KR20240101589A - Image processing method, device, and system for semiconductor electron beam defect monitoring - Google Patents

Abstract

In the image processing method, device, and system for semiconductor electron beam defect monitoring, the image processing method for semiconductor electron beam defect monitoring includes: S100 step of acquiring an original image and performing filtering on the original image to obtain a filtered image; Step S200, obtaining the filtered image and sorting the filtered image using a sorting algorithm to obtain a template image; Step S300 of obtaining a difference image by performing an image subtraction operation on the template image and a preset reference image; Step S400 of calculating the variance of the difference image, comparing and judging the variance with a preset value, and obtaining and outputting a comparison decision result. The above processing method can improve the success rate of image alignment during the defect detection process, remove unnecessary details during image alignment through filtering, and accurately judge the alignment result using differential images, allowing various alignment algorithms to be used interchangeably. and improve the accuracy of the sorting algorithm by implementing each advantage.

Description

Image processing method, device, and system for semiconductor electron beam defect monitoring

This disclosure relates to the field of semiconductor technology, and more particularly to image processing methods, devices and control systems for semiconductor electron beam defect monitoring, and non-volatile computer-readable storage media.

Yield testing is an essential step in the production and manufacturing process of semiconductor chips. Yield is generally the most important indicator of process improvement and an important criterion for testing the performance of major foundries and IDMs. Yield inspection typically compares images of two or more different dies. Ideally, the images of different dies on a wafer should be completely identical, but in the actual production process, the images of different dies may be different due to defects, and these image differences are immediately recognized as defects.

Currently, common defect detection algorithms first need to align pictures of different dies, and image alignment algorithms mainly rely on original image information. Because various noises and equipment cause fluctuations in imaging parameters at different times, the quality of the original image is easily affected by noise and equipment fluctuations, leading to image alignment failure and making subsequent yield inspection work impossible to perform smoothly. This reduces the accuracy of image alignment.

At the same time, current defect detection processes do not check the alignment results after image alignment. Second, if a defect occurs in a large area, the characteristics of the image alignment algorithm will be affected depending on the presence or absence of the defect.

Additionally, a typical sorting process typically uses a single sorting algorithm. Since the points of importance are different depending on the principles of different algorithms, it is inevitable that performance will deteriorate significantly and alignment failure will occur in certain situations. In addition, a general measure of the amount of image information is called information entropy, but in test results of thousands of pictures, it is difficult to determine whether the images are aligned using information entropy, and it is difficult to determine whether the images are aligned, and the difference image of different group images It was found that the correlation between the information entropy of (difference image) and whether the images were aligned was not clear.

Additionally, the prior art generally relies primarily on original images or simply blur-filtered images in the image alignment step. The overall light/dark unevenness of the image cannot be removed. Even if multiple defects occur, the effects of the defects cannot be selected and eliminated. Additionally, the conventional method cannot alternately use two or more alignment algorithms to reduce the failed alignment rate due to a lack of evaluation of the image alignment results.

In this respect, the present disclosure proposes an image processing method, device, and control system for semiconductor electron beam defect monitoring, and a non-volatile computer-readable storage medium, which can improve the success rate of image alignment in the defect detection process, especially This is to improve the success rate of image alignment when defects occur in a large area during the image comparison process, image quality deteriorates due to noise, or overall contrast is uneven. The present invention aligns filtered images using the Fourier transform, determines whether the alignment is successful based on the difference value of the two images, has the ability to accurately determine whether the images are aligned, and uses two or more alignment algorithms to It can be used, and by implementing each advantage, it greatly improves the accuracy of the sorting algorithm.

According to one aspect of the present disclosure, in an image processing method for monitoring semiconductor electron beam defects,

Step S100 of acquiring an original image and performing filtering processing on the original image to obtain a filtered image;

Step S200, obtaining the filtered image and sorting the filtered image using a sorting algorithm to obtain a template image;

Step S300 of obtaining a difference image by performing an image subtraction operation on the template image and a preset reference image;

An image processing method for monitoring semiconductor electron beam defects is provided, including a step S400 of calculating the variance of the difference image, comparing and judging the variance with a preset value, and obtaining and outputting a comparison decision result. do.

In one possible implementation form, optionally, obtaining a filtered image by filtering the original image in step S100 includes:

Step S110, obtaining a spectral image by Fourier transforming the original image into spectral space;

Step S120, inputting the spectral image to a Gaussian band-pass filter and performing filtering on the image spectrum to obtain a filtered spectral image; Here, the specific form of the Gaussian band-pass filter is as follows:

here:

D(u,v) is the distance to the origin in spectral space,

D ₀ is the center frequency of the pass band, W is the width of the pass band,

(u,v) is the pixel coordinate of the spectral image after Fourier transformation, the center point pixel of the spectral image is the origin of the coordinate;

Step S130 of receiving the filtered spectrum image and performing inverse Fourier transform on the filtered spectrum image to obtain the filtered image.

In one possible implementation form, optionally, obtaining a difference image by subtracting the template image from the reference image in step S300 includes:

Obtaining the sorted result processed in step S200;

Receiving the alignment result and moving the template image according to the alignment result to obtain a template translation image;

and obtaining a difference image by subtracting the template translation image from the reference image.

In one possible implementation form, optionally, calculating the variance of the difference image in step S400, comparing and judging the variance with a preset value, and obtaining and outputting a comparison decision result include,

calculating variance by obtaining grayscale values of the difference image;

If the variance is less than the preset value, image alignment is successful, and if the variance exceeds the preset value, image alignment is failed, so re-sort according to the preset alignment processing method. A step of comparing and determining values;

It includes obtaining and outputting a comparison decision result.

According to another aspect of the present disclosure, an image processing device for monitoring semiconductor electron beam defects, comprising:

It includes an original image filtering processing module, an alignment processing module, a differential image acquisition module, and a distributed calculation and judgment module that are electrically connected in that order,

The original image filtering processing module acquires an original image, performs filtering processing on the original image, and obtains a filtered image;

The alignment processing module acquires the filtered image, sorts the filtered image using a sorting algorithm, and obtains a template image;

The difference image acquisition module acquires a difference image by performing an image subtraction operation on the template image and a preset reference image;

The dispersion calculation and determination module calculates the dispersion of the difference image, compares the dispersion with a preset value, determines the dispersion, and obtains and outputs a comparison decision result. An image processing device for monitoring semiconductor electron beam defects is provided.

In one possible implementation form, optionally, the raw image filtering processing module may comprise:

a filtered spectrum image acquisition module that inputs the original image into a Gaussian band-pass filter and performs filtering on the image spectrum to obtain a filtered spectrum image; Here, the specific form of the Gaussian band-pass filter is as follows:

here:

D(u,v) is the distance to the origin in spectral space,

D ₀ is the center frequency of the passband, W is the width of the passband,

(u,v) is the pixel coordinate of the spectral image after Fourier transformation, the center point pixel of the spectral image is the origin of the coordinate;

and a spectrum image inverse Fourier transform processing module, which receives the filtered spectral image and performs inverse Fourier transform on the filtered spectral image to obtain the filtered image.

In one possible implementation form, optionally, the differential image acquisition module may comprise:

A sorting result acquisition module, which acquires the sorted sorted result in step S2;

an image translation module that receives the alignment result and moves the template image according to the alignment result to obtain a template translation image;

and an image subtraction processing module configured to obtain a difference image by subtracting the template translation image from the reference image.

In one possible implementation form, optionally, the distributed calculation and judgment module may comprise:

a variance calculation module that calculates variance by obtaining a grayscale value of the difference image;

If the variance is less than the preset value, image alignment is successful, and if the variance exceeds the preset value, image alignment is failed, so re-sort according to the preset alignment processing method. a distributed comparison judgment module that compares and determines values;

It includes an output module, which obtains and outputs comparison judgment results.

According to another aspect of the present disclosure, in a control system,

processor;

Includes a memory that stores instructions that the processor can execute,

An image processing method control system for semiconductor electron beam defect monitoring is further provided, wherein the processor, when executing the executable instructions, is configured to implement the image processing method for semiconductor electron beam defect monitoring described above.

According to another aspect of the present disclosure, there is further provided a non-volatile computer-readable storage medium storing computer program instructions, wherein the method described above is implemented when the computer program instructions are executed by a processor.

The technical effects of this application are as follows:

The present invention includes the steps of obtaining an original image and performing filtering on the original image to obtain a filtered image; Obtaining the filtered image and sorting the filtered image using a sorting algorithm to obtain a template image; Obtaining a difference image by performing an image subtraction operation on the template image and a preset reference image; By calculating the variance of the difference image, comparing and judging the variance with a preset value, and obtaining and outputting a comparison judgment result, the success rate of image alignment can be improved in the defect detection process, especially in the image comparison process. When a defect occurs in a large area, the image quality deteriorates due to noise, or the overall contrast is uneven, the present invention aligns the filtered images using Fourier transform and successfully aligns them according to the difference value of the two images. By having the ability to accurately determine whether images are aligned, removing unnecessary details during image alignment through filtering, and accurately judging the alignment results using difference images, various alignment algorithms can be used interchangeably. , Implementing each advantage significantly improves the success rate of the image alignment algorithm of the present invention, thereby significantly improving the accuracy of the alignment algorithm.

Other features and aspects of the present invention will become apparent from the detailed description of exemplary embodiments hereinafter with reference to the accompanying drawings.

The accompanying drawings, which are included in and constitute a part of the specification, are intended to explain the principles of the present invention and, together with the specification, show exemplary embodiments, features and aspects of the present invention.
Figure 1 shows a schematic flowchart of an image processing method for monitoring semiconductor electron beam defects of the present invention.
Figure 2 shows an original die image of a chip to be processed according to the present invention.
Figure 3 shows a band-pass filtered image of the present invention.
Figure 4 shows a difference image of two aligned images according to the present invention.
Figure 5 shows a difference image of two unaligned images of the present invention.
Figure 6 shows the distribution of pixel difference values (average value removed) of two aligned images according to the present invention.
Figure 7 shows the distribution of pixel difference values (average value removed) of two unaligned images of the present invention.

Hereinafter, various exemplary embodiments, features and aspects of the present disclosure will be described in detail with reference to the accompanying drawings. In the drawings, identical reference numerals indicate elements with identical or similar functions. Although various aspects of the embodiments are shown in the drawings, unless specifically noted, the drawings are not necessarily drawn to scale.

As used herein, the term “exemplary” means “for example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferable or superior to other embodiments.

Additionally, to better explain the present disclosure, numerous specific details are provided in the detailed embodiments below. Those skilled in the art will understand that the present disclosure may be practiced without specific details. In some embodiments, methods, means, components, and circuits known to those skilled in the art are not described in detail in order to emphasize the spirit of the present disclosure.

Example 1

As shown in Figure 1, according to one aspect of the present disclosure, an image processing method for semiconductor electron beam defect monitoring is provided comprising the following steps:

Step S100 of acquiring an original image and performing filtering processing on the original image to obtain a filtered image;

Figure 2 shows the original die image of the chip. First, band-pass filtering of Fourier spectrum space is performed on the original image.

In the spectral space, a Gaussian filter is used to filter the low-frequency and high-frequency components of the image to obtain a band-pass filtered image of the original image, that is, a filtered image. Filtering the high-frequency components of the Fourier spectral space can greatly reduce the impact of noise and artifacts on the final alignment result.

Specifically, the original image is Fourier transformed to obtain a spectral image (same size as the original image). Then, a band-pass filtering process is performed on the spectral image, that is, the low-frequency part and the high-frequency part are suppressed (the specific suppression method uses a Gaussian filter, which will be described in detail later), and the mid-frequency band is maintained. Afterwards, the band-pass filtered spectrum image is inverse Fourier transformed to obtain a band-pass filtered image (filtered image) of the original image. Compared to the original image, the filtered image only has a rough outline and lacks the details of the original image.

Due to variations in detection equipment parameters, noise in multiple random images can degrade the performance of alignment algorithms, often resulting in uneven contrast between different images or in unknown areas of one image. Accordingly, by filtering the low-frequency components of the Fourier spectral space, the impact of uneven brightness and darkness on image quality can be effectively reduced.

Step S200, obtaining the filtered image and sorting the filtered image using a sorting algorithm to obtain a template image;

The present invention sorts the filtered images using the Fourier transform, determines whether the alignment is successful according to the difference value between the two images, and performs a sorting process on the filtered images using a comprehensive sorting algorithm, thereby performing various sorts of sorting. Algorithms can be used interchangeably, and the success rate of the image alignment algorithm of the present invention is significantly improved by implementing the advantages of each, thereby greatly improving the accuracy of the alignment algorithm.

After processing according to the above sorting algorithm, one template image is obtained, and the template image is a band-pass filtered image, which is an image obtained by sorting the filtered image.

Step S300 of obtaining a difference image by performing an image subtraction operation on the template image and a preset reference image;

The reference image is preset in advance, and the reference image is calculated according to theoretical values, and one is selected from the image library as a reference image for image subtraction operation and used to calculate the difference value, or according to the standard image in computer simulation. One can be selected as the reference image for the image subtraction operation, and is specifically selected by the user.

The template image is subtracted from the reference image, and the subtraction involves translating the template image and then subtracting the pixel luminance values of the template image and the reference image to obtain a single difference image.

A method is provided to determine whether the images are aligned by creating difference values for the aligned images and inspecting the difference images. When two images are overlapped and aligned to create a difference value, the image of the overlapping portion obtained should be a uniform image that does not contain any information. However, the difference image obtained due to a failure in alignment clearly shows all the information patterns of the original image. It will be included. Therefore, the key to determining whether the images in the difference image are aligned to completely overlap is to identify whether the difference image is uniform and contains no information from the original image.

In this embodiment, both the template image and the reference image are subjected to band-pass filtering, and then the band-pass filtered image is aligned by applying a conventional algorithm.

Step S400 is provided in which the variance of the difference image is calculated, the variance is compared with a preset value, and a comparison judgment result is obtained and output.

In the prior art, a measure of the amount of image information is generally referred to as information entropy. Test results on thousands of pictures show that it is difficult to judge whether images are aligned using information entropy, and the correlation between the information entropy of the difference image of different group images and whether the images are aligned is not clear. It shows.

Therefore, in this embodiment, the variance of the difference image is calculated, and whether the images are aligned is determined using the variance of the difference image of the overlapping portion of the subtracted image. Here, the difference image is obtained by subtracting the band-pass filtered image, that is, the template image, from the reference image. After band-pass filtering, the luminance of the image is uniform and random noise and defects are suppressed, so the grayscale value of the obtained difference image is is close to the Normal Distribution and the variance is relatively small.

The variance of the grayscale values of the filtered difference image is calculated to determine whether the image alignment result is accurate. If the images are not aligned, the difference image definitely contains the pattern information of the original image, and the grayscale value Dispersion becomes relatively large. If the variance is less than the preset threshold, the image alignment result is accurate. Otherwise, the image alignment has failed, so alignment can be performed again by applying another alignment algorithm or replacing the template area of the template image.

The preset value can be confirmed by selecting an appropriate variance as the threshold through sample inspection, and the user can also confirm it according to test items / environment / other factors, etc. After filtering and sorting 3,000 groups of images, the variance of the grayscale values of the difference images was checked, and the variance of a few unsorted images was found to be significantly larger than the variance of the aligned images. Therefore, it is relatively easy to judge whether the image is aligned by selecting the distribution threshold as a sample, and the results can be judged without being sensitive to the threshold within a certain range.

As such, the present invention includes the steps of obtaining an original image and performing filtering processing on the original image to obtain a filtered image; Obtaining the filtered image and sorting the filtered image using a sorting algorithm to obtain a template image; Obtaining a difference image by performing an image subtraction operation on the template image and a preset reference image; The success rate of image alignment in the defect detection process can be improved through calculating the variance of the difference image, comparing and judging the variance with a preset value, and obtaining and outputting a comparison decision result. In particular, when a defect occurs in a large area during the image comparison process, the image quality deteriorates due to noise, or the overall contrast is uneven, the present invention aligns the filtered images using the Fourier transform to obtain the difference value of the two images. It determines the success of the alignment based on the alignment, has the ability to accurately judge whether the image is aligned, removes unnecessary details during image alignment through filtering, and accurately judges the alignment result using the difference image, enabling various alignment algorithms. It can be used interchangeably, and by implementing each advantage, the success rate of the image alignment algorithm of the present invention is significantly improved, thereby greatly improving the accuracy of the alignment algorithm.

As shown in FIG. 3, in one possible implementation form, optionally, the step of obtaining a filtered image by filtering the original image in step S100 includes the following steps:

Step S110, obtaining a spectral image by Fourier transforming the original image into spectral space;

Step S120, inputting the spectral image to a Gaussian band-pass filter and performing filtering on the image spectrum to obtain a filtered spectral image; Here, the specific form of the Gaussian band-pass filter is as follows:

here:

D(u,v) is the distance to the origin in spectral space,

D ₀ is the center frequency of the passband, W is the width of the passband,

(u,v) is the pixel coordinate of the spectral image after Fourier transformation, the center point pixel of the spectral image is the origin of the coordinate;

Step S120 of receiving the filtered spectrum image and performing inverse Fourier transform on the filtered spectrum image to obtain the filtered image.

As shown in Figure 3, the original image in Figure 2 was obtained by inverse Fourier transforming the previously band-pass filtered filtered spectrum image, and the image is the band-pass filtered image (filtered image).

Through comparison, it can be seen that compared to the original image, the image only has rough outline information of the original image, and the detailed information of the original image has already been lost. The reason for using band-pass filtering to process images is because when aligning two images, only the rough outlines of the images need to be aligned with each other. Excessive details in common sorting algorithms often have a confusing effect on conventional sorting algorithms, leading to incorrect results. Conventional algorithms focus on the detail noise of the image and attempt to align the image details and even the noise, thus ignoring the alignment of the rough outlines of the image.

In one possible implementation form, optionally, the step of obtaining a difference image by subtracting the template image from the reference image in step S300 includes the following steps:

Obtaining the sorted result processed in step S200;

Receiving the alignment result and moving the template image according to the alignment result to obtain a template translation image; A template image is obtained by sorting by applying a sorting algorithm that uses filtered images, and the template image is moved according to the sorting result to obtain a template translation image.

This is a step of obtaining a difference image by subtracting the template translation image from the reference image. Here, the difference image is a difference image obtained by subtracting the filtered image.

After performing band-pass filtering on both the template image and the reference image, alignment is performed by applying a conventional algorithm to the band-pass filtered image.

As shown in Figure 4, the alignment algorithm calculates a translation vector, and if the alignment algorithm result is accurate, the template image is moved according to the vector to completely overlap the template image with the reference image.

Subtraction is to translate the template image and then subtract the pixel luminance values of the template image and the reference image to obtain a single difference image. In an ideal case, at this time, the template image and the reference image completely overlap, the luminance value of each pixel point of the image obtained after subtraction is completely the same, and as shown in FIG. 4, it is one image uniformly without any pattern.

In reality, due to differences in noise and imaging conditions, as shown in FIG. 5, when the template image and the reference image are aligned and then subtracted, they are generally not constant but fluctuate on a uniform background (when there is no signal) (Similar to TV's snow screen).

In one possible implementation form of this embodiment, optionally, calculating the variance of the difference image in step S4, comparing and judging the variance with a preset value, and obtaining and outputting a comparison decision result are the following steps. Includes:

calculating variance by obtaining grayscale values of the difference image;

If the variance is less than the preset value, image alignment is successful, and if the variance exceeds the preset value, image alignment is failed, so re-sort according to the preset alignment processing method. A step of comparing and determining values;

A step of obtaining and outputting comparison judgment results.

After obtaining the difference image, grayscale value data of the difference image is acquired. The acquisition method can be directly checked in the image software, and the specific method is not limited.

After acquisition, the variance and a preset value are compared and judged to obtain and output a comparison judgment result. Specifically, by calculating the variance of the grayscale values of the difference image, if the variance is less than a preset threshold, the image alignment result is accurate. Otherwise, the image alignment has failed, so another alignment algorithm can be applied or a template of the template image. Re-sorting can be performed by replacing the area.

The grayscale value of each pixel point in the image is added and divided by the total number of pixels (i.e., the length of the image multiplied by the width) to obtain the average pixel value. Then, the average value is subtracted from each pixel point, squared, added again to obtain the sum, and then divided by the number of pixel points - 1 (N-1). In other words, it is a standard method of calculating variance in statistics.

Here, the statistical sample is the set of grayscale values of all pixel points in the image. As shown in FIG. 6, FIG. 6 is a schematic diagram showing the distribution of pixel difference values of two aligned images, and FIG. 7 is a schematic diagram showing the distribution of pixel difference values of two unaligned images. It can be clearly seen that the distribution of pixel difference values of the two aligned images is smaller and the alignment effect is better.

Note that although image filtering is processed using a Gaussian band-pass filter and the variance calculation is introduced using grayscale values as an example, those skilled in the art will understand that the present disclosure is not limited thereto. In fact, users can flexibly set the image processing and grayscale value calculation methods according to personal preferences and/or actual application scenarios, as long as they can filter and obtain grayscale value data.

Example 2

Based on Example 1, according to another aspect of the present disclosure, a semiconductor electron beam defect monitoring comprising an original image filtering processing module, an alignment processing module, a differential image acquisition module, and a distributed calculation and judgment module electrically connected in that order. An image processing device is provided for, wherein:

The original image filtering processing module acquires an original image, performs filtering processing on the original image, and obtains a filtered image;

The alignment processing module acquires the filtered image, sorts the filtered image using a sorting algorithm, and obtains a template image;

The differential image acquisition module acquires a differential image by performing an image subtraction operation on the template image and a preset reference image;

The variance calculation and judgment module calculates the variance of the difference image, compares and judges the variance with a preset value, and obtains and outputs a comparison judgment result.

For the specific operating principles of the original image filtering processing module, alignment processing module, differential image acquisition module, and distributed calculation and judgment module, refer to the technical principles of Embodiment 1 and will not be described again here. The following are alumni.

Each module or each step of the present invention described above may be implemented in a general-purpose computing device, and they may be concentrated in a single computing device or distributed in a network composed of a plurality of computing devices, and optionally, they may be programs that can be executed by the computing device. Since they can be implemented as codes, they can be stored in a storage device and executed by a computing device, each of them can be manufactured as an individual integrated circuit module, or a plurality of modules or steps among them can be manufactured into a single integrated circuit module. As such, the present invention is not limited to the combination of any specific hardware and software.

In one possible implementation form, optionally, the original image filtering processing module includes:

a filtered spectrum image acquisition module that inputs the original image into a Gaussian band-pass filter and performs filtering on the image spectrum to obtain a filtered spectrum image; Here, the specific form of the Gaussian band-pass filter is as follows:

here:

D(u,v) is the distance to the origin in spectral space,

D ₀ is the center frequency of the passband, W is the width of the passband,

(u,v) is the pixel coordinate of the spectral image after Fourier transformation, the center point pixel of the spectral image is the origin of the coordinate;

A spectral image inverse Fourier transform processing module that receives the filtered spectral image and performs inverse Fourier transform on the filtered spectral image to obtain the filtered image.

In one possible implementation form, optionally, the differential image acquisition module includes:

A sorting result acquisition module, which acquires the sorted sorted result in step S2;

an image translation module that receives the alignment result and moves the template image according to the alignment result to obtain a template translation image;

An image subtraction processing module that obtains a difference image by subtracting the template translation image from the reference image.

In one possible implementation form, optionally, the distributed calculation and judgment module includes:

a variance calculation module that calculates variance by obtaining a grayscale value of the difference image;

If the variance is less than the preset value, image alignment is successful, and if the variance exceeds the preset value, image alignment is failed, so re-sort according to the preset alignment processing method. a distributed comparison judgment module that compares and determines values;

An output module that obtains and outputs comparison judgment results.

Example 3

According to another aspect of the present disclosure, there is further provided an image processing method control system for semiconductor electron beam defect monitoring comprising:

processor;

Memory that stores instructions that a processor can execute;

Here, the processor is configured to implement the image processing method for semiconductor electron beam defect monitoring described in Embodiment 1 when executing the executable instructions.

Example 4

Additionally, according to another aspect of the present disclosure, an image processing method control system for semiconductor electron beam defect monitoring is further provided.

A control system according to an embodiment of the present disclosure includes a processor and a memory that stores instructions that the processor can execute. Here, the processor is configured to implement the image processing method for semiconductor electron beam defect monitoring described in Embodiment 1 above when executing the executable instructions.

Here, it should be noted that the number of processors may be one or more. At the same time, the control system of the embodiment of the present disclosure may further include an input device and an output device. Here, the processor, memory, input device, and output device may be connected through a bus or other methods, but are not specifically limited here.

A computer-readable storage medium for an image processing method for semiconductor electron beam defect monitoring, wherein the memory includes a software program, such as a program or module, computer-executable, corresponding to the image processing method for semiconductor electron beam defect monitoring according to an embodiment of the present disclosure. It can be used to store programs and various modules. The processor executes various functional applications and data processing of the control system by executing software programs or modules stored in memory.

An input device can be used to receive entered numbers or signals. Here, the signal may generate a key signal related to user settings and function control of the device/terminal/server. The output device may include a display device such as a display.

Although each embodiment of the present disclosure has been described above, the description is illustrative, not comprehensive, and is not limited to each disclosed embodiment. Various modifications and changes without departing from the scope and spirit of each described embodiment will be apparent to those skilled in the art. The terminology used herein has been selected to best explain the principles, practical applications, or technical improvements in the market technology of the embodiments, or to enable those skilled in the art to understand each embodiment disclosed herein.

Claims (10)

In an image processing method for semiconductor electron beam defect monitoring,
Step S100 of acquiring an original image and performing filtering processing on the original image to obtain a filtered image;
Step S200, obtaining the filtered image and sorting the filtered image using a sorting algorithm to obtain a template image;
Step S300 of obtaining a difference image by performing an image subtraction operation on the template image and a preset reference image; and
Step S400 of calculating the variance of the difference image, comparing and judging the variance with a preset value, and obtaining and outputting a comparison decision result;
An image processing method for monitoring semiconductor electron beam defects, including a. According to paragraph 1,
The step of obtaining a filtered image by filtering the original image in step S100 is,
Step S110, obtaining a spectral image by Fourier transforming the original image into spectral space;
Step S120, inputting the spectral image to a Gaussian band-pass filter and performing filtering on the image spectrum to obtain a filtered spectral image; and
Step S130 of receiving the filtered spectrum image and performing inverse Fourier transform on the filtered spectrum image to obtain the filtered image;
Including,
The specific form of the Gaussian band-pass filter is as follows:

here:
D(u,v) is the distance to the origin in spectral space,
D ₀ is the center frequency of the pass band, W is the width of the pass band,
(u,v) are the pixel coordinates of the spectral image after Fourier transformation, and the center point pixel of the spectral image is the origin of the coordinates. An image processing method for monitoring semiconductor electron beam defects. According to claim 1 or 2,
Obtaining a difference image by subtracting the template image from the reference image in step S300 includes:
Obtaining the sorted result processed in step S200;
Receiving the alignment result and moving the template image according to the alignment result to obtain a template translation image;
Obtaining a difference image by subtracting the template translation image from the reference image;
An image processing method for monitoring semiconductor electron beam defects, including a. According to paragraph 1,
The step of calculating the variance of the difference image in step S400, comparing the variance with a preset value, and making a judgment to obtain and output a comparison decision result,
calculating variance by obtaining grayscale values of the difference image;
If the variance is less than the preset value, image alignment is successful, and if the variance exceeds the preset value, image alignment is failed, so re-sort according to the preset alignment processing method. A step of comparing and determining values;
Obtaining and outputting a comparison judgment result;
An image processing method for monitoring semiconductor electron beam defects, including a. In an image processing device for monitoring semiconductor electron beam defects,
It includes an original image filtering processing module, an alignment processing module, a differential image acquisition module, and a distributed calculation and judgment module;
The original image filtering processing module acquires an original image, performs filtering processing on the original image, and obtains a filtered image;
The alignment processing module acquires the filtered image, sorts the filtered image using a sorting algorithm, and obtains a template image;
The difference image acquisition module acquires a difference image by performing an image subtraction operation on the template image and a preset reference image;
The variance calculation and determination module calculates the variance of the difference image, compares the variance with a preset value, and determines to obtain and output a comparison judgment result. An image processing device for monitoring semiconductor electron beam defects. According to clause 5,
The original image filtering processing module,
a filtered spectrum image acquisition module that inputs the original image into a Gaussian band-pass filter and performs filtering on the image spectrum to obtain a filtered spectrum image; and
a spectral image inverse Fourier transform processing module that receives the filtered spectral image and performs inverse Fourier transform on the filtered spectral image to obtain the filtered image;
Including,
The specific form of the Gaussian band-pass filter is as follows:

here:
D(u,v) is the distance to the origin in spectral space,
D ₀ is the center frequency of the passband, W is the width of the passband,
(u, v) are the pixel coordinates of the spectral image after Fourier transformation, and the center point pixel of the spectral image is the origin of the coordinates. An image processing device for monitoring electron beam defects in a semiconductor. According to claim 5 or 6,
The differential image acquisition module,
A sorting result acquisition module, which acquires the sorted sorted result in step S2;
an image translation module that receives the alignment result and moves the template image according to the alignment result to obtain a template translation image;
An image processing module for obtaining a difference image by subtracting the template translation image from the reference image. According to clause 6,
The distributed calculation and judgment module is,
a variance calculation module that calculates variance by obtaining a grayscale value of the difference image;
If the variance is less than the preset value, image alignment is successful, and if the variance exceeds the preset value, image alignment is failed, so re-sort according to the preset alignment processing method. a distributed comparison judgment module that compares and determines values;
An image processing device for monitoring semiconductor electron beam defects, comprising an output module that obtains and outputs a comparison decision result. In the control system,
processor; and
Memory that stores instructions that a processor can execute;
Including,
The control system, wherein the processor, when executing the executable instructions, is configured to implement the image processing method for semiconductor electron beam defect monitoring according to any one of claims 1 to 4. A non-volatile computer-readable storage medium storing computer program instructions, wherein the method according to any one of claims 1 to 4 is implemented when the computer program instructions are executed by a processor.