JPH10160681A - Pattern defect detection device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect a true defect by shifting a reference image and an image to be inspected in one direction from an alignment position, detecting a brightness difference at that time differentially, eliminating a noise where the level in one direction is equal to or less than a specific value is eliminated from a non-coincidence part, and providing a means for judging the true defect of an object to be inspected and a means for instructing the amount of shift. SOLUTION: An alignment part 105 aligns a reference image and an image to be inspected by pattern matching. A shift part 106 shifts the reference image in one direction by an amount of shift αaccording to an instruction from a shift amount instruction part 108. A true defect judgment part 104 extracts a part with a brightness that is equal to or more than a fixed threshold from the differential image, eliminates noises by filtering, and judges a true defect. A non-coincidence part whose brightness difference is equal to or more than a brightness threshold is extracted and at the same time a noise whose level is equal to m or less in one direction is eliminated from the non-coincidence part. When α exceeds MAX based on -MAX<=α<=-MIN, it is judged to be a true defect. On the other hand, when α is equal to or less than MIN, it is judged to be a pseudo defect.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern defect detecting apparatus and a pattern defect detecting method capable of detecting a defect in a circuit pattern of a semiconductor device such as an LSI wafer, and more particularly to a pattern capable of accurately detecting a defect. The present invention relates to a defect detection device and a pattern defect detection method.

[0002]

2. Description of the Related Art Conventionally, as an inspection method for inspecting a pattern shape, a technology for practical use of voice / image / character information edited by the Management Development Center (pp. 262-263, 1980).
The pattern comparison method described in 9.30) is known.

In this pattern comparison method, pattern inspection is basically performed by the following method. Here, a case where the semiconductor circuit pattern shown in FIG. 3 is inspected will be described as an example. FIGS. 3A and 3B are enlarged views of a portion of an image obtained by imaging a semiconductor element circuit pattern. FIG. 3A shows a reference image that does not include a defect, and an inspection target pattern to be inspected is inspected based on the reference image. FIG. 3B shows an image to be inspected including the two defects 1 and 2. FIG. 3 (b)
Is a noise (which should be determined as a pseudo defect) due to a quantization error or the like at the time of image input, and a defect 2 is a true defect due to the presence of a defect in the pattern.

Inspection of patterns is shown in FIGS. 3 (a) and 3 (b).
By comparing the luminance distributions of the images. here,
3 line a in (b), the luminance variation on b respectively G _y1
(X) and G _y2 (x). Also, let the luminance changes of the reference image on the lines a ′ and b ′ in FIG. 3A corresponding to the lines a and b be F _y1 (x) and F _y2 (x), respectively. Further, a correction formula for matching the luminance range between the reference image and the image to be inspected is H (x). At this time, the luminance changes G ′ _y1 (x) and G ′ _y2 (x) of the inspection image after correction.
G ′ _y1 (x) = G _y1 (x) + H _y1 (x) G ′ _y2 (x) = G _y2 (x) + H _y2 (x) FIGS. 3C and _3D show these G ′ _y1 (x),
It is a figure which shows _G'y2 (x) and _Gy1 (x), _Gy2 (x). FIGS. 3E and _3F respectively show G ′ _y1 (x)
FIG. 7 is a diagram simultaneously showing G _y2 (x) and F _y2 (x) corresponding thereto, and F _y1 (x) corresponding thereto. From these figures, it can be seen that there is a difference between the reference image and the image to be inspected.

Next, in order to compare the reference image and the image to be inspected, the difference between G ′ _y1 (x) and its corresponding F _y1 (x) and G ′ _y2 (x) and its corresponding F _y2 Calculate the difference of (x). The results are shown in FIGS. 3 (g) and (h), respectively. In these figures, the absolute value of the difference, ie, | F
_y1 (x) _-G'y1 (x) |, | _Fy2 (x) -G '
_y2 (x) |.

Then, the luminance difference obtained here (FIG. 3)
(G), (h)) with a certain set threshold
A value conversion process is performed, and portions where the luminance difference is larger than the threshold value are set as the mismatched portions l1 and l2. What remains after the noise is removed from the mismatched portion by the noise removal processing is detected as a true defect. Here, the noise removing process is a process of removing a noise component by removing, for example, only a portion having a certain width or more having a mismatched portion.

As a method of detecting a pattern defect based on the pattern comparison as described above, for example, Japanese Patent Publication No. 58-61
No. 448 discloses a method. This method reduces the standard pattern in the horizontal and vertical directions, creates a plurality of dead zones around the boundary, and uses this as a reference image to match all the patterns to the binarized image to be inspected. Is detected as a true defect.

In the method described in Japanese Patent Publication No. 63-32666, nine types are obtained by shifting the reference image after position adjustment by ± 1 pixel up, down, left and right, and the set luminance value ± α is added to the reference image. Total of 11 obtained by adding
A difference from a newly created reference image made of a seed is obtained, and a true defect is detected by performing binarization processing on a difference in polarity of the corresponding mismatched portion with the same polarity.

[0009]

However, the method described in Japanese Patent Publication No. 58-61448 is not suitable for detecting a defect near the boundary because of the provision of a dead zone.

In the method described in Japanese Patent Publication No. 63-26666, the number of comparisons with the reference image is large, so that a long processing time is required.

Further, in the method described in "Practical technology for voice / image / character information", for example, when the size is smaller than n × m size (n pixels in the Y direction and m pixels in the X direction) by noise removal processing, In the case of performing noise removal such as noise, FIG.
Since both l1 and l2 of (h) are less than the length m which is the X component length of the filter, they are removed as noise, and the mismatched part l caused by the original pattern defect
2 is not detected as a true defect (determined as a pseudo defect).

This problem can sometimes be detected by changing the size of the filter. However, usually, the size of the filter cannot be changed to a desired size due to limitations of hardware and the like. It takes time.

In order to detect the above-mentioned defect, it is possible to reduce the pixel size by increasing the resolution of an image used for defect detection, but on the other hand, it requires more processing time due to a decrease in throughput. Become.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem. In pattern defect detection, software change is minimized without changing hardware, and conventionally, it is determined that the defect is a pseudo defect. It is an object of the present invention to provide a pattern defect detection device and a pattern defect detection method capable of detecting a true defect.

[0015]

According to a first aspect of the present invention, there is provided a pattern defect detecting apparatus for aligning a reference image and an inspection image from an inspection object having a pattern shape with the reference image and the inspection image. Shift amount α in at least one direction from the position
Shift means for shifting the reference image after the shift, a difference means for detecting a difference in luminance between the reference image after the shift and the image to be inspected, and a non-coincidence part in which the luminance difference is equal to or greater than the luminance threshold value. True defect determining means for removing noise whose size in the direction is equal to or less than m to determine a true defect of the inspection object; determining a true defect if the size in the one direction on the image exceeds MAX. A shift amount instructing unit that instructs the shift unit to indicate a shift amount α that satisfies the following equation (I) when it is desired to determine a defect that is less than MIN as a pseudo defect;
It has.

M-MAX ≦ α ≦ m-MIN (I) In the pattern defect detection device according to the second aspect, in the pattern defect detection device according to the first aspect, a pattern in which the inspection object is repeated at a predetermined interval is provided. When the object has a shape, a reference image generating means for generating a reference image by shifting the inspection image from the inspection object by an amount corresponding to the interval is provided.

According to a third aspect of the present invention, in the pattern defect detecting method, the reference image and the inspection image from the inspection object having the pattern shape are moved in at least one direction from the position where the reference image and the inspection image are aligned. The following formula (I) and conditions (I
I) is shifted by the shift amount α that satisfies I), the luminance difference between the reference image after the shift and the image to be inspected is detected, and a non-coincidence part whose luminance difference is equal to or more than the luminance threshold value is extracted. The noise whose size in one direction is equal to or less than m is removed, and the remaining mismatched portion is determined as a true defect of the inspection object.

M-MAX ≦ α ≦ m-MIN (I) If the size in the one direction on the image exceeds MAX, it is determined to be a true defect, and if it is smaller than MIN, it is determined to be a pseudo defect. (II) In the pattern defect detection method according to the fourth aspect, in the pattern defect detection method according to the third aspect, the reference image may be moved relative to the inspected image in an upper right direction, a lower right direction, and an upper left direction. A pattern defect detection method characterized in that the pattern is shifted in four directions, that is, the lower left direction, and a true defect is determined by the four shifts.

According to a fifth aspect of the present invention, there is provided the pattern defect detecting method according to the third aspect.
When the inspection object has a pattern shape that is repeated at predetermined intervals, the inspection image from the inspection object is shifted by an integer multiple of the amount corresponding to the above-described interval to generate a reference image.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a diagram showing a configuration of a pattern defect detection apparatus according to the present embodiment. As shown in FIG. 1, the pattern defect detection device includes a reference image extraction unit 101, an inspection image extraction unit 102, an image shift processing unit 103, a true defect determination unit 10
Four. The image shift processing unit 103 includes a positioning unit 105, a shift unit 106, a difference unit 107, and a shift amount instruction unit 108.

FIG. 2 is a flow chart showing the flow of processing of a pattern detection method by the above-described pattern defect detection apparatus.
Hereinafter, the pattern defect detection apparatus of the present invention will be described with reference to this flowchart and FIG.

Here, the reference image extraction unit 10
A case will be described in which a defect of the inspected image captured by the inspected image extracting unit 102 is obtained using the reference image stored in the memory 1 or the like.

First, in step 1, the positioning unit 105 performs positioning between the reference image and the image to be inspected by pattern matching.

Next, in step 2, the shift unit 10
6 shifts the reference image by a shift amount to be described later according to the instruction from the shift amount instruction unit 108 (of course, the inspection image may be shifted).

Then, in step 3, the difference unit 10
7 obtains a difference in luminance between the image to be inspected and the reference image, thereby obtaining a difference image in which the mismatched portion is enlarged and emphasized.

Finally, in step 4, the true defect judging unit 104 extracts a portion having a luminance equal to or higher than a predetermined threshold from the difference image, and performs a noise removal process using a filter of a predetermined size to perform a true defect judgment. Make a decision.

Next, the operation of the image shift processing section 103 and the true defect judgment section 104 will be described. Here, the case of inspecting the semiconductor element circuit pattern of FIG. 3B will be specifically described. FIG. 3A shows a reference image to be compared. Lines a and b in FIG.
3 (a) correspond to F _y1 (x) and F _{y2 respectively.}
(X), which are shown in FIGS. 4 (a) and 4 (b), respectively.

Next, these reference images shifted leftward by the shift amount α are used as new reference images. They are,
FIG. 4 shows F ′ _y1 (x) and F ′ _y2 (x), respectively.
(A) and (b).

The luminance changes on the lines a and b of the pattern image to be inspected in FIG. 3B are represented by G _y1 (x) and G _y2 (x), respectively.
Further, assuming that a correction formula for matching the luminance range between the reference pattern image and the pattern image to be inspected is H (x), the luminance change G ′ after correction of the pattern image to be inspected is obtained.
_y1 (x) and _G'y2 (x) are as follows: _G'y1 (x) = _Gy1 (x) + _Hy1 (x) _G'y2 (x) = _Gy2 (x) + _Hy2 (x)

The reference pattern images F ′ _y1 (x) and F ′ corresponding to G ′ _y1 (x) and G ′ _y2 (x), respectively.
_y2 (x) is shown in FIG.
(D).

Next, in order to compare the reference image and the image to be inspected in lines a and b in FIG.
The difference (luminance difference) between _y1 (x) and G ′ _y1 (x) and F ′ _y2
The difference between (x) and G ′ _y2 (x) is obtained. FIG. 4 (e),
(F) is a diagram showing the absolute value of the difference result. The luminance difference | F ′ _y1 (x) −G ′ _y1 (x) |, |
F ′ _y2 (x) −G ′ _y2 (x) | is subjected to a binarization process at a certain set threshold value, and a portion having a luminance difference larger than the threshold value is regarded as a non-coincidence portion as l1 ′,
l2 '. 4 (e) and 4 (f) are compared with FIGS. 3 (g) and 3 (h) of the conventional example.
(E) and (f) show that the width of the mismatched portion exceeding the threshold value is larger.

Then, when a non-coincidence portion less than m is determined as noise by a noise removal process using a filter having an X-component length m, the non-coincidence portion l shown in FIG.
Since 1 'is equal to or less than the length m which is the X component length of the filter, it is removed as noise. On the other hand, the mismatched portion l2 'shown in FIG. 4F exceeds the length m and is detected as a defect. Thus, FIG.
Unmatched portions such as 11 in FIG. 3B are removed, and mismatched portions such as 12 in FIG. 3B due to pattern defects are detected as true defects.

As described above, according to the present invention, the width of the mismatched portion can be enlarged by simply shifting either the reference pattern image or the pattern image to be inspected, and the width can be arbitrarily changed by changing the shift amount. Can change. Therefore, only by adjusting the shift amount (without performing complicated arithmetic processing), it is possible to detect a defect having a desired size or more.

Here, for the sake of explanation, a one-dimensional description in the horizontal direction has been described. In addition to this, a two-dimensional pattern defect detection process can be performed by developing a similar process in the vertical direction. It is. Also, various defects can be detected by combining a plurality of shift directions.

Next, a method for determining the shift amount α will be described. When noise is removed by an nxm size filter, in the X-axis direction component, a mismatched portion having a length of Xmin or less is not detected as a defect (a pseudo defect), and a portion larger than the length Xmax is detected as a true defect. If so, the shift amount αx is set as follows.

M−Xmax ≦ αx ≦ m−Xmin (where Xmin ≦ Xmax) By setting such a shift amount, the length Xmi
Unmatched portions less than n are not detected, and those larger than Xmax are detected as true defects. Similarly, in the Y-axis direction component, n−Ymax ≦ αy ≦ n−Ymin (where Ymin
≤Ymax).

Next, a specific example of defect detection by the pattern defect detection method according to the present embodiment will be described. Here, the following model is used. That is, quantization is performed in the X and Y axis directions, the luminance value is further binarized, and the filter size for removing noise is set to 3 × 3. Also, the size of the true defect to be detected is larger than two pixels in both the horizontal and vertical directions, and the size of two or less pixels is determined as a pseudo defect. At this time, the shift amount is ± 1 pixel in both the horizontal and vertical directions according to the above equation. Note that the above values are only examples, and the generality of the present method is not lost.

Shifting an image by one in the positive direction of the X-axis is referred to as shifting to the right, shifting one in the positive direction of the Y-axis is referred to as shifting up, and those performing these shifts simultaneously are shifted to the upper right. It expresses to do.

In the case of the above conditions, an example of a defect of a projection / chuck shape in a straight line portion will be described first. FIG. 5A is a rectangular image composed of a gray portion surrounded by a solid line and does not include any defect, and is used as a reference image for pattern defect detection. FIG. 5B is an image to be inspected including a rectangular shape of the same portion as FIG. 5A, and the outer periphery of the rectangular shape of the reference image after alignment is indicated by a dotted line. In FIG. 5B, a rectangular area surrounded by a solid line and a dotted line includes protrusions 11 to 14 and chips 21 to 24 as defects. These defects are detected as mismatched portions after alignment, but are not detected. Except for 3 × 3, it is not detected as a defect.

FIG. 5C shows a non-coincidence portion when the reference image after the alignment is shifted to the upper right (+1 pixel in the X direction and +1 pixel in the Y direction) by a shaded gray area. At this time,
Since the protrusions 13 and 14 and the chips 21 and 22 are 3 × 3 areas as the defects included in FIG. 2B, they are detected as true defects. The detected true defect is shown as a dark gray area in FIG.

FIG. 5D shows the reference image at the lower left (X
(-1 pixel in the Y direction and -1 pixel in the Y direction). Similarly, the protrusions 11 and 12 and the chips 23 and 24 are detected as true defects.

Therefore, with respect to a defect having a chipping / projection shape as shown in FIG. 5 (b), if it is shifted to the upper right, a chipping at the upper side, a protrusion at the lower side, a chipping at the right side, and a protrusion at the left side are detected. If it shifts to the lower left, it can detect a protrusion on the upper side, a chip on the lower side, a protrusion on the right side, and a chip on the left side. Similarly, when shifting to the lower right, the protrusion on the upper side
Chipping on the lower side, chipping on the right side, and protrusion on the left side can be detected. When shifting to the upper left, chipping on the upper side, protrusion on the lower side, protrusion on the right side, and chipping on the left side can be detected.

Next, an example of a method for detecting a defect having a chipped shape at a 90 ° angle portion will be described. FIG.
5A is an image to be inspected including the same rectangular shape as that of FIG. 5A, and the dotted line indicates the outer periphery of the rectangular shape of the reference image after the alignment. In FIG. 6A, a rectangular area surrounded by a solid line and a dotted line is included as defects 31 to 34, and these defects are detected as non-coincidence portions after alignment, but are less than 3 × 3. Not detected as a defect. FIG. 6B shows a non-coincidence portion when the reference image after alignment is shifted to the upper right by a shaded gray area. At this time, as the defect included in FIG.
Since the area is × 3, the area is detected as a true defect. This is shown in the dark gray area in FIG. Similarly, a defect 32 is detected as a true defect when shifted to the lower right, a defect 33 is shifted when shifted to the lower left, and a defect 34 is detected as a true defect when shifted to the upper left. Can also be detected.

Next, a method for detecting a protrusion-shaped defect at a 90 ° angle portion will be described. Two types of projections are considered as minimum components. FIG. 7A includes defects including the protrusions 41 to 44, and FIG. 7B illustrates a non-coincidence portion when shifting to the upper right is indicated by a shaded gray area. At this time, the defects 42 and 43 in FIG. 7A are detected as true defects. These are shown as dark gray areas in FIG. From this, with respect to a defect having a projection shape as shown in FIG.
2, 43 are projected to the lower right.
However, when shifted to the lower left, the projections 41 and 44 are detected as true defects, and when shifted to the upper left, the projections 41 and 42 are detected as true defects.

FIG. 7C includes, as defects, projections 51 to 54 having shapes different from those of FIG. 7A.
(D) shows the unmatched part when shifting to the upper right is indicated by the shaded gray area. At this time, the defects 53 and 54 in FIG.
Are detected as true defects. These are shown as dark gray areas in FIG. From this, regarding the defect of the projection shape as shown in FIG. 7C, the projections 53 and 54 are shifted to the upper right and the projections 51 and 54 are shifted to the lower right.
When 54 is shifted to the lower left, the protrusions 51 and 52 are detected as true defects, and when shifted to the upper left, the protrusions 52 and 53 are detected as true defects. Similarly, when shifted to the lower left, the defects 51 and 52 are detected as true defects.

A defect including at least one of the two types of defects shown here as a minimum component can also be detected.
As described above, a defect at a 90 ° corner can be detected. Next, an example of detecting a defect at a 270 ° corner will be described. This 270 ° corner is substantially equal to the defect detection at the 90 ° corner. FIG. 8A shows a cross-shaped region including a 270 ° angle. FIG. 8B is an image to be inspected including the shape of the same portion as FIG. 8A, and the outer periphery of the reference image after alignment is indicated by a dotted line. In FIG. 8B, a region surrounded by a solid line and a dotted line is included as a defect of the cutouts 61 and 62 and the protrusion 63. FIG. 8C shows a non-coincidence portion at the time of shifting to the upper right by a shaded gray area. At this time, FIG.
The defect included in (b) is detected as a true defect. FIG.
Regarding the defect of the shape of the notch 61 in (a), the upper left corner and the upper right corner are shifted to the upper right, the upper right corner and the lower right corner are shifted to the lower right, and the lower left corner is shifted to the lower left. If it is shifted, the right lower corner and the lower left corner are detected as true defects, and if it is shifted to the upper left, the upper left corner and the lower left corner are detected as true defects. Regarding the defect of the shape of the notch 62 in FIG. 8 (a), the upper right corner and the lower right corner are shifted to the upper right, and the lower right corner and the lower left corner are shifted to the lower right. However, when shifted to the lower left, the lower left corner and the upper left corner are detected as true defects, and when shifted to the upper left, the upper left corner and the upper right corner are detected as true defects. A defect including at least one of the two types of defects shown here as a minimum component can also be detected. Regarding the defect of the shape of the projection 63 in FIG. 8A, the one at the lower left corner when shifted to the upper right, the one at the upper left corner when shifted to the lower right, and the upper right when shifted to the lower left. If the part angle is shifted to the upper left, the lower right corner is detected as a true defect.
A defect including the defect shown here as a minimum component can also be detected. As described above, it is possible to detect a defect at a 270 ° corner.

Next, chipping at the 45 ° hypotenuse
An example of the present invention for a defect in the shape of a protrusion will be described. FIG.
FIG. 9A is a reference image including a 45 ° hypotenuse, and FIG.
(B) is an image to be inspected including the shape of the same part as in (a) of FIG. In FIG. 9B, the area surrounded by the solid line and the dotted line is included as the defects 71 and 72 and the protrusions 73 and 74. FIG. 9C shows a non-coincidence portion when the reference image after the alignment is shifted to the upper right by a shaded gray area. At this time, the defect 71 and the protrusion 73 of the defect included in FIG. 9B are detected as true defects. The detected true defect is shown as a dark gray area in FIG. Similarly, FIG.
FIG. 9D shows a state shifted to the lower right, and FIG.
Is detected as a true defect. Therefore, regarding the defect of the chipped / projected shape as shown in FIG. 9B, the chipped portion of the upper right hypotenuse and the protrusion of the lower left hypotenuse are shifted to the upper right, and the defect of the upper left hypotenuse is shifted to the lower right. If the protrusion and the lower right hypotenuse are shifted to the lower left, the protrusion in the upper right hypotenuse and the lower left hypotenuse are missing,
In the case of shifting to the upper left, the lack of the upper left hypotenuse and the protrusion of the lower right hypotenuse are detected as true defects. As described above, a defect at the 45 ° oblique side can be detected.

As described above, according to the pattern defect detection method of the present embodiment, various defects having a desired size can be detected by a simple process. In addition, it is possible to detect a defect including these defects as the minimum constituent elements, and it is possible to cope with a thickening or thinning of a pattern formed by continuation of these defects. As described above, the defect in the linear portion can be detected.

In the pattern defect detection method of the present invention, the processing flow of FIG. 10 can be applied particularly to a repetitive pattern. Here, a repetitive pattern in the image to be inspected is used as a reference image. That is, in the configuration diagram shown in FIG. 1, the reference image extracting unit 101 extracts, as a reference image, an image obtained by shifting the image data from the inspection image extracting unit 102 by a repetition interval.

Here, first, in step 11,
The inspection image is shifted by a repetition interval (repetition shift amount) to form a reference image.

Next, in step 12, the reference image is shifted by the above-mentioned desired shift amount (the amount instructed by the shift amount instructing section) (of course, the image to be inspected may be shifted).

Then, in step 13, a difference image in which the mismatched portion is enlarged and emphasized is obtained by calculating the difference in luminance between the image to be inspected and the reference image.

Finally, in step 14, a non-coincidence portion exceeding a certain threshold value is extracted, and a noise removal process is performed by a filter of a predetermined size to determine a true defect.

In this case, the image shift is performed in two stages of step 11 and step 12. However, the reference image (or the image to be inspected) is added by the amount obtained by adding the shift amount for defect enlargement to the shift amount in advance repeatedly. ) May be shifted.

Next, a specific example of detecting a defect in a pattern having a repetitive shape will be described. Here, the following model is used. That is, quantization is performed in the XY axis directions,
Further, the luminance value is binarized, and the filter size for removing noise is set to 3 × 3. The size of the true defect to be detected is larger than two pixels in both the horizontal and vertical directions, and the size of two or less pixels (including two pixels) is determined as a pseudo defect. At this time, the shift amount is ± 1 pixel in both the horizontal and vertical directions according to the above equation.

FIG. 11 (a) is an image to be inspected including a repetitive shape.
Projections 83, 84, 85, 86 at 2,270 ° corners
And these are shown as portions surrounded by a dotted line and a solid line. Here, FIG. 11 (a) is shifted by +6 pixels in the X-axis direction and further shifted to the upper right, and is used as a reference image, and a portion that does not match the image to be inspected is indicated by a shaded gray area in FIG. 11 (b). At this time, the protrusions 81 and the protrusions 8 included in FIG.
3 is detected as a true defect. These defects 81, 83
Is also detected by the detection method shown in the flowchart of FIG. However, in the detection flow of the repetitive shape shown in FIG. 10, the protrusions 82 in FIG.
And the protrusions 85 correspond to the true defects 82 'and 8 in FIG.
Detected as 5 '. These are defects that cannot be detected only by shifting to the upper right once by the detection method according to the flow of FIG. 1, but cannot be detected without further shifting to the lower left.

As described above, in the application of the repetitive shape, shifting to the upper right (or lower left) is detected by shifting to the upper right and lower left by the method in the flow of FIG. 1 for preparing another reference image. If the shift is similarly shifted to the lower right (or upper left), it is understood that it is equal to the sum of the shifts to the lower right and upper left in the flow of FIG. In short, in the repetitive shape, it is possible to detect the same defect as one shift by two shifts. Here, the image obtained by shifting the image to be inspected by the repetition interval is used as the reference image. However, the image obtained by shifting the image to be inspected by an integral multiple of the repetition interval may be used as the reference image.

In the present invention described above, the size of a defect to be detected can be increased or decreased by increasing or decreasing the shift amount. Therefore, the present invention can be easily applied to pattern defect detection for various uses without changing hardware. can do.

Here, at the time of defect detection, the shift amount is determined in advance according to the size of a desired defect. However, the shift amount is automatically increased or decreased according to the detected defect amount. It is also possible to detect a true defect that is not a noise component.

[0061]

As described above, according to the present invention, the reference image (or the image to be inspected) is shifted by a predetermined amount in accordance with the size of the defect to be detected. Undetected defects can be detected, and the accuracy of pattern defect detection can be improved.

Since the size of a defect to be detected can be changed by increasing or decreasing the shift amount, it is possible to cope with pattern defect detection for various uses.

Further, the reference image is shifted relative to the image to be inspected in four directions of upper right direction, lower right direction, upper left direction, and lower left direction, and a true defect is determined by these four shifts. Accordingly, all the defects to be detected can be detected, and pattern defect detection can be performed more reliably.

Further, when inspecting an object having a repetitive pattern, an image obtained by moving the image to be inspected by an integral multiple of the repetition interval is used as a reference image, so that pattern defects can be detected by a small number of shift operations. Becomes possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one configuration example of a pattern defect detection device of the present invention.

FIG. 2 is a flowchart illustrating a pattern defect detection method according to the pattern defect detection method of FIG. 1;

FIG. 3 is a diagram illustrating an example of a reference image and an image to be inspected, and a conventional pattern defect detection method.

FIG. 4 is a diagram illustrating a pattern defect detection method according to the present invention.

FIG. 5 is a diagram showing an example of application of the present defect detection method to protrusions and chipped shapes.

FIG. 6 is a diagram showing an application example of the present defect detection method for a defect having a chipped shape at a 90 ° angle portion.

FIG. 7 is a diagram showing an example of application of the present defect detection method to a protrusion-shaped defect at a 90 ° angle portion.

FIG. 8 is a diagram showing an application example of the present defect detection method for a defect at a 270 ° angle portion.

FIG. 9 is a diagram showing an application example of the present defect detection method for a defect at a 45 ° angle portion.

FIG. 10 is a flowchart illustrating a pattern defect detection method of the present invention for a repeated pattern.

FIG. 11 is a diagram showing an example in which the pattern defect detection method shown in FIG. 10 is applied to a repeated pattern.

[Explanation of symbols]

1,2,11-14,41-44,51-54,63,
73, 74, 81, 82, 83-86 Defects in projection shape 21-24, 31-34, 61, 62, 71, 72 Defects in chipped shape 103 Image shift processing unit 104 True defect judgment unit 105 Positioning unit 106 Shift Unit 107 difference unit 108 shift amount indicating unit

Claims (5)

[Claims] 1. A shift means for shifting a reference image and an inspection image from an inspection object having a pattern shape by a shift amount α in at least one direction from a position where the reference image and the inspection image are aligned. Difference means for detecting a luminance difference between the shifted reference image and the image to be inspected; and extracting a non-coincidence portion where the luminance difference is equal to or greater than a luminance threshold value, and True defect determining means for determining a true defect of the inspection object by removing noise whose size in the direction is equal to or less than m; determining a true defect when the size in the one direction on the image exceeds MAX. And a shift amount instructing unit that instructs the shift unit to indicate a shift amount α that satisfies the following equation (I) when it is desired to judge that the MIN value or less is a pseudo defect. Inspection Apparatus. m-MAX ≦ α ≦ m-MIN (I) 2. The pattern defect detection device according to claim 1, wherein when the inspection object has a pattern shape repeated at a predetermined interval, the inspection image from the inspection object is displayed at the predetermined interval. A reference image generating means for generating the reference image by shifting the reference image by an amount corresponding to the pattern defect detection apparatus. 3. A reference image and an image to be inspected from an inspecting object having a pattern shape are converted from the position where the reference image and the image to be inspected are aligned in at least one direction from the following formula (I) and the condition (I). II) is shifted by a shift amount α that satisfies the above, detecting a luminance difference between the reference image after the shift and the image to be inspected, extracting a mismatch portion where the brightness difference is equal to or greater than a brightness threshold value, and A pattern defect detection method, wherein a noise having a size of m or less in the one direction is removed from a portion to determine a true defect of the inspection object. m-MAX ≦ α ≦ m-MIN (I) If the size in the one direction on the image exceeds MAX, it is determined as a true defect, and if it is less than MIN, it is determined as a pseudo defect. (II) 4. The pattern defect detection method according to claim 3, wherein the reference image is shifted relative to the image to be inspected in four directions of an upper right direction, a lower right direction, an upper left direction, and a lower left direction. A pattern defect detection method, wherein a true defect is determined by the four shifts. 5. The pattern defect detection method according to claim 3, wherein when the inspection object has a pattern shape repeated at a predetermined interval, the inspection image from the inspection object corresponds to the interval. A pattern defect detection method, wherein the reference image is generated by shifting the reference image by an integer multiple of the amount to be processed.