Research map / Video / Video segmentation and tracking
FastDVDnet: Towards Real-Time Deep Video Denoising Without Flow Estimation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 two-step cascading rather than one five-input denoising block (ablation) | FastDVDnet · Four sequences, σ=10, 30, 50 | PSNR: 0.95dB | Helped |
| Use multi-scale rather than single-scale denoising blocks (ablation) | FastDVDnet · Four sequences, σ=10, 30, 50 | PSNR: 0.55dB | Helped |
| Train the two denoising steps end-to-end rather than separately (ablation) | FastDVDnet · Video sequences | temporal artifacts | Helped |
| Replace ReLU with Leaky ReLU (ablation) | FastDVDnet · All tested sequences and noise levels | PSNR: less than 0.05dB | No clear effect |
| Replace ReLU with ELU (ablation) | FastDVDnet · All tested sequences and noise levels | PSNR: less than 0.05dB | No clear effect |
| Replace L2 loss with Huber loss (ablation) | FastDVDnet · All tested sequences and noise levels | PSNR: 0.04dB | No clear effect |
| Remove batch normalization layers (ablation) | FastDVDnet · All tested sequences and noise levels | PSNR: 0.18dB | Hurt |
| Increase input frames from 5 to 7 or 9 (ablation) | FastDVDnet · All tested sequences and noise levels | PSNR | No clear effect |
| Train a model for clipped AWGN | FastDVDnet · DAVIS, clipped AWGN | PSNR: 37.13 -> 38.45; 32.24 -> 33.52; 29.77 -> 31.23 | Helped |
Closest papers by meaning
- Classic Video Denoising in a Machine Learning World: Robust, Fast, and Controllable (CVPR 2025 · similarity 0.90)
- Universal Denoising Networks : A Novel CNN Architecture for Image Denoising (CVPR 2018 · similarity 0.89)
- Real-Time Controllable Denoising for Image and Video (CVPR 2023 · similarity 0.89)
- Efficient Multi-Stage Video Denoising With Recurrent Spatio-Temporal Fusion (CVPR 2021 · similarity 0.88)
- Toward Convolutional Blind Denoising of Real Photographs (CVPR 2019 · similarity 0.88)
- Generative Adaptive Convolutions for Real-World Noisy Image Denoising (AAAI 2022 · 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.