JP5442305B2 - Defect detection apparatus and defect detection method - Google Patents

Description

  The present invention relates to a defect detection apparatus and a defect detection method for detecting a defect included in a first pattern by comparing a first image obtained by imaging a first pattern with a second image obtained by imaging a second pattern. About.

  When detecting a defect included in the first pattern by comparing the first image obtained by imaging the first pattern and the second image obtained by imaging the second pattern, the first image and the second image Alignment with is performed. In alignment, there is a case where alignment is performed with relatively high accuracy after alignment with relatively low accuracy.

  Patent Document 1 refers to such two-stage alignment. The column “Background Technology” in Patent Document 1 refers to “positioning at a high resolution after positioning at a low resolution”. In the column of “Best Mode for Carrying Out the Invention” in Patent Document 1, the “position correction unit 52” performs alignment and then the “defect detection unit 53” performs alignment and defect detection. To mention.

JP 2005-292016 A

  However, when the first image is divided into a plurality of first regions and the second image is divided into a plurality of second regions, and the first region and the second region are compared, the above 2 If the stage alignment is performed for all the pairs of the first region and the second region, the time required for the defect detection process becomes longer.

  The present invention has been made to solve this problem, and an object of the present invention is to provide a defect detection apparatus and a defect detection method in which the time required for defect detection is shortened without reducing the accuracy of defect detection. .

In order to solve the above-mentioned problems, the first and second aspects of the invention include a first image obtained by imaging the first pattern and a second image obtained by comparing the second image obtained by imaging the second pattern. A defect detection apparatus for detecting a defect to be detected, a first image dividing unit that divides a first image into a plurality of first regions, and a second that divides a second image into a plurality of second regions. And the first region and the second region with relatively low accuracy with respect to the pairs belonging to the first group among the plurality of pairs of the first region and the second region to be compared. A first alignment unit that executes the alignment of the first and second pairs of the first region and the second region to be compared that do not belong to the first group but belong to the second group Compared to the first support unit that uses the result of alignment for the pairs belonging to the first group by the first alignment unit, Based on a state in which alignment is performed by the first alignment unit for a plurality of pairs of the first region and the second region to be registered or a state in which the alignment result is incorporated by the first assistance unit A second alignment unit that performs alignment between the first region and the second region with relatively high or the same accuracy, and a plurality of the first region and the second region to be compared A defect detection unit that detects a defect from a comparison result between the first region and the second region for the pair.

The defect detection apparatus according to claim 1 , wherein the second alignment unit is relatively configured for a pair belonging to a third group among a plurality of pairs of the first region and the second region to be compared. Alignment of the first region and the second region is performed with high or the same accuracy, and the defect detection apparatus performs the first of the plurality of pairs of the first region and the second region to be compared. A second assisting unit that uses the result of alignment of the pair belonging to the third group by the second aligning unit to the remaining pairs belonging to the fourth group without belonging to the third group.

In the defect detecting apparatus according to claim 2, sequentially determines the appropriateness determining whether each of the plurality of pairs of the first and second regions to be compared are suitable for alignment of a relatively low accuracy And the pair determined to be unsuitable for relatively low accuracy alignment by the suitability determining unit are determined to belong to the second group, and determined to be unsuitable for relatively low accuracy alignment A group determination unit that determines that a pair determined to be suitable for relatively low-precision alignment after one pair belongs to the first group;

Further, in the defect detection apparatus according to claim 2, the group determination unit performs alignment with relatively low accuracy after one of the pairs determined to be suitable for alignment with relatively low accuracy by the suitability determination unit. The pair determined to be suitable is determined to belong to the second group.
According to a third aspect of the present invention, in the defect detection device of the second aspect, is the second alignment unit relatively high for all of the plurality of pairs of the first region and the second region to be compared? Alternatively, the alignment between the first area and the second area is executed with the same accuracy.
According to a fourth aspect of the present invention, in the defect detection apparatus according to the second aspect, the second alignment unit is included in a third group of a plurality of pairs of the first region and the second region to be compared. The defect detection apparatus performs alignment between the first region and the second region to be compared by performing alignment between the first region and the second region with relatively high or the same accuracy with respect to the pair to which the member belongs. Second assistance that uses the result of alignment of the pair belonging to the third group by the second alignment unit to the remaining pair belonging to the fourth group but not belonging to the third group A section.

According to a fifth aspect of the present invention, in the defect detection device according to any of the second to fourth aspects, the suitability determination unit is included in a determination target region that is one or both of the first region and the second region. An edge detection unit that detects an edge; and a determination unit that determines that a pair is not suitable for alignment with relatively low accuracy when no edge is detected by the edge detection unit.

According to a sixth aspect of the present invention, in the defect detection device according to any of the second to fourth aspects, the suitability determination unit is included in a determination target region that is one or both of the first region and the second region. An edge detection unit for detecting an edge, an average tangent direction deriving unit for deriving an average tangent direction of the edge detected by the edge detection unit, and a determination target region for the tangent direction derived by the average tangent direction deriving unit An autocorrelation evaluation unit that evaluates the autocorrelation of a pair, and when the autocorrelation derived by the autocorrelation derivation unit is higher than a reference, it is determined that the pair is not suitable for alignment with relatively low accuracy A determination unit.

The invention according to claim 7 is the defect detection device according to any one of claims 2 to 4 , wherein the suitability determination unit is included in a determination target region that is one or both of the first region and the second region. An edge detection unit for detecting an edge, a straight line detection unit for detecting a straight line composed of pixels constituting the edge detected by the edge detection unit, and a direction for deriving the direction of the straight line detected by the straight line detection unit A derivation unit, and a determination unit that determines that the pair is not suitable for alignment with relatively low accuracy when the uniformity of the direction derived by the direction derivation unit is higher than a reference.

The invention according to claim 8 is the defect detection device according to any one of claims 2 to 4 , wherein the suitability determination unit is included in a determination target region that is one or both of the first region and the second region. An edge detection unit for detecting an edge, a histogram derivation unit for deriving a histogram in the tangential direction of the edge detected by the edge detection unit, and the number of peaks present in the histogram derived by the histogram derivation unit is only one And a determination unit that determines that the pair is not suitable for alignment with relatively low accuracy.

According to a ninth aspect of the present invention, in the defect detection device according to any one of the second to fourth aspects, the suitability determination unit is included in a determination target region that is one or both of the first region and the second region. A corner detection unit that detects a corner; and a determination unit that determines that a pair is not suitable for relatively low-precision alignment when a corner is not detected by the corner detection unit.

According to a tenth aspect of the present invention, in the defect detection device according to any one of the second to fourth aspects, the suitability determination unit is a destination of a determination target region that is one or both of the first region and the second region. A distribution deriving unit for deriving an autocorrelation distribution when the range is swayed within a range, and a pair having a relatively low accuracy when the distribution derived by the distribution deriving unit has a direction that satisfies a criterion And a determination unit that determines that the position is not suitable for the positioning.

According to an eleventh aspect of the present invention, in the defect detection apparatus according to any one of the first to tenth aspects, the first alignment unit includes an arithmetic processing circuit that performs an operation for alignment, and an alignment processing circuit. An operation dividing unit that divides the operation into a plurality of times and causes the arithmetic processing circuit to perform the operation.

According to a twelfth aspect of the present invention, in the defect detection apparatus according to any one of the first to tenth aspects, the second alignment unit includes an arithmetic processing circuit that performs an operation for alignment, and an alignment processing circuit. An operation dividing unit that divides the operation into a plurality of times and causes the arithmetic processing circuit to perform the operation.

The invention according to claims 13 and 14 detects a defect included in the first pattern by comparing the first image obtained by imaging the first pattern and the second image obtained by imaging the second pattern. A method comprising: (a) dividing a first image into a plurality of first regions; (b) dividing a second image into a plurality of second regions; and (c) being compared. Performing alignment between the first region and the second region with relatively low accuracy with respect to pairs belonging to the first group among a plurality of pairs of the first region and the second region. (D) The remaining pair belonging to the second group but not belonging to the first group of the plurality of pairs of the first region and the second region to be compared is assigned to the first pair according to the step (c). Using the result of the alignment for the pairs belonging to the group; and (e) for a plurality of pairs of the first region and the second region to be compared by the step (c) Alignment between the first region and the second region is performed with relatively high or the same accuracy based on the state where the alignment is performed or the state where the alignment result is used in the step (d). And (f) detecting a defect from a comparison result between the first region and the second region for a plurality of pairs of the first region and the second region to be compared.
14. The defect detection method according to claim 13, wherein the step (e) is relatively high for pairs belonging to the third group among a plurality of pairs of the first region and the second region to be compared. Alternatively, the first region and the second region are aligned with the same accuracy, and the defect detection method includes: (g) Of the plurality of pairs of the first region and the second region to be compared, A step of using a result of alignment of the pairs belonging to the third group in the step (e) for the remaining pairs belonging to the fourth group without belonging to the third group.
15. The defect detection method according to claim 14, wherein (g) sequentially determines whether or not each of the plurality of pairs of the first region and the second region to be compared is suitable for relatively low-precision alignment. And (h) determining that the pair determined to be unsuitable for relatively low accuracy alignment by the step (g) belongs to the second group, and is suitable for relatively low accuracy alignment. A step of determining that a pair determined to be suitable for relatively low-precision alignment after one of the pairs determined not to belong to the first group includes the step (h), A pair determined to be suitable for relatively low-precision alignment is determined to belong to the second group after one of the pairs determined to be suitable for relatively low-precision alignment in (g).

  According to the present invention, since relatively low-precision alignment is not performed for pairs belonging to the second group, the time required for defect detection is shortened. In addition, since a relatively low accuracy alignment result for the pairs belonging to the first group is used for the pairs belonging to the second group, there is little decrease in the accuracy of defect detection.

According to the first aspect of the present invention, since the alignment that is relatively high or the same accuracy is not performed for the pairs belonging to the fourth group, the time required for detecting the defect is shortened. In addition, since the result of relatively low accuracy alignment for the pairs belonging to the third group is used for the pairs belonging to the fourth group, there is little reduction in the accuracy of defect detection.

According to the invention of claim 2 , relatively low accuracy alignment is not executed for a pair that is not suitable for relatively low accuracy alignment, and relative to a pair that requires relatively low accuracy alignment. Therefore, the accuracy of defect detection is improved.

According to the second aspect of the present invention, since a relatively low accuracy alignment is not executed for a pair that does not necessarily require a relatively low accuracy alignment, the time required for detecting a defect is shortened.

According to the fifth to tenth aspects of the present invention, since it is appropriately determined whether or not it is suitable for alignment with relatively low accuracy, alignment with relatively low accuracy is appropriately executed.

According to the eleventh to twelfth aspects of the present invention, since the scale of the arithmetic processing circuit that performs the arithmetic operation for the alignment is reduced, the defect detection apparatus is simplified.

It is a figure explaining the alignment of 1st Embodiment. It is a figure which shows the state by which the to-be-inspected image was divided | segmented. It is a figure which shows the state by which the reference image was divided | segmented. It is a block diagram of the defect detection apparatus of a 1st embodiment. It is a block diagram of the position alignment part of 1st Embodiment. It is a figure explaining the alignment of 2nd Embodiment. It is a block diagram of the position alignment part of 2nd Embodiment. It is a block diagram of the global alignment necessity determination apparatus of 3rd Embodiment. It is a figure explaining the determination of the group determination part of 3rd Embodiment. It is a figure which shows the example of the determination object area | region suitable for global alignment. It is a figure explaining the autocorrelation of the determination object area | region suitable for global alignment. It is a figure which shows the example of the determination object area | region which is not suitable for global alignment. It is a figure explaining the autocorrelation of the determination object area | region which is not suitable for global alignment. It is a block diagram of the global alignment necessity determination apparatus of 4th Embodiment. It is a figure which shows the differential operator of Prewitt. It is a block diagram of the global alignment suitability determination apparatus of 5th Embodiment. It is a figure which shows the straight line detected when a determination object area | region is suitable for global alignment. It is a figure which shows the straight line detected when a determination object area | region is not suitable for global alignment. It is a block diagram of the global alignment suitability determination apparatus of 6th Embodiment. It is a figure which shows the histogram derived when the determination object area | region is suitable for global alignment. It is a figure which shows the histogram derived | led-out when the determination object area | region is not suitable for global alignment. It is a block diagram of the global alignment suitability determination apparatus of 7th Embodiment. It is a figure which shows the corner detected when a determination object area | region is suitable for global alignment. It is a block diagram of the global alignment suitability determination apparatus of 8th Embodiment. It is a figure which shows distribution of the index value of autocorrelation, when a determination object area | region is suitable for global alignment. It is a figure which shows distribution of the autocorrelation index value derived | led-out by a distribution derivation | leading-out part, when a determination object area | region is not suitable for global alignment. It is a figure explaining global alignment. It is a figure explaining fine alignment. It is a block diagram of the alignment apparatus of 9th Embodiment.

<1 First Embodiment>
The first embodiment relates to a defect detection apparatus 102 that detects a defect included in an inspected pattern by comparing an inspected image A obtained by imaging the inspected pattern with a reference image B obtained by imaging a reference pattern.

(Outline of alignment of the first embodiment)
FIG. 1 is a diagram for explaining the alignment between the areas to be inspected a1 to a12 and the reference areas b1 to b12 according to the first embodiment. 2 and 3 respectively show a state in which the inspection image A is divided into a plurality of inspection regions a1 to a12 and a state in which the reference image B is divided into a plurality of reference regions b1 to b12.

  In the first embodiment, as shown in FIG. 2, the inspected image A is divided into a plurality of inspected areas a1 to a12, and as shown in FIG. 3, the reference image B is divided into a plurality of reference areas b1 to b12. Then, for each of a plurality of pairs (i = 1, 2,..., 12) of the inspected area ai and the reference area bi to be compared, the inspected area ai and the reference area bi are aligned. Later, the inspection area ai and the reference area bi are compared to detect a defect.

  In the alignment of the first embodiment, as shown in FIG. 1, the “inspection area ai and the reference area bi” in pixel units for a “part” of a plurality of pairs of the inspection area ai and the reference area bi to be compared. A global alignment is performed to align with the. Further, fine alignment is performed in which a plurality of pairs of the inspected area ai and the reference area bi to be compared are aligned with the inspected area ai and the reference area bi in subpixel units or in units coarser than the subpixels. Is executed. That is, both global alignment and fine alignment are executed for pairs belonging to the first group. However, for the remaining pairs that do not belong to the first group but belong to the second group, the global alignment is not executed and only the fine alignment is executed. Moreover, the result of the global alignment performed immediately before is used for the pair which belongs to the 2nd group in which global alignment is not performed.

  As a result, for the pairs belonging to the first group, global alignment is executed, and fine alignment is further executed on the basis of the state where the global alignment is performed. On the other hand, for the pairs belonging to the second group, the result of the global alignment performed immediately before is used, and the fine alignment is executed based on the state where the result of the global alignment performed immediately before is used. Thereby, since global alignment is not performed about the pair which belongs to the 2nd group, time required for detection of a defect becomes short. Moreover, since the result of the global alignment about the pair which belongs to the 1st group is used for the pair which belongs to the 2nd group, there is little fall of the precision of a defect detection.

  Note that the number of divisions of the image A and the reference image B is increased or decreased as necessary. Further, the pairs belonging to the first group and the second group are changed as necessary.

(Outline of the defect detection apparatus 102)
FIG. 4 is a block diagram of the defect detection apparatus 102 according to the first embodiment. Each block in FIG. 4 may include dedicated hardware, or all or part of its functions may be realized by causing a computer to execute a program.

  As shown in FIG. 4, the defect detection apparatus 102 includes an imaging unit 104 that captures a pattern, an inspection image dividing unit 106 that divides the inspection image A into a plurality of inspection regions a1 to a12, and a reference image B. From the comparison result between the reference image dividing unit 108 that divides the plurality of reference regions b1 to b12, the alignment unit 110 that aligns the inspection region ai and the reference region bi, and the inspection region ai and the reference region bi. A defect detection unit 112 that detects defects.

(Imaging unit 104)
The imaging unit 104 captures an inspection pattern to generate an inspection image A, and also captures a reference pattern to generate a reference image B. The inspected image A is given to the inspected image dividing unit 106, and the reference image B is given to the reference image dividing unit 108.

(Inspected image dividing unit 106 and reference image dividing unit 108)
The inspected image dividing unit 106 and the reference image dividing unit 108 regularly divide the inspected image A and the reference image B in the horizontal direction and the vertical direction, respectively. The number of divisions in the horizontal direction and the vertical direction is the same for the image A and the reference image B. As a result, the inspection area ai and the reference area bi are aligned with respect to the pair of the inspection area ai and the corresponding reference area bi, and the inspection area ai and the reference area bi are compared. However, if the inspection area ai and the reference area bi are aligned and the inspection area ai and the reference area bi are compared, how the inspection image A and the reference image B are divided. Also good. For example, the inspected image A and the reference image B may be divided only in the horizontal direction or may be divided only in the vertical direction. Each of a plurality of pairs of the inspected area ai and the reference area bi to be compared is sequentially given to the alignment unit 110.

(Alignment unit 110)
FIG. 5 is a block diagram of the alignment unit 110 according to the first embodiment. Each block in FIG. 5 may also include dedicated hardware, or all or part of its functions may be realized by causing a computer to execute a program.

  As shown in FIG. 5, the alignment unit 110 divides the pair of the region to be inspected ai and the reference region bi into a first group in which global alignment is performed and a second group in which global alignment is not performed. A pair classification unit 114 that classifies, a global alignment execution unit 116 that executes global alignment, a fine alignment execution unit 118 that executes fine alignment, and a global that uses the execution result of global alignment for one pair as another pair. An alignment execution result assistance unit 120.

  The pairs classified into the first group by the pair classification unit 114 are given to the global alignment execution unit 116, and global alignment is executed for the pair. Further, the pair is given to the fine alignment execution unit 118 after the global alignment is executed, and the fine alignment is executed on the basis of the state where the global alignment is executed for the pair. The amount of positional deviation in units of pixels derived by the global alignment of the global alignment execution unit 116 is given to the global alignment execution result assisting unit 120.

  On the other hand, the pair classified into the second group by the pair classifying unit 114 is given to the global alignment execution result assisting unit 120, and the pixel unit position derived by the previous global alignment given from the global alignment executing unit 116 The shift amount is used as a positional shift amount in pixels for the pair. Therefore, for the pairs belonging to the second group, the amount of displacement in pixels is determined without performing global alignment. Further, the pair is given to the fine alignment execution unit 118 after the amount of displacement in units of pixels is used, and the inspected area ai and the reference region bi are relative to each other by the amount of position deviation in units of pixels used for the pair. Fine alignment is executed with reference to the moved state as a reference.

  After completion of the alignment process for one pair, as long as there are pairs for which alignment has not been performed, alignment is performed for other pairs for which alignment has not yet been performed. It is also desirable to perform pipeline processing for starting the alignment process for another pair before the alignment process for one pair is completed.

(Defect detection unit 112)
The defect detection unit 112 detects a defect by comparing the inspection area ai and the reference area bi with respect to the pair of the inspection area ai and the reference area bi on which fine alignment has been performed. The defect may be detected in any way, but it is typically performed by obtaining a difference or ratio between two corresponding pixel values of the inspection area ai and the reference area bi.

(2-step alignment)
In the above description, the case where fine alignment in which alignment is performed in units of sub-pixels is performed after global alignment in which alignment is performed in units of pixels is described, but more generally, alignment with relatively low accuracy is performed. The present invention is applied in the case where a two-stage alignment is performed in which a relatively high or the same accuracy of the alignment is performed after.

(pattern)
The pattern for which the defect is to be detected may be any pattern, but patterns such as a flat display panel, a printed board, a lead frame, and a photomask are typical.

<2 Second Embodiment>
The second embodiment relates to an alignment unit 210 that is employed in place of the alignment unit 110 of the first embodiment.

(Outline of alignment of the second embodiment)
FIG. 6 is a diagram for explaining alignment between the inspected areas a1 to a12 and the reference areas b1 to b12 according to the second embodiment.

  In the alignment of the second embodiment, as shown in FIG. 6, as in the alignment of the first embodiment, the “partial” of a plurality of pairs of the inspected area ai and the reference area bi to be compared is globally Although alignment is performed, unlike the alignment of the first embodiment, fine alignment is performed on “part” rather than “all” of a plurality of pairs of the inspected area ai and the reference area bi to be compared. .

  That is, as shown in FIG. 6, global alignment is executed for pairs belonging to the first group. However, global alignment is not performed for the remaining pairs that do not belong to the first group but belong to the second group. Moreover, the result of the global alignment performed immediately before is used about the pair which belongs to the 2nd group in which global alignment is not performed.

  Further, as shown in FIG. 6, fine alignment is executed for pairs belonging to the third group. However, fine alignment is not executed for the remaining pairs that do not belong to the third group and belong to the fourth group. Moreover, the result of the fine alignment performed immediately before is used about the pair which belongs to the 4th group in which fine alignment is not performed. As a result, for the pairs belonging to the first group, whether global alignment is performed and fine alignment is further performed on the basis of the state where the global alignment is performed (pairs belonging to the third group), or The result of the fine alignment performed immediately before is used (a pair belonging to the fourth group). On the other hand, for the pairs belonging to the second group, is the result of the global alignment executed immediately before being used and whether the fine alignment is executed with reference to the state where the result of the global alignment executed immediately before is used? (A pair belonging to the third group) or the result of the fine alignment performed immediately before is used (a pair belonging to the fourth group). According to such alignment, since the global alignment is not executed for the pairs belonging to the second group, the time required for detecting the defect is shortened. Moreover, since the result of the global alignment about the pair which belongs to the 1st group is used for the pair which belongs to the 2nd group, there is little fall of the precision of a defect detection. In addition, since fine alignment is not executed for the pairs belonging to the fourth group, the time required for defect detection is shortened. In addition, since the fine alignment result for the pairs belonging to the third group is used for the pairs belonging to the fourth group, there is little decrease in the accuracy of defect detection. Such omission of fine alignment is executed when it is expected that there is no large variation in the amount of positional deviation in units of subpixels.

  The pairs belonging to the first group to the fourth group are changed as necessary.

(Alignment unit 210)
FIG. 7 is a block diagram of the alignment 210 of the second embodiment. Each block in FIG. 7 may also include dedicated hardware, or all or part of its functions may be realized by causing a computer to execute a program.

  As shown in FIG. 7, the alignment unit 210 is the same as the alignment unit 110 of the first embodiment. The pair classification unit 214, the global alignment execution unit 216, the global alignment execution result assistance unit 220, and the fine alignment execution are performed. However, unlike the alignment unit 110 of the first embodiment, a fine alignment execution result assisting unit 222 that assists the execution result of fine alignment for one pair with another pair is also provided.

  The pairs classified into the first group by the pair classification unit 214 are given to the global alignment execution unit 216, and global alignment is executed for the pair. The positional deviation amount in pixel units derived by the global alignment of the global alignment execution unit 216 is given to the global alignment execution result assisting unit 220.

  On the other hand, the pair classified into the second group by the pair classifying unit 214 is given to the global alignment execution result assisting unit 220, and is a pixel unit position derived by the previous global alignment given from the global alignment executing unit 216. The shift amount is used as a positional shift amount in pixels for the pair.

  After global alignment is performed or after a positional shift amount in units of pixels is used, the pairs belonging to the third group are given to the fine alignment execution unit 218, and the state in which global alignment has been performed for the pair or immediately before Fine alignment is executed on the basis of the state where the result of the global alignment executed in step 1 is used. The amount of positional deviation in units of sub-pixels derived by the fine alignment of the fine alignment execution unit 218 is given to the fine alignment execution result assistance unit 222.

  On the other hand, after global alignment is performed or after a positional deviation amount in units of pixels is used, pairs belonging to the fourth group are given to the fine alignment execution result assistance unit 222 and given from the fine alignment execution unit 218. The positional deviation amount in subpixel units derived by the fine alignment just before is used as the positional deviation amount in subpixel units for the pair. Accordingly, for the pairs belonging to the fourth group, the amount of positional deviation in sub-pixel units is determined without performing fine alignment.

  After completion of the alignment process for one pair, as long as there are pairs for which alignment has not been performed, alignment is performed for other pairs for which alignment has not yet been performed. It is also desirable to perform pipeline processing for starting the alignment process for another pair before the alignment process for one pair is completed.

<3 Third Embodiment>
When the alignment of the first embodiment or the alignment of the second embodiment is performed, is the pair of the inspection area ai and the reference area bi belonging to the first group where the global alignment is performed manually determined? The image A to be inspected or the reference image B is pre-scanned and automatically determined. However, if global alignment is performed for a pair that is not suitable for global alignment, or if global alignment is not performed for a pair that requires global alignment, the alignment is not performed properly, and the accuracy of defect detection decreases.

  The third embodiment relates to a global alignment necessity determination device 324 that is provided to solve this problem and determines whether or not global alignment is necessary. The global alignment necessity determination device 324 of the third embodiment is mounted on the pair classification unit 114 of the first embodiment or the pair classification unit 214 of the second embodiment.

  FIG. 8 is a block diagram of the global alignment necessity determination device 324 according to the third embodiment. Each block in FIG. 8 may also include dedicated hardware, or all or part of its functions may be realized by causing a computer to execute a program.

  As shown in FIG. 8, the global alignment necessity determination device 324 includes a suitability determination unit 326 that sequentially determines whether or not each pair of the region to be inspected ai and the reference region bi to be compared is suitable for global alignment. A group determination unit 328 that determines a group to which a pair of the inspected area ai and the reference area bi to be compared belongs.

  FIG. 9 is a transition diagram for explaining the determination by the group determination unit 328. “Appropriate” and “Inappropriate” described in the nodes N102 and N104 in FIG. “Global alignment + fine alignment” and “fine alignment” described along with the arrows AR102, AR104, AR106, AR108 are appropriateness described in the nodes N102, N104 from which the arrows AR102, AR104, AR106, AR108 exit. The determination by the determination unit 326 is performed on the previous pair of the pair that is the target of the determination by the group determination unit 328, and the suitability described in the nodes N102 and N104 in which the arrows AR102, AR104, AR106, and AR108 enter This is the content of alignment executed when the determination by the determination unit 326 is performed on the pair that is the target of the determination by the group determination unit 328.

  As shown in FIG. 9, the pair determined not to be suitable for the global alignment by the suitability determining unit 326 is determined to belong to the second group, and the global alignment is not performed on the pair and only the fine alignment is performed. (Arrows AR102 and AR106). In addition, a pair determined to be suitable for global alignment after one of the pairs determined to be unsuitable for global alignment by the suitability determination unit 326 is determined to belong to the first group, and global alignment is executed for the pair. After that, fine alignment is executed (arrow AR104). Thereby, global alignment is not executed for pairs that are not suitable for global alignment, and global alignment is executed for pairs that require global alignment, so that the accuracy of defect detection is improved.

  Furthermore, a pair determined to be suitable for global alignment after one of the pairs determined to be suitable for global alignment by the suitability determination unit 326 may be determined to belong to either the first group or the second group. It is desirable to determine that the pair belongs to the second group, and it is desirable that only the fine alignment is performed for the pair without performing the global alignment (arrow AR108). As a result, since global alignment is not performed for pairs that do not necessarily require global alignment, the time required for defect detection is reduced.

  Note that, as described above, in principle, a pair determined to be unsuitable for global alignment by the suitability determination unit 326 is determined to belong to the second group, but global alignment has never been executed. Fine alignment cannot be executed when there is no execution result of the global alignment that is supported. Therefore, in such a case, the processing of the pair is postponed, and after the global alignment is executed for the subsequent pair, the exception processing for executing the fine alignment is executed.

<4 Fourth Embodiment>
In the suitability determination unit 326 of the third embodiment, whether each pair of the inspected area ai and the reference area bi to be compared is suitable for global alignment is automatically referred to the inspected image A or the reference image B. It is determined by. However, if the determination is not properly performed, the global alignment is not properly performed.

  In the fourth embodiment, a global alignment suitability determination device 430 is provided to solve this problem. The global alignment suitability determination device 430 determines whether each of the pair of the region to be inspected ai and the reference region bi to be compared is suitable for global alignment. About. The global alignment suitability determination device 430 is mounted on the suitability determination unit 326 of the third embodiment.

  FIGS. 10A to 10G are diagrams illustrating examples of determination target regions suitable for global alignment. FIG. 11 is a diagram for explaining autocorrelation of a determination target region suitable for global alignment.

  When a closed figure pattern with closed edges is included as shown in FIGS. 10A to 10C, patterns with edges having corners are included as shown in FIGS. 10D to 10F. In this case, the determination target region is suitable for global alignment when a pattern having a curved edge as shown in FIG. As shown in FIG. 11, an image before moving the determination target area JR (hereinafter referred to as “pre-movement image”) BI and an image after moving the determination target area JR (hereinafter “post-movement image”). This is because the AI is not similar regardless of the movement direction MD of the determination target area JR. That is, the autocorrelation of the determination target region JR is low with respect to an arbitrary direction.

  12A to 12H are diagrams illustrating examples of determination target regions that are not suitable for global alignment. FIG. 13 is a diagram illustrating the autocorrelation of a determination target region that is not suitable for global alignment.

  When a pattern is not included as shown in FIG. 12A, when most of the edges are composed of a pattern extending in the vertical direction as shown in FIGS. 12B to 12C, FIG. When most of the edges consist only of a pattern extending in the horizontal direction as shown in (f), a pattern in which most of the edges extend in a specific oblique direction as shown in FIGS. For example, the determination target area is not suitable for global alignment. This is because, as shown in FIG. 13, the pre-movement image BI and the post-movement image AI are similar depending on the movement direction MD of the determination target region JR. That is, this is because the autocorrelation of the determination target region JR is high in a specific direction.

  FIG. 14 is a block diagram of the global alignment suitability determination device 430 of the fourth embodiment. Each block in FIG. 14 may also include dedicated hardware, or all or part of its functions may be realized by causing a computer to execute a program.

  As illustrated in FIG. 14, the global alignment suitability determination device 430 includes an edge detection unit 432 that detects an edge included in the determination target region, an average tangent direction derivation unit 434 that derives an average tangent direction of the edge, and a specification. An autocorrelation evaluation unit 436 that evaluates the autocorrelation of the determination target region with respect to the direction of the image, and a determination unit 438 that determines whether the pair is suitable for global alignment. The “determination target area” is one or both of the inspection area ai and the reference area bi.

  The edge detection unit 432 includes the differential value dx of the density value I (x, y) in the horizontal direction (x direction) and the vertical direction (y direction) in each of the plurality of pixels PX (x, y) constituting the determination target region. A differential value dy of the density value I (x, y) is derived. The differential value may be derived in any way. For example, the Prewitt differential operator shown in FIGS. 15A to 15B is set to 3 pixels in the horizontal direction centering on the pixel from which the differential value is derived, Derived by applying to a square area of 3 pixels in the vertical direction. Note that the application of the Prewitt differential operator has the advantage of a high noise smoothing effect, but this does not preclude derivation by application of a differential operator such as Sobel, Roberts, or simple differential operation. .

  Further, the edge detection unit 432 derives the magnitude | d | of the normal vector d = (dx, dy) having the differential value dx as the horizontal component and the differential value dy as the vertical component according to the equation (1), Pixels for which | d | exceeds a threshold are detected as pixels constituting the edge.

  This edge detection utilizes the fact that the normal vector magnitude | d | becomes smaller in the flat portion and larger in the edge portion.

  The average tangent direction deriving unit 434 derives the average tangent direction θ of the edge detected by the edge detecting unit 432. The average tangential direction θ may be derived in any way, but here it is derived by the “raw data vector summing method”. In the “raw data vector summing method”, the angle of the vector derived is not summed but the angle of the vector derived is summed, and the angle of the vector summed is derived.

When the average tangential θ derived by the "raw data vector summing method", first, the normal vector d ₁ of the edge in each of the plurality of pixels constituting the edge, d _2, · · ·, a d _i Formula Add up according to (2).

The sum of the normal vectors according to the equation (2) is obtained by adding j to a vector md _j−1 obtained by summing the _first to _j−1 normal vectors d ₁ , d ₂ ,. It is performed by repeating the addition of th normal vector d _j or j-th normal vector d _j a 180 ° reversed vector -d _j. However, md ₁ = d ₁ .

Add normal vector + d _j is the inner product of the vector md _j-1 and the normal vector d _j larger vector than the angle formed [pi / 2 or [pi / 2 with md _j-1 and the normal vector d _j md _This is a case where _j−1 · d _j is 0 or positive. Add vector -d _j is the inner product md _j-1 · a vector md _j-1 and the normal vector d _j greatly angle formed than [pi / 2 radians vector md _j-1 and the normal vector d _j This is the case when _dj is negative. According to such summation, it is avoided that the sum of the normal vectors d ₁ , d _2, ..., D _i is close to the 0 vector due to the inversion of the vector due to a subtle difference in density value. Such summation is permitted by the fact that the average tangential direction to be derived is not the vector direction itself but the vector slope. Further, according to such summation, the contribution of strong edges becomes large and the contribution of weak edges becomes small, so that the influence of erroneously detected edges becomes small.

The average tangential derivation unit 434, normal vector d _1, d _2, · · ·, a horizontal component md _ix and vertical components md _iy summation vector md _i derived by summing the d _i formula ( The average tangential direction θ is derived according to 3). Tangential theta, an angle between a vector perpendicular and horizontal to the summing vector md _i, takes values from - [pi] / 2 to [pi / 2.

  The autocorrelation evaluation unit 436 evaluates the autocorrelation of the determination target region with respect to the average tangent direction θ derived by the average tangent direction derivation unit 434. The autocorrelation may be evaluated in any way. For example, the autocorrelation is evaluated based on a difference in density value in an overlapping portion between the pre-movement image and the post-movement image.

In evaluating the autocorrelation, the density value I (x, y) in the pixel PX (x, y) constituting the pre-movement image and the pixel PX ′ (x, y) constituting the post-movement image are centered. A square difference sd _ij (x, y) with each of the density values I ′ (x + i, y + j) of the plurality of pixels PX ′ (x + i, y + j) within the marginal range is derived according to the equation (4). The least square difference SD (x, y) of the square differences sd _ij (x, y) is derived according to equation (5). By shifting the position of the density value I ′ (x + i, y + j) from which the difference from the density value I (x, y) is taken in this way, the sum of squared differences SSD (to be described later) increases due to calculation accuracy and quantization error. It is suppressed.

  In equations (4) and (5), x and y are the coordinates of the pixels in the horizontal and vertical directions, respectively. In equations (4) and (5), i and j are the amounts of displacement in the horizontal direction and the vertical direction, respectively, and take a value of -1, 0, or +1.

  Further, the sum of square differences SSD of the least square differences SD (x, y) for all the pixels included in the overlapping portion is derived according to the equation (6).

  A vector s that moves the determination target region in the tangential direction θ by the distance k is derived according to Equation (7).

The size k of the vector s is desirably increased so that the amount of movement in the determination target region is several pixels or more. As a result, the pre-movement image and the post-movement image are too close to each other, and the square difference sum SSD is suppressed from becoming small despite the small autocorrelation. It is also desirable to derive a square difference sum SSD using a plurality of magnitudes k ₁ and k ₂ and adopt the largest square difference sum SSD. As a result, when the density value I (x, y) has periodicity, the amount of movement in the determination target region matches the period, so that the square difference sum SSD is reduced even though the autocorrelation is small. It is suppressed. The ratio k ₁ : k ₂ of the plurality of magnitudes k ₁ and k ₂ is preferably equal to the ratio m: n of the prime integers m and n (k ₁ , k ₂ = m: n). Thereby, it is avoided that the amount of movement of the determination target region for both the plurality of sizes k ₁ and k ₂ matches the cycle.

  The determination unit 438 determines that the pair is not suitable for global alignment when the edge detection unit 432 does not detect an edge. Thereby, global alignment is not performed in the case as shown in FIG. In addition, the determination unit 438 determines that the pair is not suitable for global alignment when the double difference total SSD is below the threshold value and the autocorrelation is higher than the reference. Thereby, global alignment is not performed in the case as shown in FIGS. In addition, the determination unit 438 determines that the pair is suitable for global alignment in other cases.

  According to such a global alignment suitability determination process, it is appropriately determined whether or not it is suitable for the global alignment, and therefore the global alignment is appropriately executed.

<5 Fifth Embodiment>
The fifth embodiment relates to a global alignment suitability determination device 530 employed in place of the global alignment suitability determination device 430 of the fourth embodiment.

  FIG. 16 is a block diagram of the global alignment suitability determination device 530 of the fifth embodiment.

  As illustrated in FIG. 16, the global alignment suitability determination device 530 includes an edge detection unit 532 that detects an edge included in a determination target region, a straight line detection unit 540 that detects a straight line including pixels, and a straight line direction. A direction deriving unit 542 for deriving and a determining unit 538 for determining whether or not the pair is suitable for global alignment.

  The edge detection unit 532 detects edges in the same manner as the edge detection unit 432 of the fourth embodiment.

  The straight line detection unit 540 detects a straight line constituted by pixels constituting the edge detected by the edge detection unit 532 by a general Hough transform or the like.

  The direction deriving unit 542 derives the direction of the straight line detected by the straight line detecting unit 540.

  FIGS. 17A to 17B are diagrams illustrating a straight line L detected by the straight line detection unit 540 when the determination target region is suitable for global alignment. 18A to 18B are diagrams illustrating a straight line L detected by the straight line detection unit 540 when the determination target region is not suitable for global alignment. When the determination target region is suitable for global alignment, the direction of the detected straight line L is non-uniform as shown in FIG. 17, and when the determination target region is not suitable for global alignment, as shown in FIG. The direction of the detected straight line L is uniform.

  Similar to the determination unit 438 of the fourth embodiment, the determination unit 538 determines that the pair is not suitable for global alignment when no edge is detected by the edge detection unit 532. Furthermore, unlike the determination unit 438 of the fourth embodiment, the determination unit 538 uses the above-described change in straight line uniformity, and the uniformity of the direction derived by the direction deriving unit 542 is higher than the reference. If it is high, it is determined that the pair is not suitable for global alignment. In addition, the determination unit 538 determines that the pair is suitable for global alignment in other cases.

  Such a global alignment suitability determination device 530 also appropriately determines whether or not it is suitable for global alignment, so that global alignment is appropriately executed.

<6 Sixth Embodiment>
The sixth embodiment relates to a global alignment suitability determination device 630 employed in place of the global alignment suitability determination device 430 of the fourth embodiment.

  FIG. 19 is a block diagram of a global alignment suitability determination device 630 according to the sixth embodiment.

  As illustrated in FIG. 19, the global alignment suitability determination device 630 includes an edge detection unit 632 that detects an edge included in the determination target region, a histogram derivation unit 644 that derives a histogram in the tangential direction of the edge, and a pair that is a global alignment. And a determination unit 638 for determining whether or not it is suitable.

  The edge detection unit 632 detects edges in the same manner as the edge detection unit 432 of the fourth embodiment.

  The histogram deriving unit 644 derives a histogram in the tangential direction of the edge detected by the edge detecting unit 632.

  FIG. 20 is a diagram illustrating a histogram derived by the histogram deriving unit 644 when the determination target region is suitable for global alignment. FIG. 21 is a diagram illustrating a histogram derived by the histogram deriving unit 644 when the determination target region is not suitable for global alignment. When the determination target region is suitable for global alignment, as shown in FIG. 20, there are two or more peaks in the histogram, and when the determination target region is not suitable for global alignment, only one peak exists in the histogram. To do.

  Similar to the determination unit 438 of the fourth embodiment, the determination unit 638 determines that the pair is not suitable for global alignment when no edge is detected by the edge detection unit 632. Furthermore, unlike the determination unit 438 of the fourth embodiment, the determination unit 638 uses the change in the number of peaks present in the above-described histogram, and the number of peaks present in the histogram derived by the histogram deriving unit 644. If there is only one, it is determined that the pair is not suitable for global alignment. In addition, the determination unit 638 determines that the pair is suitable for global alignment in other cases.

  Such a global alignment suitability determination device 630 also appropriately determines whether or not it is suitable for global alignment, so that global alignment is appropriately executed.

<7 Seventh Embodiment>
The seventh embodiment relates to a global alignment suitability determination device 730 employed in place of the global alignment suitability determination device 430 of the fourth embodiment.

  FIG. 22 is a block diagram of the global alignment suitability determination device 730 according to the seventh embodiment.

  As illustrated in FIG. 22, the global alignment suitability determination device 730 includes an edge detection unit 732 that detects an edge included in the determination target region, a corner detection unit 746 that detects a corner included in the determination target region, and a global pair And a determination unit 738 for determining whether or not it is suitable for alignment.

  The edge detection unit 732 detects an edge in the same manner as the edge detection unit 432 of the fourth embodiment.

  The corner detection unit 746 detects a corner included in the determination target region by a general Harris operator, a SUSAN operator, or the like.

  FIGS. 23A and 23B are diagrams illustrating a corner C detected by the corner detection unit 732 when the determination target region is suitable for global alignment. When the determination target region is suitable for global alignment, as shown in FIG. 23, one or more corners are detected. However, when the determination target region is not suitable for global alignment, corner C is not detected.

  Similar to the determination unit 438 of the fourth embodiment, the determination unit 738 determines that the pair is not suitable for global alignment when no edge is detected by the edge detection unit 732. Further, unlike the determination unit 438 of the fourth embodiment, the determination unit 738 uses the above-described presence / absence of corner detection, and if the corner is not detected by the corner detection unit 746, the pair is not suitable for global alignment. judge. In addition, the determination unit 738 determines that the pair is suitable for global alignment in other cases.

  Such a global alignment suitability determination device 730 also appropriately determines whether or not it is suitable for global alignment, so that global alignment is appropriately executed.

<8 Eighth Embodiment>
The eighth embodiment relates to a global alignment suitability determination device 830 employed in place of the global alignment suitability determination device 430 of the fourth embodiment.

  FIG. 24 is a block diagram of the global alignment suitability determination device 830 of the eighth embodiment.

  As shown in FIG. 24, the global alignment suitability determination device 830 includes a distribution derivation unit 848 for deriving an autocorrelation distribution when the movement destination of the determination target region is swayed within a range, and the pair is a global alignment. And a determination unit 838 for determining whether or not it is suitable.

  The eighth embodiment is different from the fourth embodiment in that the distribution derivation unit 848 swayes the movement destination of the determination target region within the range, instead of obtaining a direction with high autocorrelation from the tangential direction of the edge. Thus, an autocorrelation index value such as the difference value SSD is derived for each movement amount, and the determination unit 838 has autocorrelation with respect to a specific direction depending on whether or not the distribution has a specific directionality. It is to determine whether or not.

  FIG. 25 is a diagram showing a distribution of autocorrelation index values derived by the distribution deriving unit 848 when the determination target region is suitable for global alignment. FIG. 26 is a diagram illustrating a distribution of autocorrelation index values derived by the distribution deriving unit 848 when the determination target region is not suitable for global alignment. In FIGS. 25 and 26, the dark area has higher autocorrelation, and the bright area has lower autocorrelation. When the determination target region is suitable for global alignment, the distribution of the autocorrelation index value has no anisotropy, as shown in FIG. When the determination target region is not suitable for global alignment, the distribution of autocorrelation index values is anisotropic as shown in FIG.

  The determination unit 838 uses the presence / absence of the anisotropy of the autocorrelation distribution described above to determine whether the pair of autocorrelation index values derived by the distribution deriving unit has a direction that satisfies the criterion. It is determined that it is not suitable for alignment with relatively low accuracy. Whether or not it has directionality is determined by whether or not the direction of a straight line constituted by pixels constituting an edge is uniform, as in the fifth embodiment, with respect to the distribution of index values of autocorrelation. Check whether or not there is only one peak in the histogram in the tangential direction of the edge as in the sixth embodiment, or whether or not a corner is detected as in the seventh embodiment. It is specified by examining.

Such a global alignment suitability determination device 830 also appropriately determines whether or not it is suitable for global alignment, so that global alignment is appropriately executed.

<9 Ninth Embodiment>
When the correlation between the inspected area ai and the reference area bi is derived by the hardware processing circuit when performing the alignment of the first embodiment or the alignment of the second embodiment, the range for performing the global alignment is expanded, If the precision of fine alignment is improved, the scale of the arithmetic processing circuit increases.

For example, as shown in FIG. 27 showing the positional relationship between the region to be inspected ai and the reference region bi, if global alignment is performed within a range that extends ± P pixels in the horizontal and vertical directions, (2P + 1) ² adders Is required.

Further, as shown in FIG. 28 showing the positional relationship between the region to be inspected ai and the reference region bi, when fine alignment is performed in units of 1 / P pixel, interpolation processing for increasing one pixel to P pixels is performed, and (2P + 1) ) ^Two adders are required.

  The ninth embodiment relates to an alignment apparatus 950 provided to solve this problem. The alignment device 950 is mounted on all or part of the global alignment execution unit 116 and fine alignment execution unit 118 of the first embodiment and the global alignment execution unit 216 and fine alignment execution unit 218 of the second embodiment. The global alignment execution unit 116 and the fine alignment execution unit 118 may share a hardware arithmetic processing circuit, or the global alignment execution unit 216 and the fine alignment execution unit 218 share a hardware arithmetic processing circuit. Also good.

  FIG. 29 is a block diagram of an alignment apparatus 950 according to the ninth embodiment. The arithmetic processing circuit 952 in FIG. 29 is entirely or partially hardware, but the arithmetic processing dividing unit 954 may include dedicated hardware, or all or part of its functions may be programmed in the computer. It may be realized by executing.

  As shown in FIG. 29, the alignment device 950 includes an arithmetic processing circuit 952 that performs arithmetic processing for alignment, and an arithmetic dividing unit that divides the arithmetic for alignment into a plurality of times and causes the arithmetic processing circuit to perform the arithmetic processing circuit. 954.

  As a result, the scale of the arithmetic processing circuit that performs arithmetic processing for alignment is reduced, so that the defect detection apparatus 102 is simplified.

<10 Others>
Although the present invention has been described in detail, the above description is illustrative in all aspects, and the present invention is not limited to the description. Innumerable variations not illustrated may be envisaged without departing from the scope of the present invention.

DESCRIPTION OF SYMBOLS 102 Defect detection apparatus 106 Inspected image division part 108 Reference image division part 110,210 Position alignment part 112 Defect detection part 114,214 Pair classification part 116,216 Global alignment execution part 118,218 Fine alignment execution part 120,220 Global alignment Execution result assistance part 222 Fine alignment execution result assistance part

Claims (14)

A defect detection device that detects a defect included in a first pattern by comparing a first image obtained by imaging a first pattern with a second image obtained by imaging a second pattern,
A first image dividing unit for dividing the first image into a plurality of first regions;
A second image dividing unit for dividing the second image into a plurality of second regions;
Alignment between the first region and the second region is performed with relatively low accuracy with respect to pairs belonging to the first group among a plurality of pairs of the first region and the second region to be compared. A first alignment section;
Of the plurality of pairs of the first region and the second region to be compared, the remaining pair belonging to the second group and not belonging to the first group is added to the first group by the first alignment unit. A first assistance unit that uses the result of the alignment for the pair to which it belongs;
A state in which alignment is executed by the first alignment unit for a plurality of pairs of the first region and the second region to be compared, or a state in which the alignment result is incorporated by the first assistance unit A second alignment unit that performs alignment between the first region and the second region with relatively high or the same accuracy with respect to
A defect detection unit for detecting a defect from a comparison result between the first region and the second region for a plurality of pairs of the first region and the second region to be compared;
With
The second alignment unit is
Alignment of the first region and the second region with a relatively high or the same accuracy with respect to a pair belonging to the third group among a plurality of pairs of the first region and the second region to be compared Run
The defect detection apparatus includes:
Of the plurality of pairs of the first region and the second region to be compared, the remaining pair belonging to the fourth group and not belonging to the third group is added to the third group by the second alignment unit. A second aid part that uses the result of the alignment for the pair to which it belongs,
A defect detection apparatus further comprising: A defect detection device that detects a defect included in a first pattern by comparing a first image obtained by imaging a first pattern with a second image obtained by imaging a second pattern,
A first image dividing unit for dividing the first image into a plurality of first regions;
A second image dividing unit for dividing the second image into a plurality of second regions;
Alignment between the first region and the second region is performed with relatively low accuracy with respect to pairs belonging to the first group among a plurality of pairs of the first region and the second region to be compared. A first alignment section;
Of the plurality of pairs of the first region and the second region to be compared, the remaining pair belonging to the second group and not belonging to the first group is added to the first group by the first alignment unit. A first assistance unit that uses the result of the alignment for the pair to which it belongs;
A state in which alignment is executed by the first alignment unit for a plurality of pairs of the first region and the second region to be compared, or a state in which the alignment result is incorporated by the first assistance unit A second alignment unit that performs alignment between the first region and the second region with relatively high or the same accuracy with respect to
A defect detection unit for detecting a defect from a comparison result between the first region and the second region for a plurality of pairs of the first region and the second region to be compared;
With
A suitability determination unit that sequentially determines whether each of the plurality of pairs of the first region and the second region to be compared is suitable for alignment with relatively low accuracy;
One of the pairs determined to be unsuitable for relatively low accuracy alignment by determining that the pair determined to be unsuitable for relatively low accuracy alignment by the suitability determining unit belongs to the second group. A group determination unit that determines that a pair that is later determined to be suitable for relatively low-precision alignment belongs to the first group;
Further comprising
The group determination unit
Determining that the pair determined to be suitable for relatively low precision alignment after the pair determined to be suitable for relatively low precision alignment by the suitability determining unit belongs to the second group;
Defect detection device. The defect detection apparatus according to claim 2.
The second alignment unit is
Performing alignment of the first region and the second region with relatively high or the same accuracy for all of the plurality of pairs of the first region and the second region to be compared;
Defect detection device. The defect detection apparatus according to claim 2.
The second alignment unit is
Alignment of the first region and the second region with a relatively high or the same accuracy with respect to a pair belonging to the third group among a plurality of pairs of the first region and the second region to be compared Run
The defect detection apparatus includes:
Of the plurality of pairs of the first region and the second region to be compared, the remaining pair belonging to the fourth group and not belonging to the third group is added to the third group by the second alignment unit. A second aid part that uses the result of the alignment for the pair to which it belongs,
A defect detection apparatus further comprising: The defect detection apparatus according to any one of claims 2 to 4,
The suitability determination unit
An edge detection unit that detects an edge included in the determination target region that is one or both of the first region and the second region;
A determination unit that determines that a pair is not suitable for alignment with relatively low accuracy when an edge is not detected by the edge detection unit;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 2 to 4,
The suitability determination unit
An edge detection unit that detects an edge included in the determination target region that is one or both of the first region and the second region;
An average tangential direction deriving unit for deriving an average tangential direction of the edges detected by the edge detection unit;
An autocorrelation evaluating unit that evaluates the autocorrelation of the determination target region with respect to the tangential direction derived by the average tangential direction deriving unit;
A determination unit that determines that the pair is not suitable for alignment with relatively low accuracy when the autocorrelation derived by the autocorrelation deriving unit is higher than a reference;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 2 to 4,
The suitability determination unit
An edge detection unit that detects an edge included in the determination target region that is one or both of the first region and the second region;
A straight line detection unit for detecting a straight line constituted by pixels constituting the edge detected by the edge detection unit;
A direction deriving unit for deriving the direction of the straight line detected by the straight line detecting unit;
A determination unit that determines that the pair is not suitable for alignment with relatively low accuracy when the uniformity of the direction derived by the direction deriving unit is higher than a reference;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 2 to 4,
The suitability determination unit
An edge detection unit that detects an edge included in the determination target region that is one or both of the first region and the second region;
A histogram derivation unit for deriving a histogram in the tangential direction of the edge detected by the edge detection unit;
A determination unit that determines that a pair is not suitable for alignment with relatively low accuracy when the number of peaks present in the histogram derived by the histogram deriving unit is only one;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 2 to 4 ,
The suitability determination unit
A corner detection unit that detects a corner included in the determination target region that is one or both of the first region and the second region;
A determination unit that determines that a pair is not suitable for alignment with relatively low accuracy when a corner is not detected by the corner detection unit;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 2 to 4,
The suitability determination unit
A distribution deriving unit for deriving an autocorrelation distribution when the movement destination of the determination target region that is one or both of the first region and the second region is swung within the range;
A determination unit that determines that the pair is not suitable for alignment with relatively low accuracy when the distribution derived by the distribution deriving unit has a direction that satisfies a criterion;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 1 to 10,
The first alignment unit includes:
An arithmetic processing circuit for performing an arithmetic operation for alignment;
A calculation dividing unit for dividing the calculation for alignment into a plurality of times and causing the calculation processing circuit to perform the calculation;
A defect detection apparatus comprising: The defect detection apparatus according to any one of claims 1 to 10,
The second alignment unit is
An arithmetic processing circuit for performing an arithmetic operation for alignment;
A calculation dividing unit for dividing the calculation for alignment into a plurality of times and causing the calculation processing circuit to perform the calculation;
A defect detection apparatus comprising: A defect detection method for detecting a defect included in a first pattern by comparing a first image obtained by imaging a first pattern with a second image obtained by imaging a second pattern,
(a) dividing the first image into a plurality of first regions;
(b) dividing the second image into a plurality of second regions;
(c) Alignment of the first region and the second region with relatively low accuracy with respect to pairs belonging to the first group among a plurality of pairs of the first region and the second region to be compared. A step of executing
(d) The first group according to the step (c) is assigned to the remaining pairs belonging to the second group and not belonging to the first group among the plurality of pairs of the first region and the second region to be compared. Using the result of the alignment for pairs belonging to,
(e) A state in which the alignment of the plurality of pairs of the first region and the second region to be compared is performed in the step (c) or a state in which the result of the alignment is used in the step (d) Performing alignment of the first region and the second region with relatively high or the same accuracy relative to
(f) detecting a defect from a comparison result between the first region and the second region for a plurality of pairs of the first region and the second region to be compared;
Equipped with a,
The step (e)
Alignment of the first region and the second region with a relatively high or the same accuracy with respect to a pair belonging to the third group among a plurality of pairs of the first region and the second region to be compared Run
The defect detection method includes:
(g) The third group according to the step (e) is added to the remaining pair belonging to the fourth group and not belonging to the third group among the plurality of pairs of the first region and the second region to be compared. Using the result of alignment for pairs belonging to,
A defect detection method further comprising: A defect detection method for detecting a defect included in a first pattern by comparing a first image obtained by imaging a first pattern with a second image obtained by imaging a second pattern,
(a) dividing the first image into a plurality of first regions;
(b) dividing the second image into a plurality of second regions;
(c) Alignment of the first region and the second region with relatively low accuracy with respect to pairs belonging to the first group among a plurality of pairs of the first region and the second region to be compared. A step of executing
(d) The first group according to the step (c) is assigned to the remaining pairs belonging to the second group and not belonging to the first group among the plurality of pairs of the first region and the second region to be compared. Using the result of the alignment for pairs belonging to,
(e) A state in which the alignment of the plurality of pairs of the first region and the second region to be compared is performed in the step (c) or a state in which the result of the alignment is used in the step (d) Performing alignment of the first region and the second region with relatively high or the same accuracy relative to
(f) detecting a defect from a comparison result between the first region and the second region for a plurality of pairs of the first region and the second region to be compared;
Equipped with a,
(g) sequentially determining whether each of the plurality of pairs of the first region and the second region to be compared is suitable for relatively low accuracy alignment;
(h) It is determined that the pair determined to be unsuitable for relatively low precision alignment by the step (g) belongs to the second group, and is not suitable for relatively low precision alignment. Determining that a pair determined to be suitable for relatively low accuracy alignment after one of the pairs belongs to the first group;
Further comprising
The step (h)
Determining that a pair determined to be suitable for relatively low accuracy alignment after the pair determined to be suitable for relatively low accuracy alignment by the step (g) belongs to the second group;
Defect detection method.

Images