Method for improving CDSEM image quality

Abstract

The invention provides a method for improving CDSEM image quality. The method comprises the steps that CDSEM images in different formats are read, and mutual conversion between formats is carried out on the CDSEM images in different formats; selecting a wavelet basis function; setting a wavelet decomposition layer number N; setting threshold values respectively corresponding to a threshold function and the wavelet decomposition layer number N; performing wavelet decomposition to obtain scale function coefficients of different layers and wavelet coefficients of different layers; the threshold processing comprises the step of performing threshold processing on the wavelet coefficients of different layers to obtain the processed wavelet coefficients of each layer; the multi-dimensional wavelet reconstruction comprises the following steps of: performing multi-dimensional wavelet reconstruction by using a scale function coefficient and the processed wavelet coefficient of each layer to obtain a denoised CDSEM image; and carrying out image quality evaluation on the denoised CDSEM image.

Description

Technical Field

[0001] The invention relates to the field of semiconductor technology, and in particular to a method for improving CDSEM image quality. Background Art

[0002] As semiconductor process nodes continue to shrink, especially as 40nm and below process nodes gradually become mainstream products, OPC (optical proximity correction) has become an indispensable part of the semiconductor process. Establishing an accurate OPC model is a prerequisite for OPC correction, and the establishment of the OPC model is inseparable from high-quality CDSEM image measurement technology. Not only OPC modeling, but also the optimization analysis of OPC programs and the analysis of lithography exposure conditions are inseparable from high-quality CDSEM images. The quality of CDSEM images is easily affected by the interference of the machine itself, such as lithography parameter fluctuations, photoresist quality, development and drying, and line width measurement, which can easily cause the CDSEM image clarity to be insufficient, affecting the subsequent CDSEM measurement accuracy. Summary of the invention

[0003] In view of the shortcomings of the prior art described above, an object of the present invention is to provide a method for improving the quality of CDSEM images, which is used to solve the problem in the prior art that the quality of CDSEM images is easily affected by the interference of the machine itself such as lithography parameter fluctuations, photoresist quality, development and drying, and line width measurement, which easily causes the CDSEM image clarity to be insufficient, affecting the subsequent CDSEM measurement accuracy.

[0004] To achieve the above object and other related objects, the present invention provides a method for improving the quality of CDSEM images, which at least comprises:

[0005] Step 1, reading CDSEM images, including: reading CDSEM images of different formats and converting CDSEM images of different formats into each other;

[0006] Step 2, setting wavelet parameters, including: selecting wavelet basis functions; setting the number of wavelet decomposition layers N; setting threshold functions and threshold values ​​corresponding to the number of wavelet decomposition layers N;

[0007] Step 3: wavelet decomposition, threshold processing and multi-dimensional wavelet reconstruction;

[0008] Wavelet decomposition to obtain scaling function coefficients and wavelet coefficients at different layers;

[0009] The threshold processing includes performing threshold processing on the wavelet coefficients of the different layers to obtain processed wavelet coefficients of each layer;

[0010] The multidimensional wavelet reconstruction comprises: performing multidimensional wavelet reconstruction using the scaling function coefficients and the processed wavelet coefficients of each layer to obtain a denoised CDSEM image;

[0011] Step 4: Evaluate the image quality of the denoised CDSEM image.

[0012] Preferably, the CDSEM images in different formats in step 1 include CDSEM images in jpg, tiff, png, and bmp formats.

[0013] Preferably, the conversion between formats of CDSEM images of different formats in step 1 includes converting color images into grayscale images and converting between different grayscale formats.

[0014] Preferably, the wavelet basis function in step 2 includes Haar wavelet, db4 wavelet, coif3 wavelet.

[0015] Preferably, the number of wavelet decomposition layers N≦3 in step 2.

[0016] Preferably, the threshold function in step 2 is a soft threshold function or a hard threshold function.

[0017] Preferably, the method for obtaining the scaling function coefficients and wavelet coefficients of different layers in step three includes: performing N-layer wavelet decomposition on the CDSEM image read in step one according to the selected wavelet basis function, the number of wavelet decomposition layers N, and the threshold function and threshold value, to obtain the scaling function coefficients and wavelet coefficients of different layers.

[0018] Preferably, the method of performing threshold processing in step three includes filtering out high-frequency noise in the CDSEM image according to a set threshold function and threshold value.

[0019] Preferably, in step 4, two evaluation criteria, signal-to-noise ratio (SNR) and peak signal-to-noise ratio (PSNR), are used to evaluate the image quality of the denoised CDSEM image.

[0020] As described above, the method for improving the quality of CDSEM images of the present invention has the following beneficial effects: the present invention uses the wavelet analysis principle to filter the CDSEM image, filter out the high-frequency low-energy noise useless information in the image, retain the low-frequency high-energy useful information, greatly improve the image quality of the CDSEM after filtering and denoising, and improve the accuracy of wafer measurement. The method does not require additional equipment and does not require re-exposure of the wafer to obtain the denoised CDSEM image, which can greatly save the time and cost of subsequent image processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Shown is the original CDSEM image of a normal contact hole of the present invention;

[0022] Figure 2a Show as pair Figure 1 CDSEM image of contact hole after denoising using Haar wavelet threshold filtering;

[0023] Figure 2b Display as Figure 1 CDSEM image of contact hole using the scale coefficient of the first wavelet decomposition of Haar wavelet;

[0024] Figure 2c show Figure 1 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of Haar wavelet;

[0025] Figure 3a Show as pair Figure 1 CDSEM image of contact hole after denoising using threshold filtering of db4 wavelet;

[0026] Figure 3b Display as Figure 1 CDSEM image of contact hole using the scaling coefficients of the first wavelet decomposition of db4 wavelet;

[0027] Figure 3c show Figure 1 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of db4 wavelet;

[0028] Figure 4a Show as pair Figure 1 CDSEM image of contact hole after denoising using threshold filtering of coif3 wavelet;

[0029] Figure 4b Display as Figure 1 CDSEM image of contact hole with scale coefficients of the first wavelet decomposition using coif3 wavelet;

[0030] Figure 4c show Figure 1 CDSEM image of contact hole using the scaling coefficients of the second wavelet decomposition of coif3 wavelet;

[0031] Figure 5 The display shows the comparison of NSR and PSNR of the contact hole CDSEM image obtained by three-layer wavelet decomposition and reconstruction of the original CDSEM image of the normal contact hole using Haar wavelet, db4 wavelet and coif3 wavelet, the contact hole CDSEM image of the scale coefficient of the first layer wavelet decomposition and the contact hole CDSEM image of the scale coefficient of the second layer wavelet decomposition according to the present invention;

[0032] Figure 6Shown is the original CDSEM image of the abnormal contact hole of the present invention;

[0033] Figure 7a Show as pair Figure 6 CDSEM image of contact hole after denoising using Haar wavelet threshold filtering;

[0034] Figure 7b Display as Figure 6 CDSEM image of contact hole using the scale coefficient of the first wavelet decomposition of Haar wavelet;

[0035] Figure 7c show Figure 6 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of Haar wavelet;

[0036] Figure 8a Show as pair Figure 6 CDSEM image of contact hole after denoising using threshold filtering of db4 wavelet;

[0037] Figure 8b Display as Figure 6 CDSEM image of contact hole using the scaling coefficients of the first wavelet decomposition of db4 wavelet;

[0038] Figure 8c show Figure 6 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of db4 wavelet;

[0039] Figure 9a Show as pair Figure 6 CDSEM image of contact hole after denoising using threshold filtering of coif3 wavelet;

[0040] Figure 9b Display as Figure 6 CDSEM image of contact hole with scale coefficients of the first wavelet decomposition using coif3 wavelet;

[0041] Fig.9c show Figure 6 CDSEM image of contact hole using the scaling coefficients of the second wavelet decomposition of coif3 wavelet;

[0042] Fig.10 The display shows the comparison of NSR and PSNR of the contact hole CDSEM image obtained by three-layer wavelet decomposition and reconstruction of the original CDSEM image of the abnormal contact hole using Haar wavelet, db4 wavelet and coif3 wavelet, the contact hole CDSEM image of the scale coefficient of the first layer wavelet decomposition and the contact hole CDSEM image of the scale coefficient of the second layer wavelet decomposition according to the present invention;

[0043] Fig.11 Displayed are 1026 CDSEM images - comparison of Haar, db4 and coif3 wavelet filtering effects;

[0044] Fig.12 Shown is a flow chart of the method for improving CDSEM image quality according to the present invention. DETAILED DESCRIPTION

[0045] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0046] See also Figures 1 to 12 It should be noted that the illustrations provided in this embodiment are only used to illustrate the basic concept of the present invention in a schematic manner, and the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0047] The present invention provides a method for improving the quality of CDSEM images. The method uses the wavelet analysis principle to filter the CDSEM images, filter out the high-frequency low-energy noise useless information in the images, and retain the low-frequency high-energy useful information. The image quality of the CDSEM after denoising can be greatly improved, and the accuracy of subsequent wafer measurement can be improved. Wavelet theory has attracted more and more attention in the field of image denoising due to its good time-frequency characteristics, and has opened up a precedent for denoising using nonlinear methods. Specifically, the denoising of wavelets is mainly due to the following characteristics of wavelet transform: (1) low entropy. The sparse distribution of wavelet coefficients reduces the entropy of the image after transformation. This means that after the signal (i.e., the image) is decomposed, more wavelet basis coefficients tend to 0 (noise), while the main part of the signal is mostly concentrated in certain wavelet bases. Threshold denoising can better retain the original signal. (2) Multi-resolution characteristics. Due to the use of multi-resolution methods, the non-stationarity of the signal, such as mutations and breakpoints, can be well characterized (for example, 0-1 mutations cannot be reasonably represented by Fourier transforms), and noise can be eliminated according to the distribution of signals and noise at different resolutions. (3) Decorrelation. Wavelet transform can decorrelate the signal, and the noise tends to whiten after the transform, so the wavelet domain is more conducive to denoising than the time domain. (4) Flexible selection of basis function. Wavelet transform can flexibly select basis functions, and can also select multi-band wavelets and wavelet packets according to signal characteristics and denoising requirements (wavelet packets decompose high-frequency signals again to improve time-frequency resolution). Different wavelet basis functions can be selected for different occasions. According to the different processing methods based on wavelet coefficients, common denoising methods can be divided into three categories: (1) Denoising based on wavelet transform modulus maximum (signal and noise modulus maximum will show different change trends under wavelet transform) (2) Denoising based on correlation of adjacent scale wavelet coefficients (noise has no obvious correlation between the scales of wavelet transform, while the opposite is true for signal) (3) Denoising based on wavelet transform threshold.

[0048] like Fig.12 As shown, Fig.12 The flowchart of the method for improving the quality of CDSEM images of the present invention is shown, and the method comprises at least the following steps:

[0049] Step 1, reading CDSEM images, including: reading CDSEM images of different formats and converting CDSEM images of different formats into each other;

[0050] Furthermore, in the present invention, the CDSEM images in different formats in step 1 of this embodiment include CDSEM images in jpg, tiff, png, and bmp formats.

[0051] Furthermore, in the present invention, the conversion between formats of CDSEM images of different formats in step 1 of this embodiment includes conversion of color images into grayscale images, conversion between different grayscale formats, such as conversion of unsigned integers into floating-point numbers, etc.

[0052] Step 2, setting wavelet parameters, including: selecting wavelet basis functions; setting the number of wavelet decomposition layers N; setting threshold functions and threshold values ​​corresponding to the number of wavelet decomposition layers N;

[0053] Furthermore, in the present invention, the wavelet basis function in step 2 of this embodiment includes Haar wavelet, db4 wavelet, and coif3 wavelet.

[0054] Furthermore, in the present invention, the number of wavelet decomposition layers N≦3 in step 2 of this embodiment.

[0055] Furthermore, in the present invention, the threshold function in step 2 of this embodiment is a soft threshold function or a hard threshold function.

[0056] Step 3: wavelet decomposition, threshold processing and multi-dimensional wavelet reconstruction;

[0057] Wavelet decomposition to obtain scaling function coefficients and wavelet coefficients at different layers;

[0058] The present invention further provides that the method for obtaining the scaling function coefficients and wavelet coefficients of different layers in step three of this embodiment includes: performing N-layer wavelet decomposition on the CDSEM image read in step one according to the selected wavelet basis function, the number of wavelet decomposition layers N, and the threshold function and threshold value to obtain the scaling function coefficients and wavelet coefficients of different layers.

[0059] The threshold processing includes performing threshold processing on the wavelet coefficients of the different layers to obtain processed wavelet coefficients of each layer;

[0060] Furthermore, in the present invention, the method for performing threshold processing in step three of this embodiment includes filtering out high-frequency noise in the CDSEM image according to a set threshold function and threshold value.

[0061] The multidimensional wavelet reconstruction comprises: performing multidimensional wavelet reconstruction using the scaling function coefficients and the processed wavelet coefficients of each layer to obtain a denoised CDSEM image;

[0062] Step 4: Evaluate the image quality of the denoised CDSEM image.

[0063] Furthermore, in step 4 of this embodiment, two evaluation criteria, namely, signal-to-noise ratio (SNR) and peak signal-to-noise ratio (PSNR), are used to evaluate the image quality of the denoised CDSEM image.

[0064] The signal to noise ratio (SNR) Where P s and P n are the power of signal and noise respectively.

[0065] Peak signal to noise ratio (PSNR) Where R is the maximum fluctuation value of the input image type. If the image type is a double-precision floating point number, R = 1; if the image type is an 8-bit unsigned number, R = 255; To calculate PSNR, you must first calculate MSE (Mean Square Error). Where M and N are the number of rows and columns of the input image respectively.

[0066] Different image quality evaluation standards have different focuses. For example, MSE is based on the mean square error of the image, SNR is based on the logarithm of the ratio of signal power to noise power, and PSNR is mainly used to evaluate the peak deviation. The present invention uses SNR and PSNR as the standards for evaluating image quality.

[0067] Example

[0068] Taking the CDSEM image of a contact hole of a certain product as an example, the process and method of filtering and denoising the CDSEM image using wavelet analysis of the present invention are explained. This embodiment is divided into two cases, one is for normal CT CDSEM images, and the other is for abnormal CT CDSEM images containing more noise.

[0069] (I) CDSEM original image of normal contact hole

[0070] Step 1: Input parameters of CDSEM original image of contact hole;

[0071] The input parameters of the CDSEM original image of the contact hole are 512*512 8-bit unsigned data, the image is in jpg format, and the image size is about 65k.

[0072] Step 2: Setting of wavelet parameters;

[0073] The wavelet basis functions are Haar wavelet, db4 wavelet and coif3 wavelet, the wavelet decomposition layer number N=3, the threshold function is the hard threshold function, and the three-layer wavelet thresholds are set to 100, 150 and 200 respectively.

[0074] Step 3: wavelet decomposition, threshold processing and multi-dimensional wavelet reconstruction;

[0075] After three-layer wavelet decomposition, threshold filtering and three-layer wavelet reconstruction, three-layer scale coefficients and the reconstructed contact hole CDSEM image were obtained respectively. Figure 1 Shown is the original CDSEM image of a normal contact hole of the present invention; Figure 2a Show as pair Figure 1 CDSEM image of contact hole after denoising using Haar wavelet threshold filtering; Figure 2b Display as Figure 1 CDSEM image of contact hole using the scale coefficient of the first wavelet decomposition of Haar wavelet; Figure 2c show Figure 1 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of Haar wavelet; Figure 3a Show as pair Figure 1 CDSEM image of contact hole after denoising using threshold filtering of db4 wavelet; Figure 3b Display as Figure 1 CDSEM image of contact hole using the scaling coefficients of the first wavelet decomposition of db4 wavelet; Figure 3c show Figure 1 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of db4 wavelet; Figure 4a Show as pair Figure 1 CDSEM image of contact hole after denoising using threshold filtering of coif3 wavelet; Figure 4b Display as Figure 1 CDSEM image of contact hole with scale coefficients of the first wavelet decomposition using coif3 wavelet; Figure 4c show Figure 1 CDSEM image of the contact hole using the scaling coefficients of the second-level wavelet decomposition of the coif3 wavelet.

[0076] Step 4: Evaluate the image quality of the denoised CDSEM image;

[0077] Figure 5 The figure shows the comparison of NSR and PSNR of the contact hole CDSEM image (Threshold) obtained by three-layer wavelet decomposition and reconstruction of the original CDSEM image of the normal contact hole using Haar wavelet, db4 wavelet and coif3 wavelet, the scale coefficient contact hole CDSEM image of the first layer wavelet decomposition (App lev1) and the scale coefficient contact hole CDSEM image of the second layer wavelet decomposition (App lev2). Figure 2a-4c and Figure 5It can be found that the comparison of Haar, db4 and coif3 wavelet filtering effects on normal contact hole CDSEM images has the following conclusions: the three wavelet filtering effects are significantly improved compared with the original image, and the image noise is greatly suppressed; the filtering effects of the three wavelets are similar, and the comprehensive filtering effect of coif3 wavelet is the best; after the three wavelets are decomposed, the first layer scale coefficient method is taken, and its corresponding PSNR and SNR are greater than the threshold filtering method and the second layer scale coefficient method, and the filtering effect is the best; most of the high-frequency noise in the original image is concentrated in the first layer of wavelet decomposition coefficients, so only taking the first layer of scale coefficients can obtain a satisfactory filtering effect; since most of the noise is concentrated in the first and second layers of wavelet coefficients, it is not recommended to exceed 3 layers of wavelet decomposition, and the image quality may deteriorate.

[0078] (II) CDSEM original image of abnormal contact hole

[0079] The difference between the abnormal contact hole CDSEM original image and the normal contact hole CDSEM original image is that the abnormal contact hole CDSEM original image contains more noise caused by the machine, and other wavelet filtering conditions are exactly the same as those of the normal contact hole CDSEM original image.

[0080] Figure 6 Shown is the original CDSEM image of the abnormal contact hole of the present invention; Figure 7a Show as pair Figure 6 CDSEM image of contact hole after denoising using Haar wavelet threshold filtering; Figure 7b Display as Figure 6 CDSEM image of contact hole using the scale coefficient of the first wavelet decomposition of Haar wavelet; Figure 7c show Figure 6 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of Haar wavelet;

[0081] Figure 8a Show as pair Figure 6 CDSEM image of contact hole after denoising using threshold filtering of db4 wavelet; Figure 8b Display as Figure 6 CDSEM image of contact hole using the scaling coefficients of the first wavelet decomposition of db4 wavelet; Figure 8c show Figure 6 CDSEM image of contact hole using the scaling coefficient of the second wavelet decomposition of db4 wavelet;

[0082] Figure 9a Show as pair Figure 6 CDSEM image of contact hole after denoising using threshold filtering of coif3 wavelet; Figure 9b Display as Figure 6CDSEM image of contact hole with scale coefficients of the first wavelet decomposition using coif3 wavelet; Fig.9c show Figure 6 CDSEM image of the contact hole using the scaling coefficients of the second-level wavelet decomposition of the coif3 wavelet.

[0083] Fig.10 It shows the NSR and PSNR comparison of the contact hole CDSEM image (Threshold) obtained by three-layer wavelet decomposition and reconstruction of the original CDSEM image of the abnormal contact hole using Haar wavelet, db4 wavelet and coif3 wavelet respectively, the scale coefficient contact hole CDSEM image of the first layer wavelet decomposition (App lev1) and the scale coefficient contact hole CDSEM image of the second layer wavelet decomposition (App lev2).

[0084] from Figure 7a-9c and Fig.10 It can be found that the following conclusions are drawn from the comparison of the Haar, db4 and coif3 wavelet filtering effects on the CDSEM images of abnormal contact holes: the three wavelet filtering effects are significantly improved compared with the original image, and the noise is greatly suppressed; the filtering effects of the three wavelets are similar, and the coif3 comprehensive filtering effect is the best; the threshold filtering method of the three wavelets is better than the first layer proximity coefficient method and the second layer proximity coefficient method, and the corresponding PSNR and SNR are larger, and the filtering effect is the best. Most of the high-frequency noise in the original image is significantly reduced after the first and second layers of wavelet decomposition coefficients, so the threshold filtering method, the first layer proximity coefficient method and the second layer proximity coefficient method are flexibly selected according to the actual situation; since most of the noise is concentrated in the first and second layers of wavelet coefficients, it is not recommended to exceed 3 layers of wavelet decomposition, and the image quality may deteriorate.

[0085] (III) 1026 contact hole SEM images

[0086] like Fig.11 As shown, Fig.11 Displayed are 1026 CDSEM images - comparison of Haar, db4 and coif3 wavelet filtering effects. Figure 5 , Fig.10 and Fig.11 The comparison found that Fig.11 The conclusions and Figure 5 , Fig.10 Consistent.

[0087] In summary, the present invention uses the wavelet analysis principle to filter the CDSEM image, filter out the high-frequency low-energy noise useless information in the image, retain the low-frequency high-energy useful information, greatly improve the image quality of the CDSEM after filtering and denoising, and improve the accuracy of wafer measurement. This method does not require additional equipment and does not require re-exposure of the wafer to obtain the denoised CDSEM image, which can greatly save the time and cost of subsequent image processing. Therefore, the present invention effectively overcomes the various shortcomings of the prior art and has a high industrial utilization value.

[0088] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for improving CDSEM image quality, characterized in that: At least: Step 1, reading CDSEM images, including: reading CDSEM images of different formats and converting CDSEM images of different formats into each other; Step 2, setting wavelet parameters, including: selecting wavelet basis functions; setting the number of wavelet decomposition layers N; setting threshold functions and threshold values ​​corresponding to the number of wavelet decomposition layers N; Step 3: wavelet decomposition, threshold processing and multi-dimensional wavelet reconstruction; Wavelet decomposition to obtain scaling function coefficients and wavelet coefficients at different layers; The threshold processing includes performing threshold processing on the wavelet coefficients of the different layers to obtain processed wavelet coefficients of each layer; The multidimensional wavelet reconstruction comprises: performing multidimensional wavelet reconstruction using the scaling function coefficients and the processed wavelet coefficients of each layer to obtain a denoised CDSEM image; Step 4: Evaluate the image quality of the denoised CDSEM image.

2. The method for improving CDSEM image quality according to claim 1, characterized in that: The CDSEM images in different formats described in step 1 include CDSEM images in jpg, tiff, png, and bmp formats.

3. The method for improving CDSEM image quality according to claim 1, characterized in that: In step 1, the CDSEM images of different formats are converted into each other, including the conversion of color images into grayscale images and the conversion between different grayscale formats.

4. The method for improving CDSEM image quality according to claim 1, characterized in that: The wavelet basis functions in step 2 include Haar wavelet, db4 wavelet, and coif3 wavelet.

5. The method for improving CDSEM image quality according to claim 1, characterized in that: The number of wavelet decomposition layers N≦3 in step 2.

6. The method for improving CDSEM image quality according to claim 1, characterized in that: The threshold function package in step 2 is a soft threshold function or a hard threshold function.

7. The method for improving CDSEM image quality according to claim 1, characterized in that: The method for obtaining the scaling function coefficients and wavelet coefficients of different layers in step three includes: performing N-layer wavelet decomposition on the CDSEM image read in step one according to the selected wavelet basis function, the number of wavelet decomposition layers N, the threshold function and the threshold value, to obtain the scaling function coefficients and wavelet coefficients of different layers.

8. The method for improving CDSEM image quality according to claim 1, characterized in that: The method of performing threshold processing in step three includes filtering out high-frequency noise in the CDSEM image according to a set threshold function and threshold value.

9. The method for improving CDSEM image quality according to claim 1, characterized in that: In step 4, the image quality of the denoised CDSEM image is evaluated using two evaluation criteria: signal-to-noise ratio (SNR) and peak signal-to-noise ratio (PSNR).

Images