to perform the operation on the area where the density value of the masked image MS is “0”. Since there is no processing, the efficiency of the processing can be improved.

Next, the same processing is performed on the reference image R. Therefore, first, the mask unit 22 (FIG. 2) uses the defect candidate image Q to mask a portion other than the defect candidate region D in the reference image R, and generates a masked image MR. That is, for each pixel of the reference image R, the defect candidate image Q
When the value of the corresponding pixel in “1” is “1”, the gradation value of the pixel of the reference image R is set to the same value as the original gradation value, and conversely,
When the value of the corresponding pixel in the defect candidate image Q is “0”, the pixel of the reference image R is masked and the gradation value is set to “0”, whereby the masked image MR can be obtained. . In FIG. 3, each pixel of each of the regions D1 to D5 indicated by outline in the masked image MR has the original tone value (of the image S to be inspected), and the other pixels have the value “0”. have.

The feature extracting section 32 outputs the masked image M
Feature extraction is performed in R. Here, although the result is not shown, the average density B1 to B1 which is the same characteristic amount CR is also obtained from the mask image MR corresponding to the mask image MS in FIG.
B5 is extracted.

The judging unit 40 determines the feature amounts CS and CR extracted by the feature extracting unit 31 and the feature extracting unit 32.
The defect is determined based on Specifically, the average density A
1 to A5 and B1 to B5 are compared with each other to determine whether or not there is a defect. For example, the average density A1
By binarizing the absolute value of the difference between B1 and B1 using a predetermined threshold value and making a determination, it is possible to determine whether or not the defect candidate area D1 is defective.

Here, a comparison process by binarization using a threshold value is also performed. Since the comparison process is related to a feature amount called an average density for a plurality of pixels,
This is an advanced (including more information) information processing as compared with a simple gradation value comparison process between one pixel. In other words, the content of the pattern can be examined in more detail with respect to the defect candidate area D having a spread as a surface, and the defect is determined after adding information such as graphic information in the defect candidate area D. Can be determined. Therefore, a defect can be detected more accurately, and erroneous detection and the like can be suppressed as compared with a case where the tone values of the corresponding pixels are simply compared between the inspection image S and the reference image R. . In addition, such comparison of the feature amounts may be performed only for the defect candidate area D, and need not be performed for all pixels. Therefore, it is possible to increase the processing efficiency.

Next, a case where the processing result of the autocorrelation filter is used as the characteristic amounts CS and CR will be described. FIG. 5 is a diagram illustrating an example of an autocorrelation filter for obtaining an autocorrelation function. Here, it is assumed that the first-order autocorrelation filter FT1 of FIG. 5A is used. This autocorrelation filter FT1 is configured by a 3 × 3 matrix,
The value “1” at the center position and the position adjacent above the center position
And at other positions it has the value “0”. Then, each pixel position of the processing target image (the image to be inspected S or the reference image R) is associated with the center position of the autocorrelation filter FT1, and pixels existing at positions corresponding to the value “1” in the autocorrelation filter FT1 are compared. And the product of
A value obtained by dividing the product by a predetermined value (for example, an average density value or a maximum density value) is set as a new tone value at the pixel position. By performing such processing sequentially on each pixel of each processing target image, feature extraction can be performed. In this case, for each defect candidate area Di, a new sum (or average) of new tone values may be obtained as a feature amount for each defect candidate area Di, but a new sum for each pixel in each defect candidate area Di may be obtained. The gradation value may be obtained as a feature amount for each pixel.

The processing result obtained by applying the autocorrelation filter FT1 to the masked image MS has a large value in the case of a figure (for example, a line) connected in the vertical direction, and a figure connected in the vertical direction. Otherwise, it has a small value. That is, the result of processing by the autocorrelation filter is obtained as an index indicating a graphic feature having directivity in a specific direction (in this case, directivity in a direction at 90 degrees to the horizontal direction). Here, the “vertical direction” and the “horizontal direction” are defined as meaning the column direction and the row direction of the matrix pixel array, respectively.

Therefore, the inspection image S having a pattern in which the horizontal stripes appear repeatedly in the vertical direction.
In FIG. 5, if an autocorrelation filter as shown in FIG. 5A is used and the processing result is detected as being large, it is determined that there is a high probability that a graphic (for example, a line) extending in the vertical direction exists, and Can be determined to be a defect.

As described above, the processing by such an autocorrelation filter is performed on each of the image to be inspected S and the reference image R, and the results are obtained as feature quantities CS and CR. By comparing the values and determining whether or not the defect exists, more accurate defect detection can be performed.

The same applies to the case where the autocorrelation filters FT2 to FT4 shown in FIGS. 5B to 5D are used.
An index indicating a graphic feature having directivity in the degree direction can be obtained, and each of the autocorrelation filters FT2 to FT4 can be obtained.
According to the characteristics of the pattern to be inspected, it is possible to perform appropriate defect detection by selecting one of them or using a combination of some of them. Further, by performing detection using an autocorrelation filter over a wider range (for example, an autocorrelation filter having a size of 5 × 5 or a size of 10 × 10), it is also possible to add advanced texture analysis.

As described above, according to the pattern inspection apparatus 1A of the first embodiment, the area extracted as the defect candidate area D is further subjected to feature extraction and the like to determine whether or not it is a defect. More detailed defect determination can be performed, and the rate of detecting so-called pseudo defects other than true defects can be reduced. Furthermore, in the binarization in the comparison inspection unit 10, if as many areas as possible are extracted as the defect candidate areas D using a small value as the threshold value TH, the rate of missing true defects can be suppressed low. True defects can also be accurately detected.
Further, the defect candidate area D is extracted prior to the comparison determination by the feature extraction, and the feature extraction is performed only in the defect candidate area D. Therefore, it is not necessary to perform the feature extraction for all the pixels, and the processing efficiency is improved. be able to.

<Second Embodiment> In the first embodiment, when performing a defect inspection on an image to be inspected, features are also extracted from a reference image R (more precisely, an image MR to be masked), and the extracted features are extracted. The defect is determined based on the quantity CR. In order to perform more accurate detection, it is preferable to provide such a configuration, but the present invention is not limited to this. For example, the feature value CR regarding the reference image R
May be determined based on the feature amount CS extracted with respect to the masked inspection image MS without extracting. In the second embodiment, such a case will be described.

FIG. 6 is a functional block diagram showing a schematic configuration of a pattern inspection apparatus 1B according to the second embodiment. FIG.
As shown in FIG.
The pattern inspection apparatus according to the first embodiment includes the comparison inspection unit 10, the mask unit 21, the feature extraction unit 31, and the determination unit 40, but does not include the mask unit 22 and the feature extraction unit 32 for the reference image R. This is different from the device 1A. Further, the determination unit 40 determines the characteristic amount CR regarding the masked image MR.
The first embodiment is different from the first embodiment in that a defect is detected based on a feature amount CS extracted with respect to a masked image MS without being based on.

In the second embodiment, the judgment unit 40
Determines the defect by comparing the feature amount CS extracted by the feature extraction unit 31 with respect to the masked image MS with a predetermined reference value related to the feature amount CS.

For example, the average density value A is used as the characteristic amount CS.
And the average density value Ai calculated for each defect candidate area Di is larger (or smaller) than a predetermined reference value.
In this case, it can be determined as a defect. This is particularly useful when, for example, extracting a defect appearing in a part of the inspection image S having a small (or large) gradation value that is uniform throughout.

Alternatively, when the processing result of the autocorrelation filter is larger (or smaller) than a predetermined reference value as the characteristic amount CS, it can be determined as a defect. As an example, an autocorrelation filter as shown in FIG. 5A is used for an image S to be inspected having a stripe-like pattern in which lines extending in the horizontal direction repeatedly appear in the vertical direction, and the processing result is a large value. Is detected, it is considered that there is a high probability that a figure (for example, a line) connected in the vertical direction is present, and that part may be determined to be defective.

As described above, when the defect has a specific tendency based on the graphic characteristics of the inspection image S, it is possible to make a determination based on the tendency.

Further, in the pattern inspection apparatus 1B of the second embodiment, since it is not necessary to extract the feature amount relating to the reference image R, it is possible to further improve the processing efficiency.

<Third Embodiment> In the first and second embodiments, the feature is not directly extracted from the defect candidate image Q. In the third embodiment, a case will be described in which feature extraction is performed directly from the defect candidate image Q and a defect is determined based on the extracted feature amount.

FIG. 7 is a functional block diagram showing a schematic configuration of a pattern inspection apparatus 1C according to the third embodiment. FIG.
As shown in (1), the pattern inspection apparatus 1C includes a feature extraction unit 33 in addition to the functional elements of the pattern inspection apparatus 1A. The feature extracting unit 33 performs feature extraction on the defect candidate image Q and extracts a feature amount CQ. As this feature extraction, for example, one relating to the area can be performed. That is, a region where the pixels having the value “1” are continuous in the defect candidate image Q (defect candidate region D)
Is counted, and this number can be regarded as “area”. This “area” is the feature amount CQ.

Then, the judgment unit 40 determines that the masked image M
Defect detection is performed based on the feature values CQ extracted for the defect candidate image Q in addition to the feature values CS and CR for S and MR.

Here, a rule for performing the above-described comparison / determination operation (comparison / determination operation between the feature amounts CS and CR) only for the defect candidate region D having an area equal to or larger than a predetermined threshold value is adopted. . For example, among the defect candidate regions Di, the comparison between the feature amounts CS and CR is performed only for the defect candidate region D having the area equal to or larger than the threshold value. Is
The comparison between the feature amounts CS and CR may not be performed. According to this, it is not necessary to perform the above-described comparison and determination operation on the defect candidate region D having a small area which is considered to be based on erroneous detection due to the influence of noise or the like, thereby further improving the processing efficiency. Can be planned.

<Others> If the technical principle in each of the above embodiments is generalized, according to the present invention, the inspection by the feature extraction method is performed on the image to be inspected including the pattern only in the range selected by the screening by the comparative inspection. By doing so, both accuracy and speed of the defect inspection are achieved.

In each of the above embodiments, the image to be inspected S is input using the two-dimensional CCD camera 8, but the image to be inspected S is relatively scanned by the line sensor on the object. You may enter it.

Furthermore, in calculating the average density value in the first and second embodiments, the sum of the density values in each defect candidate area Di is divided by the number Ni of pixels included in each defect candidate area Di. Required by
The number of pixels Ni can be obtained by calculating the number of pixels that are not “0” when adding the density value.
Similarly to the embodiment, the number of pixels “1” can be obtained in advance by totaling the number of pixels of “1” using the defect candidate image Q.

In each of the above embodiments, with respect to the defect candidate image Q, the value of the pixel existing at the defect candidate position is expressed as “1”, and the value of the pixel existing at a position other than the defect candidate position is expressed as “0”. Although the binarized image is used, “1” and “0” may be expressed in reverse.

Further, in each of the above embodiments, the characteristic amount is exemplified by the area, the average density, the autocorrelation characteristic, and the like. However, the present invention is not limited to this, and other characteristics (eg, geometric characteristics) may be used. It may be an index to indicate. Further, the comparison determination may be performed by appropriately combining the feature amounts.

Further, in each of the above embodiments, the pattern inspection apparatus 1 is realized by configuring a part of the functions as software in a computer system. However, the present invention is not limited to this. Each function may be implemented by hardware. In that case, higher-speed processing becomes possible.

Further, in each of the above embodiments, the defect candidate area D is obtained by binarizing the absolute value of the difference between the inspection image S and the reference image R with the threshold value TH, and pixels larger than the threshold value TH are used as the defect candidate area. Although the pixel is specified as a pixel, a peripheral pixel (adjacent pixel) may be added thereto and specified as a defect candidate area. In particular, when the feature value extraction requires even the gradation values of the peripheral pixels of each defect candidate area Di (such as when a large-size autocorrelation filter is used), the peripheral pixels are also included. It is preferable that the defect candidate area D is set. Further, by using a larger area as a defect candidate, detection of a pseudo defect can be excluded, and oversight of a true defect can be avoided.

In specifying the defect candidate area D, as shown in FIG. 8, the defect candidate image Q is divided into a plurality of divided areas RD (each divided area RD is composed of k × k pixels). In addition, among the plurality of divided regions RD, a divided region RD (including a region specified here as DA) specified by binarization of the absolute value of the difference between the image to be inspected S and the reference image R described above ( 8 may be specified as the defect candidate area D. FIG. 8 shows areas DA1 and DA2 as areas DA specified by binarization, and a divided area R including these areas DA1 and DA2.
This shows a case where D1 and RD2 are specified as defect candidate areas D. The defect candidate area D specified in this way
By performing defect detection as described above for
It is possible to achieve both accuracy and speed of defect inspection.

[0055]

As described above, claims 1 to 6 are as described above.
According to the pattern inspection apparatus described in (1), since a defect candidate area is specified and further feature extraction is performed to determine a defect, accurate defect detection can be performed. In addition, since it is not necessary to perform feature extraction for all pixels of the image to be inspected, processing efficiency can be improved.

In particular, according to the pattern inspection apparatus of the third aspect, feature extraction is also performed on a defect candidate area of the reference image, and not only the feature amount in the masked inspection image but also the masked reference image. Since the defect is determined based on the feature amount, more accurate detection can be performed.

According to the pattern inspection apparatus of the fifth aspect, feature extraction is also performed on the defect candidate area of the defect candidate image, and based on the extracted feature amount,
Since the defect is determined, the processing can be performed efficiently.

Furthermore, according to the pattern inspection method of the present invention, since the inspection by the feature extraction method is performed only for the range selected by the screening by the comparative inspection, both accuracy and speed are achieved. Then, a defect inspection can be performed.

[Brief description of the drawings]

FIG. 1 is a conceptual diagram illustrating a hardware configuration of a pattern inspection apparatus 1A according to a first embodiment of the present invention.

FIG. 2 is a functional block diagram illustrating a schematic configuration of a pattern inspection apparatus 1A.

FIG. 3 is a conceptual diagram showing images S, R, MS, MR, Q, and a defect candidate area D.

FIG. 4 is a conceptual diagram showing a part of a masked image MS.

FIG. 5 is a diagram illustrating an example of an autocorrelation filter for obtaining an autocorrelation function.

FIG. 6 is a functional block diagram illustrating a schematic configuration of a pattern inspection apparatus 1B according to a second embodiment.

FIG. 7 is a functional block diagram illustrating a schematic configuration of a pattern inspection apparatus 1C according to a third embodiment.

FIG. 8 is an explanatory diagram relating to identification of a defect candidate area D in a modification.

[Explanation of symbols]

 1, 1A-1C pattern inspection apparatus 8 CCD camera 9a XY table 9b object CQ, CR, CS feature D, D1-D5, Di defect candidate area FT1-FT4 auto-correlation filter MR, MS masked image Q defect candidate image R Reference image S Image to be inspected W Semiconductor wafer

Continued on front page F term (reference) 2F065 AA49 BB02 CC19 DD00 DD06 FF04 FF42 FF61 JJ03 JJ09 JJ26 PP12 QQ00 QQ03 QQ05 QQ21 QQ24 QQ25 QQ33 QQ37 QQ42 RR05 RR06 RR08 5B057 AA03 BA02 DC33 DA07 DA07 DA07

Claims (10)

[Claims] 1. A pattern inspection apparatus for performing a defect inspection of a pattern by image processing, comprising: generating a defect candidate image for specifying a defect candidate area in the inspection image by comparing the inspection image including the pattern with a reference image. Candidate image generating means for performing the operation, first mask means for using the defect candidate image to mask areas other than the defect candidate area in the image to be inspected, and first feature extraction for performing characteristic extraction in the masked image to be inspected A pattern inspection apparatus, comprising: a determination unit configured to determine a pattern defect of the inspection target image based on the first feature amount extracted by the first feature extraction unit. 2. The pattern inspection apparatus according to claim 1, wherein the defect candidate image generation unit determines: (a) a predetermined value for each pixel corresponding to each other in the inspection image and the reference image; Pixel value matching means to be determined by a criterion, (b) among the pixels whose pixel values are disagreeed with each other by the criterion, collectively identify spatially continuous ones to identify a defect candidate area, Defect candidate area defining means for defining the defect candidate image as a set of areas, wherein the first feature value is extracted from the masked inspected image for each of the defect candidate areas; A pattern inspection apparatus, wherein a defect is determined by comparing the first characteristic amount with a predetermined threshold value. 3. The pattern inspection apparatus according to claim 1, wherein a second mask unit that masks a portion other than the defect candidate area using the defect candidate image in the reference image; Second feature extraction in reference image
Pattern extracting means for determining a pattern defect of the image to be inspected based on the second characteristic amount extracted by the second characteristic extracting means. Inspection equipment. 4. The pattern inspection apparatus according to claim 3, wherein: the defect candidate image generation unit determines: (a) a predetermined value of each pixel corresponding to each other in the inspection image and the reference image; Pixel value matching means to be determined by a criterion, (b) among the pixels whose pixel values are disagreeed with each other by the criterion, collectively identify spatially continuous ones to identify a defect candidate area, A defect candidate area defining unit that defines the defect candidate image as a set of areas, wherein the first feature quantity is extracted from the masked inspection image for each of the defect candidate areas; Is extracted from the reference image for each of the defect candidate areas, and the determining means performs a defect determination by comparing the first characteristic amount and the second characteristic amount with each other. apparatus . 5. The pattern inspection apparatus according to claim 1, further comprising: a third feature extracting unit configured to perform feature extraction on the defect candidate image; A pattern inspection apparatus characterized in that a pattern defect of the image to be inspected is determined based also on a third feature amount extracted by a feature extraction unit. 6. The pattern inspection apparatus according to claim 5, wherein: the defect candidate image generation unit determines: (a) a predetermined value for each pixel corresponding to each other in the inspection image and the reference image; Pixel value matching means to be determined by a criterion, (b) among the pixels whose pixel values are disagreeed with each other by the criterion, collectively identify spatially continuous ones to identify a defect candidate area, A defect candidate area defining unit that defines the defect candidate image as a set of areas, wherein the first feature quantity is extracted from the masked inspection image for each of the defect candidate areas; Is extracted from the reference image for each of the defect candidate areas, the third feature quantity is calculated for each of the defect candidate areas, and the determination unit compares the third feature quantity with a predetermined threshold value, Said ma Classifying means for classifying the checked image to be inspected into the inspection target image and the non-inspection target image; and, of the inspection target image and the non-inspection target image, only the inspection target image, the area of the inspection target image. A pattern inspection apparatus comprising: a selective determination unit configured to determine a defect by comparing the first characteristic amount with the second characteristic amount. 7. A pattern inspection method characterized in that an inspection image including a pattern is inspected by a feature extraction method only in a range selected by screening by a comparative inspection. 8. The pattern inspection method according to claim 7, wherein a comparison inspection is performed in which each pixel value of the inspection image including the pattern is compared with each pixel value of the reference image. Obtaining a first feature amount of each selected area of the inspection image selected while excluding an area whose pixel value matches the reference image by inspection, and determining a pattern defect of the inspection image based on the first characteristic amount. A pattern inspection method characterized by performing a judgment. 9. The pattern inspection method according to claim 8, wherein a second characteristic amount of the reference image is obtained for each of the selected regions, and the first characteristic amount and the second characteristic amount are obtained for each of the selected regions. A pattern inspection method, wherein a pattern defect of the image to be inspected is determined by comparison. 10. The pattern inspection method according to claim 8, wherein a third feature value is obtained from each of the selected regions, and the third feature value does not satisfy a predetermined criterion for each of the selected regions. A pattern inspection method, comprising: comparing the first characteristic amount with the second characteristic amount to determine a pattern defect in the image to be inspected while excluding a region.