JPH06295330A - Picture processing method - Google Patents

Abstract

PURPOSE:To improve the size measuring accuracy of an objective pattern and the detection accuracy of pattern existence, a defect, etc., by processing picture data with uneven gradation, correcting SEM picture data including the objective pattern and removing the unevenness of gradation. CONSTITUTION:This picture processing method is provided with the 1st step for scanning a pattern (A) to be quantized with an electronic beam 12 and obtaining the 1st picture data (B) based upon a secondary electronic signal S3, the 2nd step for executing space filtering processing by using a filter matrix at least over N times the range of one picture element in the 1st picture data and obtaining the 2nd picture data (C) and the 3rd step for adding the mean value of the gradation values of the 1st picture data to a difference between the 1st and 2nd picture data and constituted so as to obtain picture data from which the influence of background gradation is removed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method, and more particularly to processing each SEM image data obtained by scanning an electron beam on a fine circuit pattern of the LSI with an electron microscope (SEM). This is for improving the reliability in the dimension measurement of the pattern formed in the manufacturing process and the detection of foreign matters and defects.

[0002]

2. Description of the Related Art Conventionally, when inspecting a fine circuit pattern formed on a substrate in each LSI manufacturing process, S
EM images have been used. The SEM image is raster scan image data obtained by scanning the sample including the circuit pattern with an electron beam.

In the LSI manufacturing process, image processing such as averaging processing and spatial filtering processing is performed on this SEM image to smooth the image and detect pattern dimensions, pattern defects, contamination by foreign substances, and the like. Running the algorithm.

As described above, in the conventional image processing method, noise is removed from the obtained SEM image data by performing weighted average processing or spatial filtering processing on the SEM image data, and accurate pattern inspection or the like is performed. It is expected.

[0005]

In the conventional image processing method, noise removal is taken into consideration, but due to charge-up, secondary electron detection method, etc., SEM image data has a gradation of gradation and non-uniformity. In this case, there is a problem that this cannot be corrected and cannot be dealt with.

That is, S obtained by scanning the electron beam
In the EM image, the potential may not be uniform due to charge-up or the like even if the sample shape or material does not change within the electron beam irradiation surface. For this reason, the gradation is often non-uniform over the entire SEM image obtained. For example, the left and right sides of the screen are often dark and the right side is bright. Further, depending on the secondary electron detection method, the gray value of a part of the image may be larger than that of the other part, or the gray level may be uneven on the left and right of the image.

Therefore, in the SEM image data,
The gray value of the board part where no pattern exists may be higher than the gray value of the peak value of the pattern edge part,
Line profile data will also be out of scope. Therefore, when attempting to measure a pattern dimension using such SEM image data, erroneous detection of an edge or variation in the measurement result often occurs.

Then, at the time of pattern detection and defect detection,
If a gray level close to the gray level of the target pattern exists in the background part other than the target pattern, the problem that the target pattern and the background part cannot be clearly separated when binarization processing is performed and accurate detection cannot be performed. Will happen.

The present invention solves the problems of the prior art as described above, corrects the SEM image data including the target pattern by processing the image data having uneven gradation, and eliminates the unevenness of gradation. Accordingly, it is an object of the present invention to obtain an image processing method capable of improving the dimensional measurement accuracy of a target pattern and the detection accuracy of the existence and defects of the pattern.

[0010]

In order to achieve the above object, the present invention provides a first step of scanning a pattern to be quantized with an electron beam to obtain first image data, and the first step. A second step of obtaining second image data by subjecting the image data to spatial filtering processing using a filter matrix for smoothing a pixel in the first image data by a plurality of pixels around the pixel; A third step of obtaining corrected third image data by adding the average value of the grayscale values of the first image data to the difference between the first image data and the second image data An image processing method comprising:

[0011]

According to the above-mentioned means, in the image processing method of the present invention, the first image data corresponding to the pattern to be quantized is obtained in the first step, and the first image data is obtained in the second step. Spatial filtering without weighting is performed to obtain the second image data that reflects the shading distribution of the background of the image, and in the third step, the difference between the first image data and the second image data is obtained. By adding the average value of the shade of the first image data to this, the third image data in which the influence of the shade of the background is removed is obtained.

[0012]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic block diagram of an apparatus for executing an image processing method according to an embodiment of the present invention.
As shown in the figure, in the SEM 1, a sample 5 such as an LSI substrate having a fine pattern formed thereon is placed on the XY stage 4. The SEM 1 projects an electron beam toward the sample 5, and the electron beam is deflected by the deflector 2 and scans the sample 5. Secondary electrons are emitted from the sample 5 projected with the electron beam, and the secondary electrons are detected by the detector 3. The deflector 2 is controlled and operated by the deflection signal S2 from the scan generator 6. Secondary electron signal S detected by the detector 3
3 is input to the image processing device 8. Further, a deflection synchronization signal S1 synchronized with the deflection signal S2 given to the deflector 2 is given from the scan generator 6 to the image processing device 8.
The image processing device 8 processes the secondary electronic signal S3,
The processing result is output to the image processing device 8. On the other hand, computer 7
Controls the image processing device 8 and receives the image processing result from the image processing device 8 to execute an analysis algorithm of the sample 5, that is, processing such as pattern dimension measurement, pattern detection, and defect detection.

The operation of the above-described structure will be described below with reference to FIGS. By the way, FIG. 2 shows a cross-sectional view of the sample 5 in FIG. 1, which is formed by forming the pattern 11 on the substrate 10. Further, FIG. 3 is an explanatory diagram of a process of image processing, and FIG.
Is the scanning position of the surface of the sample 5 by the electron beam 12,
(B) is a line profile of the first image data based on the secondary electron signal S3 obtained from the SEM1, (C) is a line profile of the second image data obtained by processing the first image data, ( D) shows the line profiles of the finally obtained third image data.

In the SEM 1, the XY stage 4 is moved to bring the portion of the sample 5 to be image-analyzed into the electron beam scanning range. Sample 5 in SEM1
The electron beam projected toward is deflected by the deflector 2 provided with the deflection signal S2 from the scan generator 6, and scans a specific range on the sample 5.

As shown in FIG. 2, the sample 5 has the pattern 11 formed on the substrate 10, and the intensity of the secondary electrons emitted corresponding to the electron beam changes based on the state of the surface thereof. Come on. The secondary electrons are detected by the detector 3 and sent to the image processing device 8 as a secondary electron signal S3.

The image processing device 8 synchronizes with the deflection synchronization signal S1 from the scan generator 6, that is, in synchronization with the deflection state of the electron beam by the deflector 2, and outputs the secondary electron signal S3.
Is digitized (for example, converted into 8-bit, 256-gradation image data) and is captured in a frame memory (image memory) not shown. That is, the surface condition of the substrate 10 is rasterized and converted into image data.

The image data from the SEM1 thus obtained is used as the first image data. The scanning results of a plurality of scans are averaged and further smoothed by the spatial filtering process to obtain S / N. Image data with improved
Image data may be used.

The first image data obtained as described above is as shown in FIG. 3B for the surface condition of the sample 5 as shown in FIG. That is, pattern 1
In addition to the shade of the edge portion of 1, the line profile has a shade gradient in the background portion where the pattern 11 does not exist. This is due to the charge-up of the sample 5, the secondary electron detection method by the detector 3, and the like.

Next, the image processing device 8 processes the first image data, but performs the unweighted spatial filtering process on the first image data shown in FIG. 3B. Here, “no weight” means filter processing in which each element is “1”. The filter size of the spatial filtering process is about (15 × 15) or more, for example, (21 × 21). By the way, this filter size may be of a size such that the grayscale signal of the pattern 11 is sufficiently filled, that is, smoothed. Therefore, the filter size can generally be (N × N) or (N × M). Here, N and M are both positive integers, and N ≠ M.

As a result of the above processing, as shown in FIG. 3C, the second image data obtained by blunting the grayscale signal corresponding to the pattern 11 part and extracting only the overall grayscale gradient is obtained. be able to. That is, the second image data is image data that reflects the light and shade distribution of the background portion on the surface of the substrate 10.

Next, the second image data is subtracted from the first image data to obtain difference image data corresponding to 11 parts of the pattern from which the shade components included in the first image data have been removed. On the other hand, the average value of the grayscale values of the first image data is obtained, and this is the background image data corresponding to the non-gradient grayscale of the background portion of the first image data.
Then, by adding the background image data to the previously obtained difference image data, the third image in which the shade of the background portion is added is added.
The image data of can be obtained. As shown in FIG. 3D, the third image data obtained in this way becomes an image in which the unevenness of the shade is corrected without the density gradient or the density difference distribution in the background portion.

The first image data, the second image data, and the third image data obtained as described above can be sequentially monitored by the monitor 9 throughout the process.

Now, the third image data in which the unevenness of light and shade is corrected by the image processing device 8 is subjected to a binarization process under the control of the computer 7, and the pattern 11 to be inspected.
It is used to detect patterns and defects, measure pattern dimensions, etc. by clearly separating the background portion from the background.

Further, by extracting the line profile of the third image data and executing the pattern dimension measuring algorithm, it becomes possible to detect the peak of the pattern edge portion with high accuracy without erroneous detection.

In the above embodiment, image processing is performed by the LSI.
Although the case of using it for the pattern inspection in the manufacturing process is exemplified, it can be practically used for the purpose such as pattern inspection performed for other purposes and image recognition.

Further, as described above, in the above-described embodiment, the case where the filter size of 21 × 21 of the pixels of the image data is applied in the spatial filtering process has been exemplified, but this is based on the quantization size and the pattern size to be detected. Depends on the difference. In the case of a typical SEM, this is a filter size of at least about 15 × 15.

[0027]

As described above, according to the image processing method of the present invention, when pattern dimension measurement, pattern and defect detection are performed using SEM image data obtained by scanning an electron beam, There is an effect that the unevenness of the concentration caused by the charge-up or the secondary electron detection method is corrected to prevent the measurement variation and the erroneous detection.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of an apparatus for realizing an image processing method according to an embodiment of the present invention.

FIG. 2 is a cross-sectional view of the sample in FIG.

FIG. 3 is an explanatory diagram for explaining the method of the present invention.

[Explanation of symbols]

 1 SEM 2 deflector 3 detector 4 XY stage 5 sample 6 scan generator 7 computer 8 image processor 9 monitor 10 substrate 11 pattern 12 electron beam

Claims (5)

[Claims] 1. A first step of scanning a pattern to be quantized with an electron beam to obtain first image data; and, for the first image data, a pixel in the first image data is A second step of performing a spatial filtering process using a filter matrix smoothed by a plurality of pixels around the pixel to obtain second image data, and a difference between the first image data and the second image data And a third step of obtaining corrected third image data by adding an average value of the grayscale values of the first image data to the image processing method. 2. The range of the filter matrix is (N ×
The method of claim 1, wherein N), where N is a positive integer. 3. The range of the filter matrisk is (N ×
M) [N, M are positive integers and N ≠ M],
the method of. 4. The image processing method according to claim 1, wherein the range of the filter matrix is at least 15 times as large as one pixel. 5. The image processing method according to claim 1, wherein the pattern to be quantized is a fine circuit pattern, and the first image data is an SEM image.