Real-Time Video Deblurring via Lightweight Motion Compensation
Not on the map. Why?
Abstract (arXiv)
While motion compensation greatly improves video deblurring quality, separately performing motion compensation and video deblurring demands huge computational overhead. This paper proposes a real-time video deblurring framework consisting of a lightweight multi-task unit that supports both video deblurring and motion compensation in an efficient way. The multi-task unit is specifically designed to handle large portions of the two tasks using a single shared network, and consists of a multi-task detail network and simple networks for deblurring and motion compensation. The multi-task unit minimizes the cost of incorporating motion compensation into video deblurring and enables real-time deblurring. Moreover, by stacking multiple multi-task units, our framework provides flexible control between the cost and deblurring quality. We experimentally validate the state-of-the-art deblurring quality of our approach, which runs at a much faster speed compared to previous methods, and show practical real-time performance (30.99dB@30fps measured in the DVD dataset).
Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 31 of its 53 references that are in the map. First posted on arXiv 25 May 2022.
Builds on
31 of 53 references.
References · 31
- 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
- 2018YOLOv3: An Incremental Improvementcited by 2,652
- 2015Spatial Transformer Networkscited by 1,423
- 2017Mixed Precision Trainingcited by 1,135
- 2015FlowNet: Learning Optical Flow with Convolutional Networkscited by 1,054
- 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 966
- 2015Accurate Image Super-Resolution Using Very Deep Convolutional Networkscited by 902
- 2017GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networkscited by 816
- 2021Multi-Stage Progressive Image Restorationcited by 813
- 2018Multi-Task Learning as Multi-Objective Optimizationcited by 809
- 2016Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurringcited by 722
- 2017Video Enhancement with Task-Oriented Flowcited by 612
19 more
- 2019EDVR: Video Restoration with Enhanced Deformable Convolutional Networkscited by 492
- 2017DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networkscited by 366
- 2018Scale-recurrent Network for Deep Image Deblurringcited by 332
- 2016UberNet: Training a `Universal' Convolutional Neural Network for Low-, Mid-, and High-Level Vision using Diverse Datasets and Limited Memorycited by 200
- 2016EnhanceNet: Single Image Super-Resolution Through Automated Texture Synthesiscited by 177
- 2019Spatio-Temporal Filter Adaptive Network for Video Deblurringcited by 106
- 2020Cascaded Deep Video Deblurring Using Temporal Sharpness Priorcited by 85
- 2021Iterative Filter Adaptive Network for Single Image Defocus Deblurringcited by 85
- 2021ARVo: Learning All-Range Volumetric Correspondence for Video Deblurringcited by 54
- 2021Single Image Defocus Deblurring Using Kernel-Sharing Parallel Atrous Convolutionscited by 54
- 2017Online Video Deblurring via Dynamic Temporal Blending Networkcited by 53
- 2021Recurrent Video Deblurring with Blur-Invariant Motion Estimation and Pixel Volumescited by 41
- 2015Generalized Video Deblurring for Dynamic Scenescited by 40
- 2019EdgeStereo: An Effective Multi-Task Learning Network for Stereo Matching and Edge Detectioncited by 39
- 2018Learning Dual Convolutional Neural Networks for Low-Level Visioncited by 35
- 2016Integrated perception with recurrent multi-task neural networkscited by 24
- 2022Reference-based Video Super-Resolution Using Multi-Camera Video Tripletscited by 19
- 2015Hand-held Video Deblurring via Efficient Fourier Aggregationcited by 8
- 2019DAVID: Dual-Attentional Video Deblurringcited by 3
Built on by
Citing papers per year: 2023: 2; 2024: 0; 2025: 1.
3 papers.