Research map / Video / Video segmentation and tracking
Learning to Extract Flawless Slow Motion From Blurry Videos
Meiguang Jin, Zhe Hu, Paolo Favaro
Check 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 |
|---|---|---|---|
| Use the two-network TNTT design instead of the strongest listed conventional deblurring baseline | DeblurNet + InterpNet (TNTT) · Videos #1–#8; video #9 is an exception | PSNR: 33.9 vs 33.2; 34.6 vs 32.5; 29.3 vs 28.0; 28.0 vs 27.0; 26.3 vs 26.0; 38.3 vs 36.5; 31.4 vs 29.4; 35.5 vs 32.1 | Improved |
| Use TNTT instead of the strongest conventional deblurring/interpolation combination | DeblurNet + InterpNet (TNTT) · Videos #1–#7 and #9 | PSNR: 32.4 vs 30.7; 33.2 vs 30.5; 28.8 vs 26.0; 27.5 vs 25.1; 26.9 vs 24.2; 37.5 vs 35.2; 31.8 vs 29.2; 36.5 vs 36.3 | Improved |
| Add the temporal-transition error term to the two-network method (ablation) | TN / TNTT · Synthetic test videos | temporal smoothness (optical-flow gradient error) | Improved |
| Add the temporal-transition error term (ablation) | TN / TNTT · Video #1 (GoPro) | PSNR: 32.2 -> 32.4 | Improved |
| Add the temporal-transition error term (ablation) | TN / TNTT · Video #1 (GoPro) | PSNR: 33.9 -> 33.9 | No clear effect |
| Replace the naïve single-network output with the two-network design, without the temporal-transition term (ablation) | Slow-motion reconstruction method · Video #1 (GoPro) | PSNR: 32.3 -> 32.2 | No clear effect |
| Use TNTT rather than the naïve all-frames-at-once approach (ablation) | Slow-motion reconstruction method · Video #1 (GoPro) | PSNR: 32.3 -> 32.4 | Improved |
| Use TNTT instead of the conventional deblurring baseline | DeblurNet · Video #9 (Sony) | PSNR: 39.2 -> 38.7 | Hurt |
Closest papers by meaning
- Unifying Motion Deblurring and Frame Interpolation With Events (CVPR 2022 · similarity 0.90)
- Learning to Synthesize Motion Blur (CVPR 2019 · similarity 0.90)
- Bringing Alive Blurred Moments (CVPR 2019 · similarity 0.88)
- Learning to Extract a Video Sequence From a Single Motion-Blurred Image (CVPR 2018 · similarity 0.88)
- Event-Based Frame Interpolation With Ad-Hoc Deblurring (CVPR 2023 · similarity 0.88)
- Motion-blurred Video Interpolation and Extrapolation (AAAI 2021 · similarity 0.87)
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.
