Research map / Video / Video segmentation and tracking
Video Reflection Removal Through Spatio-Temporal Optimization
Ajay Nandoriya, Mohamed Elgharib, Changil Kim, Mohamed Hefeeda, Wojciech Matusik
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 per-frame layer separation with temporal layer redundancy and layer prior (full spatio-temporal objective) versus Xue et al. ++ video extension | video reflection removal · Controlled synthetic sequence (Lena + buildings) | temporal PSD, high frequency >6 Hz: 0.0195 -> 0.0107 | Improved |
| add the full method versus Yang et al. ++ video extension | video reflection removal · Controlled synthetic sequence (Lena + buildings) | temporal PSD, high frequency >6 Hz: 0.0843 -> 0.0107 | Improved |
| add the full method versus Xue et al. ++ | video reflection removal · Controlled real sequence (black box and red box behind glass) | mean NCC: 0.77 -> 0.82 | Improved |
| add the full method versus Xue et al. ++ | video reflection removal · Controlled real sequence (black box and red box behind glass) | mean SSIM: 0.78 -> 0.82 | Improved |
| add the full method versus Yang et al. ++ | video reflection removal · Controlled real sequence (black box and red box behind glass) | mean NCC: 0.81 -> 0.82 | Improved |
| add the full method versus Yang et al. ++ | video reflection removal · Controlled real sequence (black box and red box behind glass) | mean SSIM: 0.79 -> 0.82 | Improved |
| add the full method versus Xue et al. ++ | video reflection removal · Controlled real sequence (black box and red box behind glass) | temporal PSD, high frequency >6 Hz: 0.0934 -> 0.0128 | Improved |
| add the full method versus Yang et al. ++ | video reflection removal · Controlled real sequence (black box and red box behind glass) | temporal PSD, high frequency >6 Hz: 0.01752 -> 0.0128 | Improved |
| add each objective term (image compositing, temporal layer redundancy, layer prior) (ablation) | full objective · Real sequences (Fig. 4, qualitative) | separation completeness | Improved |
| add iterative alignment instead of direct matching (ablation) | alignment strategy · Real sequences (Fig. 5, qualitative) | separation quality | Improved |
| add post-hoc temporal averaging filter ('++') to the '+' extension | Xue et al. video extension · Real sequences (Fig. 6, 11) | temporal flicker versus moving-object blur | Mixed |
Closest papers by meaning
- EvReflection: Event-Driven Micro-Dynamics for Reflection Removal (ICML 2026 · similarity 0.86)
- Single Image Reflection Removal Beyond Linearity (CVPR 2019 · similarity 0.86)
- Flash-Split: 2D Reflection Removal with Flash Cues and Latent Diffusion Separation (CVPR 2025 · similarity 0.84)
- FIRM: Flexible Interactive Reflection ReMoval (AAAI 2025 · similarity 0.84)
- MotionAura: Generating High-Quality and Motion Consistent Videos using Discrete Diffusion (ICLR 2025 · similarity 0.84)
- Revisiting Single Image Reflection Removal In the Wild (CVPR 2024 · similarity 0.84)
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.
