Research map / Video / Video segmentation and tracking
Deep Kalman Filtering Network for Video Compression Artifact Reduction
Guo Lu, Wanli Ouyang, Dong Xu, Xiaoyun Zhang, Zhiyong Gao, Ming-Ting Sun
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 quantized prediction residual as input to the measurement network (ablation) | Measurement network · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 37.15 -> 37.49 | Improved |
| add recursive filtering using previous restored frame to the temporal mapping network (ablation) | Temporal mapping network · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 37.35 -> 37.76 | Improved |
| combine measurement and temporal networks through Kalman update (ablation) | DKFN · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 37.76 -> 37.93 | Improved |
| combine measurement and temporal networks through Kalman update vs measurement only (ablation) | DKFN · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 37.49 -> 37.93 | Improved |
| replace convolutional GRU recurrence with Kalman filter (ablation) | DKFN · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 37.10 -> 37.93 | Improved |
| add prediction residual to measurement network (ablation) | Measurement network · Vimeo-90K JPEG2000 q=20 | SSIM: 0.967 -> 0.968 | Improved |
| add recursive filtering to temporal network (ablation) | Temporal mapping network · Vimeo-90K JPEG2000 q=20 | SSIM: 0.967 -> 0.970 | Improved |
| use DKFN instead of Tolow | DKFN · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 36.92 -> 37.93 | Improved |
| use DKFN instead of DnCNN | DKFN · Vimeo-90K JPEG2000 q=20 | PSNR (dB): 37.26 -> 37.93 | Improved |
| use DKFN instead of DnCNN | DKFN · Vimeo-90K HEVC qp=37 | PSNR (dB): 33.01 -> 33.23 | Improved |
| use DKFN instead of HEVC loop filter | DKFN · Vimeo-90K HEVC qp=37 | PSNR (dB): 31.98 -> 33.23 | Improved |
| use DKFN instead of Tolow on unseen data | DKFN · MPI Sintel (cross-dataset) | PSNR (dB): 34.78 -> 37.01 | Improved |
Closest papers by meaning
- One-To-Many Network for Visually Pleasing Compression Artifacts Reduction (CVPR 2017 · similarity 0.86)
- Neural Compression-Based Feature Learning for Video Restoration (CVPR 2022 · similarity 0.86)
- Learning End-to-End Lossy Image Compression: A Benchmark (TPAMI 2021 · similarity 0.85)
- Insights From Generative Modeling for Neural Video Compression (TPAMI 2023 · similarity 0.85)
- Learning to Extract a Video Sequence From a Single Motion-Blurred Image (CVPR 2018 · similarity 0.84)
- Learned Video Compression (ICCV 2019 · 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.
