Research map / Video / Video segmentation and tracking
Online Video Deblurring via Dynamic Temporal Blending Network
Tae Hyun Kim, Kyoung Mu Lee, Bernhard Scholkopf, Michael Hirsch
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add spatio-temporal recurrence (STRCNN) to the plain CNN, 5 input frames (ablation) | video deblurring CNN · 25 test videos, 5 frames | PSNR (dB): 29.25 -> 29.63 | Improved |
| add dynamic temporal blending to STRCNN, 5 input frames (ablation) | STRCNN · 25 test videos, 5 frames | PSNR (dB): 29.63 -> 29.76 | Improved |
| add dynamic temporal blending to STRCNN, 3 input frames (ablation) | STRCNN · 3 input frames | PSNR (dB): 29.45 -> 29.61 | Improved |
| add dynamic temporal blending to STRCNN, 7 input frames (ablation) | STRCNN · 7 input frames | PSNR (dB): 29.37 -> 29.50 | Improved |
| add spatio-temporal recurrence to CNN across blur strengths (ablation) | video deblurring CNN · varying blur strength, 5 frames | average PSNR gain over baseline: 0.2dB higher than the baseline | Improved |
| add recurrence and dynamic temporal blending across blur strengths (ablation) | video deblurring CNN · varying blur strength, 5 frames | average PSNR gain over baseline: 0.37dB against the baseline | Improved |
| increase input blurry frames from 5 to 7 in STRCNN+DTB (ablation) | STRCNN+DTB · 25 test videos | PSNR (dB): 29.76 -> 29.50 | Hurt |
| reduce input frames from 5 to 3 in STRCNN+DTB (ablation) | STRCNN+DTB · 25 test videos | PSNR (dB): 29.76 -> 29.61 | Hurt |
| STRCNN+DTB vs Su et al. FLOW baseline | video deblurring · 25 test videos | PSNR (dB): 28.81 -> 29.02 | Improved |
| STRCNN+DTB online run vs Su et al. NOALIGN | video deblurring · 100 HD frames | run time (sec): ~21 -> ~12.5 | Improved |
| place dynamic temporal blending between encoder and decoder rather than inside them (ablation) | STRCNN+DTB · layout choice | PSNR | Improved |
Closest papers by meaning
- Efficient Spatio-Temporal Recurrent Neural Network for Video Deblurring (ECCV 2020 · similarity 0.91)
- Real-World Video Deblurring: A Benchmark Dataset and an Efficient Recurrent Neural Network (IJCV 2022 · similarity 0.91)
- Learning Blind Motion Deblurring (ICCV 2017 · similarity 0.91)
- Spatio-Temporal Filter Adaptive Network for Video Deblurring (ICCV 2019 · similarity 0.90)
- Frequency-aware Event-based Video Deblurring for Real-World Motion Blur (CVPR 2024 · similarity 0.88)
- Blur-aware Spatio-temporal Sparse Transformer for Video Deblurring (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
