Optics and Precision Engineering, Volume. 31, Issue 16, 2430(2023)
Multi-stage frame alignment video super- resolution network
Figures & Tables(16)
Fig. 9. Performance and model complexity comparison on Vid4 dataset for BD task
Table 1. Quantitative comparison［10］
View tableView in ArticleTable 1. Quantitative comparison［10］
对比网络 参数量/M 运行时间/ms BI（PSNR/SSIM） BD（PSNR/SSIM） REDS4 Vid4 Vid4 Vimeo-90K-T Bicubic — — 26.14/0.729 2 23.78/0.634 7 21.80/0.524 6 31.30/0.868 7 VESPCN［ 13 ］— — — 25.35/0.755 7 — — TOF［ 14 ］— — 27.98/0.799 0 25.89/0.765 1 — 34.62/0.921 2 DUF［ 19 ］5.8 974 28.63/0.825 1 — 27.38/0.832 9 36.87/0.944 7 FRVSR［ 33 ］5.1 137 — — 26.69/0.810 3 35.64/0.931 9 SPMC［ 15 ］- - — 25.88/0.775 2 — — EDVR［ 12 ］20.6 378 31.09/0.880 0 27.35/0.826 4 27.85/0.850 3 37.81/0.952 3 EDVR-M［ 12 ］3.3 118 30.53/0.869 9 27.10/0.818 6 27.45/0.840 6 37.33/0.948 4 MuCAN［ 39 ］— — 30.88/0.875 0 — — — TGA［ 40 ］5.8 — 30.88/0.875 0 — — — RLSP［ 24 ］4.2 49 — — 27.48/0.838 8 36.49/0.940 3 RSDN［ 32 ］6.2 94 — — 27.92/0.850 5 37.23/0.947 1 RRN［ 41 ］3.4 45 — — 27.92/0.850 5 37.23/0.947 1 PFNL［ 43 ］3.0 295 29.63/0.850 2 26.73/0.802 9 27.16/0.833 5 - RBPN［ 42 ］12.2 1 507 30.09/0.859 0 27.12/0.818 0 — 37.20/0.945 8 BaiscVSR［ 10 ］6.3 63 31.42/0.890 9 27.24/0.825 1 27.96/0.855 3 37.53/0.949 8 MSVSR 3.5 245 31.64/0.895 8 27.41/0.833 4 28.12/0.858 2 37.70/0.951 4
Table 2. Ablation tests on Vid4 dataset for BD task
View tableView in ArticleTable 2. Ablation tests on Vid4 dataset for BD task
FFB Pre-aligned PSNR SSIM √ √ 28.12 0.858 2 √ × 28.11 0.858 8 × √ 28.10 0.858 6 × × 28.08 0.856 9
Table 3. Quantitative comparison of different convolution decomposition approaches on Vid4 dataset for BD task
View tableView in ArticleTable 3. Quantitative comparison of different convolution decomposition approaches on Vid4 dataset for BD task
Method Params PSNR SSIM Shift-MLP（DWConv） 3.53 M 28.12 0.858 2 Shift-MLP（Conv） 3.91 M 28.14 0.859 9
Tools
Get Citation
Copy Citation Text
Sen WANG, Yang ZHU, Yinhui ZHANG, Qingjian WANG, Zifen HE. Multi-stage frame alignment video super- resolution network[J]. Optics and Precision Engineering, 2023, 31(16): 2430
Paper Information
Category: Information Sciences
Received: Dec. 14, 2022
Accepted: --
Published Online: Sep. 5, 2023
The Author Email: ZHANG Yinhui (zhangyinhui@kust.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20