Super resolution SEM image implementing device and method thereof
Some example embodiments relate to a super resolution scanning electron microscope (SEM) image implementing device and/or a method thereof. Provided a super resolution scanning electron microscope (SEM) image implementing device comprising a processor configured to crop a low resolution SEM image to generate a first cropped image and a second cropped image, to upscale the first cropped image and the second cropped image to generate a first upscaled image and a second upscaled image, and to cancel noise from the first upscaled image and the second upscaled image to generate a first noise canceled image and a second noise canceled image.
1. A super resolution scanning electron microscope (SEM) image implementing device comprising:
a processor configured to execute machine-readable instructions to cause the device to crop a low resolution SEM image to generate a first cropped image and a second cropped image, the low resolution SEM image including an image of at least one of fins of a semiconductor device, shallow trench isolations of the semiconductor device, or gate lines of the semiconductor device,
to upscale the first cropped image and the second cropped image to generate a first upscaled image and a second upscaled image using a ground truth image having high-resolution and corresponding to the first and second cropped image, respectively, and
to cancel noise from the first upscaled image and the second upscaled image to generate a first noise canceled image and a second noise canceled image, wherein
the noise canceling includes canceling noise based on a deep learning network trained on a layout of the semiconductor device, the layout corresponding to a field of view of the SEM image.
2. The super resolution SEM image implementing device of claim 1 , further comprising:
a buffer configured to store a first position of the cropping of the first cropped image and a second position of the cropping of the second cropped image.
3. The super resolution SEM image implementing device of claim 1 , wherein the processor is configured to cause the device to,
based on a boundary line corresponding to the first cropped image and the second cropped image contacting each other,
generate a first-first cropped image further including a first area permeating from the boundary line into the second cropped image by a first length with respect to the first cropped image, and
generate a second-first cropped image further including a second area permeating from the boundary line into the first cropped image by a second length with respect to the second cropped image.
4. The super resolution SEM image implementing device of claim 3 , wherein the processor is configured to cause the device to upscale the first-first cropped image and the second-first cropped image to generate a first-first upscaled image and a second-first upscaled image.
5. The super resolution SEM image implementing device of claim 4 , wherein the processor is configured to cause the device to cancel noise from the first-first upscaled image and the second-first upscaled image to generate a first-first noise canceled image and a second-first noise canceled image.
6. The super resolution SEM image implementing device of claim 1 , wherein the processor is configured to cause the device to
perform convolution on the first cropped image and the second cropped image,
perform deconvolution on the first cropped image on which the convolution is performed and the second cropped image on which the convolution is performed, and
generate the first noise canceled image and the second noise canceled image by applying a selective skip connection to the first cropped image and the second cropped image to cancel the noise.
7. The super resolution SEM image implementing device of claim 1 , wherein the processor is further configured to cause the device to merge the first noise canceled image and the second noise canceled image with each other to generate a super resolution SEM image.
8. A super resolution SEM image implementing method comprising:
generating a first cropped image and a second cropped image by cropping a low resolution SEM image by a processor, the low resolution SEM image including an image of at least one of fins of a semiconductor device, shallow trench isolations of the semiconductor device, or gate lines of the semiconductor device;
generating a first upscaled image and a second upscaled image by upscaling the first cropped image and the second cropped image using a ground truth image having high-resolution and corresponding to the first and second cropped image, respectively; and
generating a first noise canceled image and a second noise canceled image by canceling noise from the first upscaled image and the second upscaled image, wherein
the noise canceling includes canceling noise based on a deep learning network trained on a layout of the semiconductor device, the layout corresponding to a field of view of the SEM image.
9. The super resolution SEM image implementing method of claim 8 , further comprising:
storing, in a buffer, a first position of the cropping of the first cropped image and a second position of the cropping of the second cropped image.
10. The super resolution SEM image implementing method of claim 8 , further comprising:
generating a super resolution SEM image by merging the first noise canceled image and the second noise canceled image with each other.
11. The super resolution SEM image implementing method of claim 8 , wherein the generating of the first noise canceled image and the second noise canceled image includes:
generating the first noise canceled image and the second noise canceled image by performing a selective skip connection on the first cropped image and the second cropped image to cancel the noise, by performing convolution on the first cropped image and the second cropped image, and by performing deconvolution on the first cropped image on which the convolution is performed and the second cropped image on which the convolution is performed.
12. The super resolution SEM image implementing method of claim 8 , further comprising:
based on a boundary line corresponding to the first cropped image and the second cropped image contacting each other,
generating a first-first cropped image further including a first area permeating from the boundary line into the second cropped image by a first length with respect to the first cropped image, and
generating a second-first cropped image further including a second area permeating from the boundary line into the first cropped image by a second length with respect to the second cropped image.
13. The super resolution SEM image implementing method of claim 12 , further comprising:
generating a first-first upscaled image and a second-first upscaled image by upscaling the first-first cropped image and the second-first cropped image.
14. A super resolution SEM image implementing system comprising:
a central processing unit;
a bus connected to the central processing unit; and
a super resolution SEM image implementing device communicating with the central processing unit via the bus,
wherein the super resolution SEM image implementing device is configured to crop a low resolution SEM image to generate a first cropped image and a second cropped image, the low resolution SEM image including an image of at least one of fins of a semiconductor device, shallow trench isolations of the semiconductor device, or gate lines of the semiconductor device,
to upscale the first cropped image and the second cropped image to generate a first upscaled image and a second upscaled image using a ground truth image having high-resolution and corresponding to the first and second cropped image, respectively, and
to cancel noise from the first upscaled image and the second upscaled image to generate a first noise canceled image and a second noise canceled image, wherein
the noise canceling includes canceling noise based on a deep learning network trained on a layout of the semiconductor device, the layout corresponding to a field of view of the SEM image.
15. The super resolution SEM image implementing system of claim 14 , wherein the super resolution SEM image implementing device further includes:
a buffer configured to store a first position of the cropping of the first cropped image and a second position of the cropping of the second cropped image.
16. The super resolution SEM image implementing system of claim 14 , wherein super resolution SEM image implementing device is configured to,
based on a boundary line corresponding to the first cropped image and the second cropped image contacting each other,
generate a first-first cropped image further including a first area permeating from the boundary line into the second cropped image by a first length with respect to the first cropped image, and
generate a second-first cropped image further including a second area permeating from the boundary line into the first cropped image by a second length with respect to the second cropped image.
17. The super resolution SEM image implementing system of claim 16 , wherein super resolution SEM image implementing device is configured to upscale the first-first cropped image and the second-first cropped image to generate a first-first upscaled image and a second-first upscaled image.
18. The super resolution SEM image implementing system of claim 17 , wherein the super resolution SEM image implementing device is configured to cancel noise from the first-first upscaled image and the second-first upscaled image to generate a first-first noise canceled image and a second-first noise canceled image.
19. The super resolution SEM image implementing system of claim 14 , wherein the super resolution SEM image implementing device is configured to,
perform convolution on the first cropped image and the second cropped image; and
perform deconvolution on the first cropped image on which the convolution is performed and the second cropped image on which the convolution is performed, and
generate the first noise canceled image and the second noise canceled image by performing a selective skip connection on the first cropped image and the second cropped image to cancel the noise.
20. The super resolution SEM image implementing system of claim 14 , wherein the super resolution SEM image implementing device is configured to merge the first noise canceled image and the second noise canceled image with each other to generate a super resolution SEM image.