2022
DOI: 10.48550/arxiv.2206.02146
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

Abstract: Video restoration aims at restoring multiple high-quality frames from multiple lowquality frames. Existing video restoration methods generally fall into two extreme cases, i.e., they either restore all frames in parallel or restore the video frame by frame in a recurrent way, which would result in different merits and drawbacks. Typically, the former has the advantage of temporal information fusion. However, it suffers from large model size and intensive memory consumption; the latter has a relatively small mo… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1

Citation Types

0
14
0

Year Published

2022
2022
2023
2023

Publication Types

Select...
3
1
1

Relationship

1
4

Authors

Journals

citations
Cited by 6 publications
(14 citation statements)
references
References 70 publications
0
14
0
Order By: Relevance
“…However, training with 30 frames takes much longer time than with 16 frames, which makes this method uneconomical. In Table 7 we also compare the state-of-the-art contemporaneous work RVRT [23]. When RVRT uses 30 frames for training, it can achieve similar performance to PSRT-recurrent when it was trained with 16 frames.…”
Section: Reconstructionmentioning
confidence: 99%
See 2 more Smart Citations
“…However, training with 30 frames takes much longer time than with 16 frames, which makes this method uneconomical. In Table 7 we also compare the state-of-the-art contemporaneous work RVRT [23]. When RVRT uses 30 frames for training, it can achieve similar performance to PSRT-recurrent when it was trained with 16 frames.…”
Section: Reconstructionmentioning
confidence: 99%
“…The alignment module is critical for CNN-based VSR networks, because the locality inductive bias of CNNs only allows them to utilize spatial-close distributed information effectively. Many VSR networks have achieved better performance by introducing more advanced alignment methods [43,4,39,23]. these VSR Transformers still retain complex alignment modules.…”
Section: Introductionmentioning
confidence: 99%
See 1 more Smart Citation
“…Video deblurring aims at removing the blur artifacts from the input videos. Depending on the number of required input frames, there are multiple-frame [46,16,18,33,13,21,23] and single-frame [49,46,20] deblurring methods. EDVR [53] restores high-quality deblurred frames by first extracting features of multiple inputs and then conducting feature alignment and fusion.…”
Section: Related Workmentioning
confidence: 99%
“…Video super-resolution reconstructs HR video frames from the corresponding LR frames. There are several VSR methods [5,48,41,52,57,23] that use optical flow for explicit temporal alignment. Recently, RBPN [12] combines ideas from single-and multiple-frame SR for VSR, and estimates inter-frame motion to generate SR frames.…”
Section: Related Workmentioning
confidence: 99%