JPS63118884A - Pattern defect detector - Google Patents

Abstract

PURPOSE:To easily and appropriately detect defects by expanding a contour by a fixed width based on a contour binarization image, contracting the expanded contour by width larger than the expanding one and detecting the defects of a pattern. CONSTITUTION:A space differentiation processing part 3 differentiates the pattern, thereby obtaining an image where the brightness 18-1 of the contour of an upper layer pattern and that 21-1 of the contour of a lower layer pattern are larger than other parts. If a threshold is set, the binarization image of two patterns with contours 18-2 and 21-2 can be obtained. An expansion processing part 5 expands the binarization image of the contours obtained in such a way, and a contract processing part 6 contracts said image. The contract ratio at that time equals or exceeds the contour width W-2 (=W-1+n) of a normal part after expansion. A defect decision part 7 checks whether the image has an extracted part or not.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a pattern defect detection device, and more particularly to a pattern defect detection device that optically detects and inspects a target pattern.

(Prior Art) Conventionally, one of the methods for pattern inspection is the "expansion/contraction method." Examples of this method include, for example, "Flaw Extraction Device Method for Complex Patterns," Journal of the Institute of Electrical Engineers of Japan (C)
νo1.94-CN15, PP89~96°1974
There was something written in.

The principle of this conventional pattern defect detection method will be explained below with reference to FIG.

In FIG. 2(a), reference numeral 8 indicates a background, and reference numeral 9 indicates a target pattern, and this target pattern 9 has a black flaw 11-1 and a white flaw 10-l.

Now, as shown in FIG. 2(b), the target pattern 9
When inflated uniformly, the white wound 1O-1 is crushed,
When this is contracted, as shown in Figure 2 (c),
Only the black scratch 11-1 is restored. Therefore, Fig. 2 (a
) and each corresponding point in FIG. 2(c), and by determining different portions, it is possible to extract a white minute portion 10-3 as shown in FIG. 2(d). Similarly,
First, as shown in FIG. 2(f), when the target pattern 9 in FIG. 2(a) is uniformly shrunk, the black scratches 11-
8 disappears, and when the target pattern 9 is expanded again, only the white scratch 1O-1 is restored as shown in FIG. 2(g). Here, by comparing each corresponding point in FIG. 2(a) and FIG. 2(g) and determining the different parts, the second
As shown in Figure (h), only the black minute portion 11-3 is restored.

Therefore, by merging FIGS. 2(d) and 2(h), it is possible to extract minute portions 10-z and 114 representing all the flaws, as shown in FIG. 2(e).

FIG. 3 is a functional block diagram of pattern defect detection using the conventional expansion and contraction method. A second contraction processing section 26, a first contraction processing section 27, and a first comparison processing section 28 are provided, and a second contraction processing section 29, a second expansion processing section 30, and a second comparison processing section 31 are provided in parallel. Furthermore, a defect determination section 32 that combines the first comparison processing section 28 and the second comparison processing section 31 is provided.

(Problems to be Solved by the Invention) However, the conventional method has the following problems.

FIG. 4 shows a pattern before etching during the printed wiring board manufacturing process and after photoresist development, as one example in which it is difficult to apply the conventional method.

Here, FIG. 4(a) is an image when a pattern is optically detected by a normal method, and includes a photoresist pattern 12-3, a background 13-+, and its boundary 14-1.
and the defect 15, and FIG. 4(b) shows its brightness cross section. In this case, due to the high transparency of the photoresist, the brightness of the pattern is 12-t and the brightness of the background is 13-!, as shown in FIG. 4(b). There was a problem in that the difference between the patterns and the background almost disappeared, making it impossible to stably obtain a binarized image divided into a pattern and a background, and that defects could not be detected using conventional methods.

FIG. 5 shows a defect in a two-layer wiring board as another example where it is difficult to apply the conventional method.

Here, FIG. 5(a) shows an image when the pattern is optically detected, and shows the upper layer pattern 16-1 and its defect 1.
7. Lower layer pattern 19-I and its defects 20. Background 22-
I, and FIG. 5(b) shows its brightness cross section. In this case, the dark value at the time of binarization is A in Fig. 5(b).
If the brightness is set between the brightness 16-1 of the upper layer pattern and the brightness 19-2 of the lower layer pattern as shown in FIG. Also,
The brightness of the lower pattern is 19-! as shown in B in FIG. 5(b). and background brightness 22-! If set between
Defects 17 occurring in the upper layer pattern as shown in FIG. 3(d) cannot be extracted. For this reason, in the conventional method, it was necessary to take measures such as changing the darkness value and performing the inspection twice.

Furthermore, in the conventional method, as shown in the above-mentioned FIG. 3, the minute portions that become defects are extracted from FIGS. 2(C) and (g), and FIGS. In order to seek
Figure 2 (a) and Figure 2 (c) and Figure 2 (a) and Figure 2
Comparison processing with Figure (g) was required.

In this way, in the conventional method, the processing is not a simple sequential process, and in particular, since the images in the image storage section 25 are referred to in the first comparison processing section 28 and the second comparison processing section 31, This poses a problem in that it is not suitable for pipelining, which is used to speed up processing.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a pattern defect detection device that eliminates the above-mentioned problems and can easily and appropriately detect defects using simple sequential processing.

(Means for Solving the Problems) In order to solve the above problems, the present invention, in a pattern defect detection device, binarizes a contour image obtained directly or by differential processing and expands it by a certain width. After the expansion process, a contraction process is performed to contract the expanded outline by a width larger than the expansion width.

(Function) According to the present invention, with the above configuration, it is possible to detect objects for which only the outline of the pattern can be obtained by optical methods, or to obtain two images with one darkness value, such as a pattern with a multilayer structure.
Even when the necessary pattern cannot be obtained by converting into values, defects can be detected easily and appropriately. Also,
Since it is a simple sequential process, it can be easily pipelined to speed up the process.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a functional block diagram showing a pattern defect detection processing method according to an embodiment of the present invention. 3. Binarization processing unit 4, expansion processing unit 5, contraction processing unit 6,
It consists of a defect determination section 7.

In this image input section 1, the pattern to be inspected is detected by an imaging means such as an industrial television camera, and a video signal of the pattern to be inspected is obtained. This video signal is converted into a digital signal by the A/D converter 2.

Next, if the pattern to be inspected cannot be contour extracted by binarization processing, the contour is extracted by the spatial differentiation processing section 3, and the contour is straightened by the * 2 + a processing section 4, and the contour is A valued image is obtained.

This process will be explained using two examples shown in FIGS. 4 and 5 mentioned above.

FIG. 6 shows an example after photolithography development, where it is difficult to separate the pattern and background by binarization.

Note that FIGS. 6(a) and 6(b) are similar to FIGS. 4(a) and 4(b).

Here, as shown in the brightness cross section in FIG. 6(b), the brightness is lower at the boundary 14-2 between the pattern and the background than in other parts, so the darkness value is set to A in FIG. 6(b). By setting as shown in FIG. 6(C), a binarized image of the outline of the pattern can be obtained.

Furthermore, Fig. 7 shows defects in a two-layer wiring board, and Fig. 7(a) and Fig. 7(b) show defects in Fig. 5 (
This is similar to FIG. 5(a) and FIG. 5(b).

Since a contour image cannot be obtained from this pattern by direct binarization, by differentiating it in the spatial differentiation processing section 3 shown in FIG.
As shown in c), the brightness of the contour of the upper layer pattern is 18
-9 and the brightness 21-8 of the outline of the lower pattern is higher than other parts of the image. The threshold value is set to C in Fig. 7(c).
If the settings are made as shown in FIG. 7(d), a binarized image of the contour 18-z+21-2 between the two patterns can be obtained.

The binary image of the contour thus obtained is expanded by the expansion processing section 5 shown in FIG.

The expansion width n at this time is set so that the contours do not come into contact with each other due to expansion in a normal pattern.

FIGS. 8(a) and 8(b) show an example of this expansion process, and FIG. 8(a) shows a protrusion defect 33. This is a binarized contour image of I and contour width W-, which includes an undersized pattern width defect 34-1.

Now, as shown in FIG. 8(b), when the tube is uniformly inflated with an expansion width n, the contour width W-2 of the normal portion becomes w-1+n.

On the other hand, the contour width W'-2 of the expanded protrusion defect 33-2 and the contour width W'' of the expanded pattern width insufficient defect 34-1
-2 is wider than the contour width W-, +n of the inflated normal part.

Next, this expanded contour binarized image is contracted by the contraction processing section 6 shown in FIG.

The contraction width at this time is the contour width of the normal part after expansion, Vl.
, (=W-, +n) or higher.

FIG. 8(c) shows an example of this shrinkage process for FIG. 8(b), and by shrinking the shrinkage width; W4+α (α≧O),
The outline of the normal portion in FIG. 8(b) completely disappears, and only the protrusion defect 33-3 and the pattern width undersizing defect 34-1 are extracted.

Finally, the defect determination unit 7 shown in FIG. 1 examines the presence or absence of the extracted portion in the image (in the case of a binary image with positive logic of 0 and 1, a pixel with a value of 1), If there is no error, the process is determined to be normal, and if it is, it is determined to be abnormal and the location of the defect is output as necessary.

Note that the present invention is not limited to the above embodiments,
Various modifications are possible based on the spirit of the present invention, and these are not excluded from the scope of the present invention.

(Effects of the Invention) As described above in detail, according to the present invention, the contour 2
Based on the digitized image, the contour is expanded by a certain width, and then the expanded contour is contracted by a width larger than the expansion width to detect pattern defects. This method is easy to use for objects for which only the outline of the pattern can be obtained using optical methods, or for cases where the required pattern cannot be obtained by binarization using a single dark value, such as a pattern with a multilayer structure. Furthermore, pattern defects can be detected appropriately.

(2) Since simple sequential processing is required, pipelining for speeding up processing can be easily promoted.

[Brief explanation of the drawing]

Fig. 1 is a functional block diagram showing a pattern defect detection processing method according to an embodiment of the present invention, Fig. 2 is a principle diagram of the conventional expansion and contraction method, and Fig. 3 is a functional block diagram of the conventional expansion and contraction method. , FIG. 4 is a diagram explaining the first problem of the prior art, FIG. 5 is a diagram explaining the second problem of the prior art, and FIG. 6 is a diagram explaining the first problem of the present invention.
Fig. 7 is an explanatory diagram of the second outline binarized image detection method of the present invention, and Fig. 8 is an explanatory diagram of the defect extraction method using expansion and contraction processing of the present invention. It is an explanatory diagram. 1... Image input section, 2... A/D conversion section, 3...
Spatial differentiation processing section, 4... Binarization processing section, 5... Expansion processing section, 6... Shrinkage processing section, 7... Defect determination section, 1
2-I...pattern, 12-! ...Pattern brightness, 13-1...Background, 13-! ...Background brightness, 14
-1... Boundary, 14-2... Boundary brightness, Ics...
・Extracted boundary, 15...defect, 16-t...
Upper layer pattern, 16-3... Brightness of upper layer pattern, 1
7... Defect of upper layer pattern, 18-1... Brightness of outline of upper layer pattern, 18-1... Outline of upper layer pattern,
19-1... Lower layer pattern, 19-2... Luminance of lower layer pattern, 20... Defect of lower layer pattern, 21-1
... Brightness of contour of lower layer pattern, 21-8 ... Outline of lower layer pattern, 22-1 ... Background, 22-2 ... Brightness of background, 33-1 ... Protrusion defect, 33-2 ... Expanded protrusion defect, 33-3... Extracted protrusion defect, 34-1... Pattern width too small defect, 34-2... Expanded pattern width too small defect, 34-1...・Extracted pattern width defect.

Claims (1)

[Claims] (a) means for obtaining a binarized image in which the outline of a two-dimensional pattern of the object is extracted; (b) means for expanding the outline extracted by the means by a certain width; ( A pattern defect detection device characterized by comprising: c) means for contracting the expanded contour by a width larger than the expanded width; and (d) means for extracting only the defective portion using the means.