Research map / Image generation / Image restoration
EDformer: Transformer-Based Event Denoising Across Varied Noise Levels
Bin Jiang, Bo Xiong, Bohan Qu, M. Salman Asif, You Zhou, Zhan Ma
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 |
|---|---|---|---|
| use both large-scale and small-scale branches instead of a single fixed-scale branch (ablation) | EDformer · DND21 hotel-bar 5 Hz/pixel | AUC: 0.9628 -> 0.9845 | Improved |
| add SCformer and GXformer to LXformer (ablation) | EDformer · DND21 hotel-bar 5 Hz/pixel | AUC: 0.9720 -> 0.9841 | Improved |
| add GXformer (global attention) in the large-scale branch (ablation) | EDformer · DND21 hotel-bar 5 Hz/pixel | AUC: 0.9845 -> 0.9841 | Hurt |
| reduce KNN neighbors in LXformer from 16 to 8 (ablation) | EDformer · DND21 driving 5 Hz/pixel | AUC: 0.9424 -> 0.9396 | Hurt |
| increase KNN neighbors in LXformer from 16 to 32 (ablation) | EDformer · DND21 driving 5 Hz/pixel | AUC: 0.9424 -> 0.9324 | Hurt |
| reduce ball query window from 9 to 5 in SCformer (ablation) | EDformer · DND21 driving 5 Hz/pixel | AUC: 0.9424 -> 0.9379 | Hurt |
| use EDformer instead of MLPF at high noise rate | event denoising · DND21 hotel-bar 10 Hz/pixel | AUC: 0.9634 -> 0.9699 | Improved |
| use EDformer instead of EventZoom | event denoising · DND21 (MESR) | MESR: 1.059 -> 1.041 | Hurt |
| use only LXformer to speed up inference (ablation) | EDformer · inference speed | time per event: 22 us -> ~9 us (accuracy decreases) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- EventZoom — EventZoom: Learning To Denoise and Super Resolve Neuromorphic Events (CVPR 2021) · better on 1 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LED: A Large-scale Real-world Paired Dataset for Event Camera Denoising (CVPR 2024 · similarity 0.90)
- BiEvLight: Bi-level Learning of Task-Aware Event Refinement for Low-Light Image Enhancement (CVPR 2026 · similarity 0.87)
- Simultaneous Motion And Noise Estimation with Event Cameras (ICCV 2025 · similarity 0.87)
- Low Cost and Latency Event Camera Background Activity Denoising (TPAMI 2022 · similarity 0.87)
- Towards Robust Event-guided Low-Light Image Enhancement: A Large-Scale Real-World Event-Image Dataset and Novel Approach (CVPR 2024 · similarity 0.85)
- Motion-adaptive Transformer for Event-based Image Deblurring (AAAI 2025 · 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.
