# US 2026/0073485 A1 · Scanning electron microscope (sem) image improvement method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0073485 A1
- Document ID: US20260073485A1
- Status: Published application
- Kind: A1
- Application number: 19/056,648
- Filed: 2025-02-18
- Published: 2026-03-12
- Earliest claimed: 2024-09-10
- Assignees: Samsung Electronics Co., Ltd.
- Applicants: Samsung Electronics Co., Ltd.
- Inventors: Dongeun Kim; Jinchoel Choi
- Web page: https://patentgrep.com/patent/US20260073485A1

## CPC

- G06T 5/70: Denoising; Smoothing
- G06T 5/73: Deblurring; Sharpening
- G06T 5/50: using two or more images, e.g. averaging or subtraction
- G06T 5/60: using machine learning, e.g. neural networks
- G06T 2207/10061: from scanning electron microscope
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/20221: Image fusion; Image merging
- G06T 2207/30148: Semiconductor; IC; Wafer
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

Provided is a scanning electron microscope (SEM) image improving method including obtaining a plurality of first SEM images by using the SEM, generating a plurality of second SEM images by removing noise from each of the first SEM images, aligning the plurality of second SEM images, and generating a third SEM image by combining the aligned plurality of second SEM images.

## Claims

### Claim 1

A scanning electron microscope (SEM) image improvement method, the method comprising:
obtaining a plurality of first SEM images by using the SEM;
generating a plurality of second SEM images by removing noise from each of the plurality of first SEM images;
aligning the plurality of second SEM images; and
generating a third SEM image by combining the aligned plurality of second SEM images.

### Claim 2 (depends on 1)

The method of claim 1, wherein each of the plurality of first SEM images is an image obtained by using a semiconductor process resultant product as a target.

### Claim 3 (depends on 2)

The method of claim 2, wherein each of the plurality of first SEM images comprises one frame obtained by one-time irradiation of the SEM.

### Claim 4 (depends on 1)

The method of claim 1, wherein the generating of the plurality of second SEM images by removing the noise comprises generating the plurality of second SEM images by removing the noise from each of the plurality of first SEM images by using machine learning.

### Claim 5 (depends on 4)

The method of claim 4, wherein the machine learning comprises learning by using an auto-encoder/decoder, a convolution autoencoder (CAE), a variational autoencoder (VAE), or a generative adversarial network (GAN).

### Claim 6 (depends on 1)

The method of claim 1, wherein the aligning of the plurality of second SEM images comprises:
setting any one of the plurality of second SEM images as a reference image, setting other second SEM images of the plurality of second SEM images as input images, and aligning the reference image and the input images such that a mean squared error (MSE) between the reference image and each of the input images is minimized while relatively moving the input images with respect to the reference image.

### Claim 7 (depends on 6)

The method of claim 6, wherein the generating of the third SEM image comprises generating the third SEM image by combining the reference image with the input images moved to have a minimum MSE in relation to the reference image.

### Claim 8 (depends on 6)

The method of claim 6, wherein the MSE is calculated using Formula 1 below:
M
⁢
S
⁢
E
⁡
(
x
,
y
)
=
∑
M
·
N
[
Ref
⁢
(
m
,
n
)
-
input
⁢
(
m
-
x
,
n
-
y
)
]
2
M
·
N
,
Formula
⁢
1
wherein Ref(m, n) is a pixel value of the reference image at a coordinate (m, n), input (m−x, n−y) is a pixel value of the input image at a coordinate moved from the coordinate (m, n) by (x, y), M is the number of pixels in a first direction, and N is the number of pixels in a second direction,
wherein the first direction is a longitudinal axis direction of the reference image, and
wherein the second direction is a vertical axis direction of the reference image.

### Claim 9

A scanning electron microscope (SEM) image improvement method, the method comprising:
obtaining a plurality of first SEM images, each including one frame, by using the SEM;
generating a plurality of second SEM images by removing noise from each of the plurality of first SEM images by using machine learning;
aligning the plurality of second SEM images; and
generating a third SEM image by combining the aligned plurality of second SEM images.

### Claim 10 (depends on 9)

The method of claim 9, wherein each of the plurality of first SEM images is an image obtained by using a semiconductor process resultant product as a target.

### Claim 11 (depends on 9)

The method of claim 9, wherein the machine learning comprises learning by using an auto-encoder/decoder, a convolution autoencoder (CAE), a variational autoencoder (VAE), or a generative adversarial network (GAN).

### Claim 12 (depends on 9)

The method of claim 9, wherein the aligning of the plurality of second SEM images comprises:
setting any one of the plurality of second SEM images as a reference image, setting other second SEM images of the plurality of second SEM images as input images, and aligning the reference image and the input images such that a mean squared error (MSE) between the reference image and each of the input images is minimized while relatively moving the input images with respect to the reference image.

### Claim 13 (depends on 12)

The method of claim 12, wherein the generating of the third SEM image comprises generating the third SEM image by combining the reference image with the input images moved to have a minimum MSE in relation to the reference image.

### Claim 14 (depends on 13)

The method of claim 13, wherein the MSE is calculated by using Formula 1 below:
M
⁢
S
⁢
E
⁡
(
x
,
y
)
=
∑
M
·
N
[
Ref
⁢
(
m
,
n
)
-
input
⁢
(
m
-
x
,
n
-
y
)
]
2
M
·
N
,
Formula
⁢
1
wherein Ref(m, n) is a pixel value of the reference image at a coordinate (m, n), input (m−x, n−y) is a pixel value of the input image at a coordinate moved from the coordinate (m, n) by (x, y), M is the number of pixels in the first direction, and N is the number of pixels in the second direction,
wherein the first direction is a longitudinal axis direction of the reference image
wherein the second direction is a vertical axis direction of the reference image.

### Claim 15

A scanning electron microscope (SEM) image improvement method, the method comprising:
obtaining a first SEM image by using a semiconductor process resultant product as a target by using the SEM;
generating a plurality of second SEM images by removing noise from each of the first SEM images by using machine learning;
aligning the plurality of second SEM images; and
generating a third SEM image by combining the aligned plurality of second SEM images.

### Claim 16 (depends on 15)

The method of claim 15, wherein each of the plurality of first SEM images comprises one frame obtained by one-time irradiation of the SEM.

### Claim 17 (depends on 15)

The method of claim 15, wherein the machine learning comprises learning by using an auto-encoder/decoder, a convolution autoencoder (CAE), a variational autoencoder (VAE), or a generative adversarial network (GAN).

### Claim 18 (depends on 15)

The method of claim 15, wherein the aligning of the plurality of second SEM images comprises:
setting any one of the plurality of second SEM images as a reference image;
setting other second SEM images of the plurality of second SEM images as input images;
calculating a movement amount that minimizes a mean squared error (MSE) of the reference image and each of the input images while relatively moving the input images with respect to the reference image; and
aligning the input images moved by the movement amount with the reference image.

### Claim 19 (depends on 15)

The method of claim 15, wherein the generating of the third SEM image comprises generating the third SEM image by combining the reference image with the input images moved to have a minimum MSE in relation to the reference image.

### Claim 20 (depends on 18)

The method of claim 18, wherein the MSE is calculated by using Formula 1 below:
M
⁢
S
⁢
E
⁡
(
x
,
y
)
=
∑
M
·
N
[
Ref
⁢
(
m
,
n
)
-
input
⁢
(
m
-
x
,
n
-
y
)
]
2
M
·
N
,
Formula
⁢
1
wherein Ref(m, n) is a pixel value of the reference image at a coordinate (m, n), input (m−x, n−y) is a pixel value of the input image at a coordinate moved from the coordinate (m, n) by (x, y), M is the number of pixels in the first direction, and N is the number of pixels in the second direction,
wherein the first direction is a longitudinal axis direction of the reference image, and
wherein the second direction is a vertical axis direction of the reference image.
