《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (4): 1277-1284.DOI: 10.11772/j.issn.1001-9081.2023040523
所属专题: 多媒体计算与计算机仿真
收稿日期:
2023-05-05
修回日期:
2023-08-18
接受日期:
2023-08-24
发布日期:
2023-12-04
出版日期:
2024-04-10
通讯作者:
刘蓉
作者简介:
刘扬(1999—),男,湖南长沙人,硕士研究生,主要研究方向:计算机视觉基金资助:
Yang LIU, Rong LIU(), Ke FANG, Xinyue ZHANG, Guangxu WANG
Received:
2023-05-05
Revised:
2023-08-18
Accepted:
2023-08-24
Online:
2023-12-04
Published:
2024-04-10
Contact:
Rong LIU
About author:
LIU Yang, born in 1999, M. S. candidate. His research interests include computer vision.Supported by:
摘要:
面对运动幅度较大的复杂场景,当前的视频超分辨率(VSR)算法在处理长序列时无法充分利用不同距离的帧间信息,难以精确地恢复遮挡、边界和多细节区域。为解决上述问题,提出一种基于帧间跨越光流机制的VSR模型。首先,通过密集残差块(RDB)提取低分辨率视频帧(LR)的浅层特征;其次,通过光流空间金字塔网络(SPyNet)以不同时间长度的跨越光流对视频帧进行运动估计和运动补偿,并通过RDB对帧间信息进行深层特征提取与矫正;最后,融合浅层特征与深层特征,并通过上采样得到高分辨率视频帧(HR)。在REDS4公开数据集上的实验结果表明,所提模型与经典的非显式运动补偿的动态上采样滤波器视频超分辨率网络(DUF-VSR)相比,峰值信噪比(PSNR)和结构相似性(SSIM)分别提升了1.07 dB和0.06。验证了所提模型可有效提高视频图像重建的质量。
中图分类号:
刘扬, 刘蓉, 方可, 张心月, 王光旭. 基于帧间跨越光流的视频超分辨率重建网络[J]. 计算机应用, 2024, 44(4): 1277-1284.
Yang LIU, Rong LIU, Ke FANG, Xinyue ZHANG, Guangxu WANG. Video super-resolution reconstruction network based on frame straddling optical flow[J]. Journal of Computer Applications, 2024, 44(4): 1277-1284.
模型 | 运行时间/ms | 参数量/106 | 浮点运算量/ GFLOPs | REDS4 | Vid4 | SPMC | UDM10 | ||||
---|---|---|---|---|---|---|---|---|---|---|---|
PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | ||||
Bicubic | — | — | — | 26.14 | 0.72 | 21.80 | 0.52 | 23.29 | 0.64 | 28.47 | 0.82 |
EDVR-m | 156 | 3.3 | 200.0 | 28.11 | 0.80 | 23.54 | 0.69 | 27.27 | 0.75 | 37.42 | 0.93 |
DUF-VSR | 559 | 5.8 | 736.6 | 29.40 | 0.81 | 25.01 | 0.73 | 27.96 | 0.79 | 37.97 | 0.93 |
BasicVSR | 1 306 | 6.3 | 163.7 | 30.21 | 0.85 | 24.94 | 0.74 | 28.52 | 0.82 | 38.20 | 0.94 |
RawVSR | 1 610 | 4.5 | 356.9 | 30.20 | 0.86 | 25.04 | 0.75 | 28.54 | 0.84 | 38.44 | 0.96 |
OVSR | 1 152 | 1.8 | 110.0 | 30.17 | 0.85 | 25.15 | 0.79 | 28.57 | 0.83 | 38.18 | 0.94 |
本文模型 | 1 544 | 6.3 | 336.4 | 30.47 | 0.87 | 25.17 | 0.79 | 28.59 | 0.83 | 38.31 | 0.95 |
表1 不同模型在4个数据集上的结果对比
Tab. 1 Results comparison of different models on four datasets
模型 | 运行时间/ms | 参数量/106 | 浮点运算量/ GFLOPs | REDS4 | Vid4 | SPMC | UDM10 | ||||
---|---|---|---|---|---|---|---|---|---|---|---|
PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | ||||
Bicubic | — | — | — | 26.14 | 0.72 | 21.80 | 0.52 | 23.29 | 0.64 | 28.47 | 0.82 |
EDVR-m | 156 | 3.3 | 200.0 | 28.11 | 0.80 | 23.54 | 0.69 | 27.27 | 0.75 | 37.42 | 0.93 |
DUF-VSR | 559 | 5.8 | 736.6 | 29.40 | 0.81 | 25.01 | 0.73 | 27.96 | 0.79 | 37.97 | 0.93 |
BasicVSR | 1 306 | 6.3 | 163.7 | 30.21 | 0.85 | 24.94 | 0.74 | 28.52 | 0.82 | 38.20 | 0.94 |
RawVSR | 1 610 | 4.5 | 356.9 | 30.20 | 0.86 | 25.04 | 0.75 | 28.54 | 0.84 | 38.44 | 0.96 |
OVSR | 1 152 | 1.8 | 110.0 | 30.17 | 0.85 | 25.15 | 0.79 | 28.57 | 0.83 | 38.18 | 0.94 |
本文模型 | 1 544 | 6.3 | 336.4 | 30.47 | 0.87 | 25.17 | 0.79 | 28.59 | 0.83 | 38.31 | 0.95 |
模型 | 运行时间/ms | 参数量/106 | 浮点运算量/ GFLOPs | REDS4-000 | REDS4-011 | REDS4-015 | REDS4-020 | ||||
---|---|---|---|---|---|---|---|---|---|---|---|
PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | ||||
ModelX | 1 129 | 4.6 | 192.2 | 27.72 | 0.80 | 30.13 | 0.85 | 32.16 | 0.88 | 29.18 | 0.85 |
BasicVSR | 1 306 | 6.3 | 163.7 | 27.67 | 0.81 | 30.83 | 0.87 | 32.75 | 0.90 | 29.61 | 0.87 |
本文模型 | 1 544 | 6.3 | 336.4 | 27.83 | 0.82 | 31.13 | 0.88 | 33.12 | 0.91 | 29.82 | 0.88 |
表2 不同模型在REDS4数据集上的性能对比
Tab. 2 Performance comparison of different models on REDS4 dataset
模型 | 运行时间/ms | 参数量/106 | 浮点运算量/ GFLOPs | REDS4-000 | REDS4-011 | REDS4-015 | REDS4-020 | ||||
---|---|---|---|---|---|---|---|---|---|---|---|
PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | PSNR/dB | SSIM | ||||
ModelX | 1 129 | 4.6 | 192.2 | 27.72 | 0.80 | 30.13 | 0.85 | 32.16 | 0.88 | 29.18 | 0.85 |
BasicVSR | 1 306 | 6.3 | 163.7 | 27.67 | 0.81 | 30.83 | 0.87 | 32.75 | 0.90 | 29.61 | 0.87 |
本文模型 | 1 544 | 6.3 | 336.4 | 27.83 | 0.82 | 31.13 | 0.88 | 33.12 | 0.91 | 29.82 | 0.88 |
P | R | PSNR/dB | SSIM | 运行时间/ms | 参数量/106 | 浮点运算量/ GFLOPs |
---|---|---|---|---|---|---|
4 | 6 | 30.38 | 0.85 | 1624 | 7.1 | 439.4 |
6 | 6 | 30.26 | 0.83 | 1731 | 7.7 | 484.0 |
3 | 6 | 30.41 | 0.87 | 1558 | 6.8 | 405.8 |
3 | 4 | 30.17 | 0.83 | 1436 | 5.8 | 267.0 |
3 | 5 | 30.47 | 0.87 | 1544 | 6.3 | 336.4 |
表3 不同的预处理残差块数和特征矫正残差块数对REDS4数据集上性能的影响
Tab. 3 Different numbers of PRDB and RDB on performance on REDS4 dataset
P | R | PSNR/dB | SSIM | 运行时间/ms | 参数量/106 | 浮点运算量/ GFLOPs |
---|---|---|---|---|---|---|
4 | 6 | 30.38 | 0.85 | 1624 | 7.1 | 439.4 |
6 | 6 | 30.26 | 0.83 | 1731 | 7.7 | 484.0 |
3 | 6 | 30.41 | 0.87 | 1558 | 6.8 | 405.8 |
3 | 4 | 30.17 | 0.83 | 1436 | 5.8 | 267.0 |
3 | 5 | 30.47 | 0.87 | 1544 | 6.3 | 336.4 |
模型 | 5帧前向分支 | 3帧后向分支 | 预处理 | 长距离复用 | PSNR/dB | SSIM | 运行时间/ms | 参数量/106 | 浮点运算量/GFLOPs |
---|---|---|---|---|---|---|---|---|---|
Model1 | 29.92 | 0.81 | 1 066 | 3.7 | 125.2 | ||||
Model2 | √ | 30.00 | 0.81 | 1 282 | 4.5 | 183.1 | |||
Model3 | √ | √ | 30.25 | 0.83 | 1 481 | 5.4 | 270.7 | ||
Model4 | √ | √ | √ | 30.33 | 0.84 | 1 543 | 6.3 | 336.4 | |
本文模型 | √ | √ | √ | √ | 30.47 | 0.87 | 1 544 | 6.3 | 336.4 |
表4 各模块在REDS4数据集上的有效性实验结果
Tab. 4 Validation experimental results of different modules on REDS4 dataset
模型 | 5帧前向分支 | 3帧后向分支 | 预处理 | 长距离复用 | PSNR/dB | SSIM | 运行时间/ms | 参数量/106 | 浮点运算量/GFLOPs |
---|---|---|---|---|---|---|---|---|---|
Model1 | 29.92 | 0.81 | 1 066 | 3.7 | 125.2 | ||||
Model2 | √ | 30.00 | 0.81 | 1 282 | 4.5 | 183.1 | |||
Model3 | √ | √ | 30.25 | 0.83 | 1 481 | 5.4 | 270.7 | ||
Model4 | √ | √ | √ | 30.33 | 0.84 | 1 543 | 6.3 | 336.4 | |
本文模型 | √ | √ | √ | √ | 30.47 | 0.87 | 1 544 | 6.3 | 336.4 |
1 | SON S, LEE S, NAH S, et al. NTIRE 2021 challenge on video super-resolution [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2021: 166-181. |
2 | ZHANG L, ZHANG H, SHEN H,et al. A super-resolution reconstruction algorithm for surveillance images[J]. Signal Processing,2010,90(3): 848-859. 10.1016/j.sigpro.2009.09.002 |
3 | SCHULTZ R R, STEVENSON R L. Extraction of high-resolution frames from video sequences[J]. IEEE Transactions on Image Processing, 1996, 5(6): 996-1011. 10.1109/83.503915 |
4 | CHAN K C K, WANG X, YU K, et al. BasicVSR: the search for essential components in video super-resolution and beyond [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 4945-4954. 10.1109/cvpr46437.2021.00491 |
5 | TIAN Y, ZHANG Y, FU Y, et al. TDAN: temporally-deformable alignment network for video super-resolution [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 3360-3369. 10.1109/cvpr42600.2020.00342 |
6 | DAI J, QI H, XIONG Y, et al. Deformable convolutional networks [C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 764-773. 10.1109/iccv.2017.89 |
7 | WANG X, CHAN K C K, YU K, et al. EDVR: video restoration with enhanced deformable convolutional networks [C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2019: 1954-1963. 10.1109/cvprw.2019.00247 |
8 | LI W, TAO X, GUO T,et al. MuCAN: multi-correspondence aggregation network for video super-resolution[C]// Proceedings of the 16th European Conference on Computer Vision. Cham: Springer, 2020: 335-351. 10.1007/978-3-030-58607-2_20 |
9 | JO Y, OH S W, KANG J, et al. Deep video super-resolution network using dynamic upsampling filters without explicit motion compensation [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 3224-3232. 10.1109/cvpr.2018.00340 |
10 | CAO J, LI Y, ZHANG K, et al. Video super-resolution transformer [EB/OL]. [2023-05-01]. . |
11 | ISOBE T, JIA X, GU S, et al. Video super-resolution with recurrent structure-detail network [C]// Proceedings of the 16th European Conference on Computer Vision. Cham: Springer, 2020: 645-660. 10.1007/978-3-030-58610-2_38 |
12 | ISOBE T, ZHU F, JIA X, et al. Revisiting temporal modeling for video super-resolution [EB/OL]. [2023-05-01]. . |
13 | JIANG L, WANG N, DANG Q, et al. PP-MSVSR: multi-stage video super-resolution [EB/OL]. [2023-05-01]. . |
14 | ISOBE T, JIA X, TAO X, et al. Look back and forth: video super-resolution with explicit temporal difference modeling [C]// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 17411-17420. 10.1109/cvpr52688.2022.01689 |
15 | XIAO Y, YUAN Q, JIANG K, et al. Local-global temporal difference learning for satellite video super-resolution [EB/OL]. [2023-05-01]. . 10.1109/tcsvt.2023.3312321 |
16 | CHAN K C K, ZHOU S, XU X, et al. BasicVSR++: improving video super-resolution with enhanced propagation and alignment [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 5972-5981. 10.1109/cvpr52688.2022.00588 |
17 | CHAN K C K, ZHOU S, XU X, et al. Investigating tradeoffs in real-world video super-resolution [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2021: 5962-5971. 10.1109/cvpr52688.2022.00587 |
18 | LIANG J, CAO J, SUN G, et al. SwinIR: image restoration using swin transformer [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 1833-1844. 10.1109/iccvw54120.2021.00210 |
19 | RANJAN A, BLACK M J. Optical flow estimation using a spatial pyramid network [C] // Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 4161-4170. 10.1109/cvpr.2017.291 |
20 | SONG Y, ZHU Y, DU X. Dynamic residual dense network for image denoising [J]. Sensors, 2019, 19(17): 3809. 10.3390/s19173809 |
21 | WANG X, XIE L, DONG C, et al. Real-ESRGAN: training real-world blind super-resolution with pure synthetic data [C] // Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 1905-1914. 10.1109/iccvw54120.2021.00217 |
22 | YAMANAKA J, KUWASHIMA S, KURITA T. Fast and accurate image super resolution by deep CNN with skip connection and network in network [C]// Proceedings of the 24th International Conference Neural Information Processing. Cham: Springer, 2017: 217-225. 10.1007/978-3-319-70096-0_23 |
23 | CHEN Y, WU T. SATVSR: scenario adaptive transformer for cross scenarios video super-resolution [J]. Journal of Physics: Conference Series, 2023, 2456: 012028. 10.1088/1742-6596/2456/1/012028 |
24 | CABALLERO J, LEDIG C, AITKEN A, et al. Real-time video super-resolution with spatio-temporal networks and motion compensation [C] // Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 4778-4787. 10.1109/cvpr.2017.304 |
25 | LAI W-S, HUANG J-B, AHUJA N, et al. Fast and accurate image super-resolution with deep laplacian pyramid networks [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(11): 2599-2613. 10.1109/tpami.2018.2865304 |
26 | NAH S, BAIK S, HONG S, et al. NTIRE 2019 challenge on video deblurring and super-resolution: dataset and study [C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway: IEEE, 2019: 1996-2005. 10.1109/cvprw.2019.00251 |
27 | HARIS M, SHAKHNAROVICH G, UKITA N. Recurrent back-projection network for video super-resolution [C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 3897-3906. 10.1109/cvpr.2019.00402 |
28 | KINGMA D P, BA J. Adam: a method for stochastic optimization [EB/OL]. [2023-05-01]. . |
29 | LOSHCHILOV I, HUTTER F. SGDR: stochastic gradient descent with warm restarts [EB/OL]. [2023-05-01]. . |
30 | 佟雨兵,张其善,祁云平.基于PSNR与SSIM联合的图像质量评价模型[J].中国图象图形学报,2006, 11(12): 1758-1763. 10.11834/jig.2006012307 |
TONG Y B, ZHANG Q S, QI Y P. Image quality assessing by combining PSNR with SSIM [J]. Journal of Image and Graphics, 2006, 11(12): 1758-1763. 10.11834/jig.2006012307 | |
31 | LIU X, SHI K, WANG Z, et al. Exploit camera raw data for video super-resolution via hidden Markov model inference [J]. IEEE Transactions on Image Processing, 2021, 30: 2127-2140. 10.1109/tip.2021.3049974 |
32 | YI P, WANG Z, JIANG K, et al. Omniscient video super-resolution [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 4429-4438. 10.1109/iccv48922.2021.00439 |
[1] | 张晔, 刘蓉, 刘明, 陈明. 基于多通道注意力机制的图像超分辨率重建网络[J]. 《计算机应用》唯一官方网站, 2022, 42(5): 1563-1569. |
[2] | 刘士豪, 胡学敏, 姜博厚, 张若晗, 孔力. 基于生成对抗双网络的虚拟到真实驾驶场景的视频翻译模型[J]. 计算机应用, 2020, 40(6): 1621-1626. |
[3] | 张家岗, 李达平, 杨晓东, 邹茂扬, 吴锡, 胡金蓉. 基于深度卷积特征光流的形变医学图像配准算法[J]. 计算机应用, 2020, 40(6): 1799-1805. |
[4] | 聂可卉, 刘文哲, 童同, 杜民, 高钦泉. 基于自适应可分离卷积核的视频压缩伪影去除算法[J]. 计算机应用, 2019, 39(5): 1473-1479. |
[5] | 廖斌, 吴文. 区域配对引导的光照传播视频阴影去除方法[J]. 计算机应用, 2019, 39(2): 556-563. |
[6] | 魏震宇, 文畅, 谢凯, 贺建飚. 光流估计下的移动端实时人脸检测[J]. 计算机应用, 2018, 38(4): 1146-1150. |
[7] | 胡学敏, 易重辉, 陈钦, 陈茜, 陈龙. 基于运动显著图的人群异常行为检测[J]. 计算机应用, 2018, 38(4): 1164-1169. |
[8] | 丁飞飞, 杨文元. 信息熵约束下的视频目标分割[J]. 计算机应用, 2018, 38(10): 2782-2787. |
[9] | 朱明敏, 胡茂海. 基于相关滤波器的长时视觉目标跟踪方法[J]. 计算机应用, 2017, 37(5): 1466-1470. |
[10] | 魏玮, 马瑞, 王小芳. 视频中人脸位置的定量检测[J]. 计算机应用, 2017, 37(3): 801-805. |
[11] | 厉丹, 鲍蓉, 孙金萍, 肖理庆, 党向盈. 多分辨率LK光流联合SURF的跟踪方法[J]. 计算机应用, 2017, 37(3): 806-810. |
[12] | 李雪君, 张开华, 宋慧慧. 融合时空多特征表示的无监督视频分割算法[J]. 计算机应用, 2017, 37(11): 3134-3138. |
[13] | 邹洪中, 许悦雷, 马时平, 李帅, 张文达. 基于视皮层V1模型的随机点视频序列运动特征提取[J]. 计算机应用, 2016, 36(6): 1677-1681. |
[14] | 高智勇, 唐文峰, 贺良杰. 基于运动显著性的移动镜头下的运动目标检测[J]. 计算机应用, 2016, 36(6): 1692-1698. |
[15] | 潘磊, 周欢, 王明辉. 适用于密集人群的异常事件实时检测方法[J]. 计算机应用, 2016, 36(6): 1719-1723. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||