TY - GEN
T1 - Video Super Resolution Based on Motion Compensation
AU - Chen, Jiyuan
AU - Guo, Suping
AU - Deng, Jun
AU - Yang, Maolin
AU - Shen, Feng
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - An image super-resolution method and device provided by acquiring the captured video of the power equipment. The method according to the time sequence of the image frames in the video of the power equipment, extracts the appearance image of the power equipment, selects the reference image, obtains the target image, and obtains the motion vector of any pixel moving from the position coordinates in the target image to the position coordinates in the reference image. The reference images are input into the spatiotemporal sub-pixel convolution network, and then the two-dimensional super-resolution image is output, which improves the image clarity, and can image the moving power equipment with high resolution. In this paper We use time fusion strategy in the model of video SR. Our model constructs a motion compensation mechanism based on spatial transformation network. This mechanism can be trained jointly for video SR. The results obtained by our method are proved to be superior in peak signal to noise ratio (PSNR).
AB - An image super-resolution method and device provided by acquiring the captured video of the power equipment. The method according to the time sequence of the image frames in the video of the power equipment, extracts the appearance image of the power equipment, selects the reference image, obtains the target image, and obtains the motion vector of any pixel moving from the position coordinates in the target image to the position coordinates in the reference image. The reference images are input into the spatiotemporal sub-pixel convolution network, and then the two-dimensional super-resolution image is output, which improves the image clarity, and can image the moving power equipment with high resolution. In this paper We use time fusion strategy in the model of video SR. Our model constructs a motion compensation mechanism based on spatial transformation network. This mechanism can be trained jointly for video SR. The results obtained by our method are proved to be superior in peak signal to noise ratio (PSNR).
KW - motion compensation
KW - super-resolution
KW - video procession
UR - https://www.scopus.com/pages/publications/85127457028
U2 - 10.1109/CISAI54367.2021.00142
DO - 10.1109/CISAI54367.2021.00142
M3 - 会议稿件
AN - SCOPUS:85127457028
T3 - Proceedings - 2021 International Conference on Computer Information Science and Artificial Intelligence, CISAI 2021
SP - 704
EP - 707
BT - Proceedings - 2021 International Conference on Computer Information Science and Artificial Intelligence, CISAI 2021
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 4th International Conference on Computer Information Science and Artificial Intelligence, CISAI 2021
Y2 - 17 September 2021 through 19 September 2021
ER -