JPH0777495A - Pattern inspection method - Google Patents

Abstract

PURPOSE:To provide the pattern inspection method capable of decreasing the effect of shading or the like and improving the accuracy of defect inspection. CONSTITUTION:The image of an object to be inspected formed with a synchronous pattern is fed to the right and left screens by a television camera. Omni-directional differentiation is applied to the multi-valued images 5, 6 thus fed, then they are binarized at a prescribed threshold value to extract the contour portion of the pattern. The exclusive logical sum operation is applied to the picture elements at the same coordinates of binary images 7, 8. Portions 9b having a minor difference of the prescribed reference or below of the exclusive logic image 9 obtained as a result are deleted from the original binary images 7, 8. Binary images 10, 11 thus obtained are compared with each other to detect irregularities of the pattern.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern inspection method for inspecting a pattern defect such as a reticle used for pattern formation of a semiconductor device.

[0002]

2. Description of the Related Art Conventionally, as a pattern inspection method of this kind, two TV cameras capture images of two objects to be inspected, and the captured multi-valued images are vertically, horizontally and horizontally.
A method is known in which differentiation and binarization are performed in an oblique direction, edge components are extracted, and comparison operations are performed. However, in this method, since the edge portion is emphasized by differentiation, when the pattern is dense or has a complicated shape, the defect portion is buried in the peripheral edge component when the image differentiation is performed, and the defect There was a problem that it was difficult to detect. Therefore, in order to solve such a problem, the present applicant simply binarizes the two captured images to extract the contour of the pattern,
An exclusive OR operation is performed on these binary images, and a minute difference portion of the exclusive OR image obtained as a result of which is equal to or smaller than a predetermined reference is deleted from the original binary image. A method of comparing and calculating images to detect pattern irregularities has been proposed (see Japanese Patent Application No. 3-291652). Then, according to this method, it is possible to reliably detect defects of a complicated shape, a close proximity, and a dense pattern.

[0003]

However, when such a method is used, there are the following problems. That is, when capturing an image with a television camera, it is unavoidable that some degree of shading will occur in the captured image, and since two different optical systems are used,
It is difficult to uniformly binarize the screen and to similarly binarize two screens to be compared. As a result, in order to suppress the occurrence of pseudo defects, the detection sensitivity had to be lowered and the inspection had to be performed.

The present invention has been made in view of the above points of the conventional example, and an object thereof is to provide a pattern inspection method capable of reducing the influence of shading and the like and improving the accuracy of defect detection. is there.

[0005]

According to the present invention, as shown in, for example, FIGS. 1 and 2, images of two inspected objects 1 on which synchronous patterns are formed are taken in corresponding screens (steps). ), Omnidirectional differentiation processing is performed on the captured images 5 and 6 (step), and then binarization is performed with a predetermined threshold value to extract the contour portion of the pattern (step). Exclusive-OR operation is performed between pixels having the same coordinate between them (step), and the minute difference portion 9b of the exclusive-OR image 9 obtained as a result of which is equal to or less than a predetermined reference is deleted from the original binary images 7 and 8. Then, the binary images 10 and 11 thus obtained are compared and calculated to detect pattern irregularities (step).
It is a thing.

[0006]

In the present invention having such a configuration, since the images 5 and 6 captured on the respective screens are omnidirectionally differentiated and then binarized by a predetermined threshold value, shading or the like is eliminated. It is possible to extract the contour portion of the pattern without being affected by the gradual change in luminance over the entire screen. Then, after such processing is performed, if a minute difference 9b between the binary images 7 and 8 is extracted by exclusive OR and is deleted from each of the original binary images 7 and 8, the defect with high sensitivity can be obtained. Even when the detection is performed, it is possible to suppress the occurrence of the pseudo defect due to the unevenly binarized portion or the positional shift between the images.

[0007]

Embodiments of the pattern inspection method according to the present invention will be described below with reference to FIGS.

FIG. 2 shows a schematic structure of an apparatus used in this embodiment. As shown in the figure, this device
It has a pair of television cameras 2 for picking up an image of a reticle 1 as an inspection object. The television camera 2 is a CCD camera with a microscope and is configured to capture images of the same portion of the reticle 1 on which the same pattern is formed. Then, each television camera 2 is connected to a CPU 3 including an A / D converter, a memory and the like for arithmetically processing the captured image signal. Furthermore, CPU
Display means 4 for displaying the position of the detected defect is connected to 3.

Next, the pattern inspection method of this embodiment will be described with reference to FIG.
It will be described based on. Here, the images from step to step show a process of capturing an image of the pattern on the reticle 1 on the same screen and performing digital image processing.

In the method of this embodiment, first, the image of each reticle 1 is captured by the television camera 2 to obtain multivalued images 5 and 6 (step). In this case, the protruding portion 6a of the multi-valued image 6 captured on the right screen is a defect to be detected.

Next, omnidirectional differentiation processing is performed on these multivalued images 5 and 6 by the following method (step). In the present embodiment, a mask table (differential operator) as shown in FIG. 3 is used to perform a mask operation on the entire image to approximately perform Laplacian (second derivative) processing in the vertical, horizontal, and diagonal directions. , The result is 2
Quantify. That is, E ′ = {E− (A + B + C + F + I + H + G + D) / with respect to the central pixel (E) of the 3 × 3 pixel area shown in FIG.
8} is performed. Then, this calculation is performed for all pixels in the screen. However, in view of the fact that this calculation is the second derivative, each calculation result is converted into an absolute value in preparation for the next binarization (FIG. 4, step).

The data obtained by such a differential (difference) process is not easily affected by a gradual change in luminance such as shading over the entire screen, and in particular, in the case of a pattern having a thin contour portion, shading is performed. Data that is almost the same as the data that is obtained by performing simple binarization on a screen that does not have any is obtained.

Next, the multivalued images 5 and 6 in which the edges are emphasized by the above-described processing are simply binarized by a predetermined threshold value, and only the contour of the pattern is extracted (step).
Thereby, the binary images 7 and 8 are obtained.

Generally, when a multi-valued image is captured and binarized, image distortion due to the configuration of the optical system, nonuniformity of the illumination light amount on the screen, quantization by A / D conversion, etc. Due to the influence, it is difficult to faithfully binarize the original pattern, and in most cases, a jagged image is extracted. In particular, when different images are captured from different optical systems and compared as in the present embodiment, the jaggedness occurs as a difference between the images, resulting in a pseudo minute defect. Therefore, in the present embodiment, in order to reduce the influence of such jaggedness of the binary image, a process for deleting a minute difference as described below is added. That is, pattern defects often appear as lumps when images are compared with each other, whereas jagged lines due to binarization tend to appear as fine dots or thin lines. A process to remove small differences by a kind of filtering is performed.

Specifically, first, in step 2,
The exclusive OR operation is performed on the binarized binary data between the same coordinate pixels on the left and right screens to extract the difference between the binary images 7 and 8. In this case, as can be seen from the figure, in the exclusive OR image 9, in addition to the portion 9a to be extracted as a defect, a portion 9b relating to a minute difference due to the above-described reason appears. Then, if this portion 9b is less than or equal to a predetermined amount, processing is performed to delete the portion 9b from the binarized left and right binary images 7 and 8 (step).
As a result of this processing, for example, as shown in FIG. 5, the data of the pixels of the portion 9b relating to the minute difference among the portions where the data extracted as the contour of the pattern is 1 becomes 0, and as a result, almost all except for the defective portion 11a. Binary images 10 and 11 having no difference are obtained.

The conditions for deleting the minute difference in the exclusive OR image 9 are as follows. That is, as shown in FIG. 6, 1 located at the center of any 5 × 5 pixel
When paying attention to one pixel I, if the surrounding pixels satisfy any one of the expressions (1) to (12), the pixel I is deleted. In this case, according to the formulas (1) to (4), a pixel having a width of one pixel in the vertical, horizontal, and diagonal directions is deleted, and according to the formulas (5) to (12), the vertical, horizontal, and diagonal directions are deleted. Pixels having a width of 2 pixels in the direction are deleted. Note that this deletion is performed by rewriting both the corresponding pixels on the left and right binary screens to 0 (or 1).

If such a process is performed for the entire range on the screen, the difference between the binary images 7 and 8 within the width of 2 pixels is deleted. As a result, when the exclusive OR is taken, the defect portion in which the difference appears in a lump remains, and the minute difference such as the positional deviation that causes the pseudo defect is removed.

Then, the left and right binary images 10 and 11 obtained by the above-mentioned processing are subjected to the following comparison operation to extract the defect 12 (step). That is, as shown in FIG. 7, an arbitrary 8 × 8 pixel area is cut out from the left and right screens and each pixel having the same coordinates is compared. here,
The portion of "1" is the portion extracted as the contour of the pattern. In this case, the combination of the values of the left and right pixels is (0,0), (0,1), (1,0), (1,1).
There are four possible ways. Usually the left and right values are equal (0,
0) and (1,1) are normal, and left and right values are different (0,1) and (1,0) are abnormal. Either of the values obtained by the following two equations is considered. If it is equal to or larger than the predetermined threshold value, there is an abnormality in this area, and if not, it is determined to be normal.

[0019]

[Equation 1]

Then, the area of 8 × 8 pixels is sequentially cut out by shifting one pixel vertically and horizontally over the entire range of the screen, and finally one screen is judged. The detected defect is displayed on the display means 4 of the apparatus shown in FIG.

As described above, according to this embodiment, the influence of shading and the like can be greatly reduced, so that pseudo defects are less likely to appear even if the sensitivity is increased during defect detection. As a result,
The accuracy of defect detection can be improved.

The present invention is not limited to the above-mentioned embodiment, but the value of the mask table at the time of omnidirectional differentiation processing can be changed. However, it is preferable that the weighting at that time is smaller than the normal differential processing.

Further, although the reticle used for forming the pattern of the semiconductor device has been described as an example in the above-mentioned embodiments, the present invention is not limited to this, and can be applied to inspection of various patterns.

[0024]

As described above, according to the present invention, the omnidirectional differential processing is performed on each image captured on each screen, and then binarized by a predetermined threshold value to extract the contour portion of the pattern. In addition, since the difference between the two images is compared and the irregularity of the pattern is detected after the small difference is removed by the exclusive OR, the influence of shading and the like can be greatly reduced, and as a result, the high sensitivity can be obtained. The defect detection can be performed, and the accuracy of the defect detection can be improved.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a processing procedure of this embodiment.

FIG. 2 is a block diagram showing a schematic configuration of an apparatus used in the embodiment.

FIG. 3 is an explanatory diagram showing an example of a mask table used in the same embodiment.

FIG. 4 is an explanatory diagram of an omnidirectional differentiation process of the same embodiment.

FIG. 5 is an explanatory diagram showing a process of deleting a minute difference between binary images in the embodiment.

FIG. 6 is an explanatory diagram showing a condition for deleting a minute difference of an exclusive OR image in the example.

FIG. 7 is an explanatory diagram showing pixels cut out for defect extraction in the example.

[Explanation of symbols]

 1 Inspected object 5, 6 Multi-valued image 7, 8, 10, 11 Binary image 9 Exclusive OR image 12 Defect

Claims (1)

[Claims] 1. An image of two inspection objects on which synchronous patterns are formed is captured in respective corresponding screens,
After performing an omnidirectional differentiation process on each of the captured images, the image is binarized with a predetermined threshold value to extract the contour portion of the pattern, and the exclusive OR operation is performed on the same coordinate pixels between these binary images. , The minute difference part of the exclusive OR image obtained as a result, which is equal to or less than a predetermined reference, is deleted from the original binary image, and the binary images thus obtained are compared and calculated to detect the irregularity of the pattern. A pattern inspection method characterized by the above.