现代电子技术2025，Vol.48Issue(9)：36-42,7.DOI:10.16652/j.issn.1004-373x.2025.09.006
基于时空可变形卷积的多帧视频质量增强方法研究
Multi-frame video quality enhancement method based on spatio-temporal deformable convolution
摘要
Abstract
The existing multi-frame video quality enhancement methods typically rely on a single neighboring frame or peak quality frame(PQF),and fails to fully exploit the temporal characteristics of the video stream.In view of the above,this paper proposes a multi-frame video quality enhancement method based on spatio-temporal deformable convolution.Time information is aggregated by spatio-temporal deformable convolution,so as to improve the accuracy of motion estimation.Spatio-temporal deformable fusion and spatio-temporal feature fusion are used to extract spatio-temporal features from PQF and neighboring frames,respectively.The two groups of feature maps extracted are fused and then processed by the feature refinement module.The enhanced feature map is outputted and added to the residual of the frame to be enhanced,resulting in the final enhanced frame.Experimental results show that the performance of the proposed method outperforms that of the contrastive methods on public test datasets.It achieves a gain of 0.76 dB in peak signal-to-noise ratio(PSNR)and improvements of 0.20 dB and 0.11 dB over MFQE2.0 and STDF-R1,respectively.关键词
时空可变形卷积/多帧视频/时空特征/视频质量增强/运动估计/特征融合Key words
spatio-temporal deformable convolution/multi-frame video/spatio-temporal feature/video quality enhancement/motion estimation/feature fusion分类
信息技术与安全科学引用本文复制引用
李娟,何月顺,何璘琳,庞振宇,戴天峦,邓钰嫣,许亚男..基于时空可变形卷积的多帧视频质量增强方法研究[J].现代电子技术,2025,48(9):36-42,7.基金项目
江西省重点研发项目(20232ABC03A09) （20232ABC03A09）
2022年度博士科研启动基金项目:旅游与交通多模型多尺度关联关系研究(DHBK2022004) （DHBK2022004）
