Research map / Image generation / Image restoration
From Abyssal Darkness to Blinding Glare: A Benchmark on Extreme Exposure Correction in Real World
Bo Wang, Huiyuan Fu, Zhiye Huang, Siru Zhang, Xin Wang, Huadong 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 |
|---|---|---|---|
| add contrastive loss LCL separating over/under-exposed latents (ablation) | CLIER · REED | PSNR: 15.29 -> 19.25 | Improved |
| add CLIP image-embedding conditioning to the diffusion model (ablation) | CLIER SEDM · REED | PSNR: 18.67 -> 19.25 | Improved |
| set encoder depth to 3 rather than 1 or 2 (ablation) | CLIER encoder · REED | PSNR: 16.10 -> 19.25 (k=1 vs k=3) | Improved |
| increase encoder depth from 3 to 4 (ablation) | CLIER encoder · REED | PSNR: 19.25 -> 18.92 | Hurt |
| add Luminance Normalizer and moving variance loss to RECNet (ablation) | RECNet · REED | PSNR: 18.50 -> 18.92 | Improved |
| add CLIP-guided iterative refinement (CIRS) at inference (ablation) | RECNet · REED | PSNR: 18.92 -> 19.13 | Improved |
| add LN, Lvar and CIRS to CSEC (ablation) | CSEC · REED | PSNR: 18.71 -> 19.18 | Improved |
| add CIRS to CLIER (ablation) | CLIER · REED | PSNR: 19.13 -> 19.25 | Improved |
| iterate the model more than five times (ablation) | RECNet · MSEC-trained RECNet iterations | image quality | Hurt |
| train on extreme-exposure REED-aware method CLIER vs non-extreme methods | CLIER vs CSEC · REED | PSNR: 18.71 -> 19.25 | Improved |
Closest papers by meaning
- Luminance-Aware Statistical Quantization: Unsupervised Hierarchical Learning for Illumination Enhancement (NeurIPS 2025 · similarity 0.88)
- Rethinking No-reference Image Exposure Assessment from Holism to Pixel: Models, Datasets and Benchmarks (NeurIPS 2024 · similarity 0.87)
- Real-Time Exposure Correction via Collaborative Transformations and Adaptive Sampling (CVPR 2024 · similarity 0.86)
- AGLLDiff: Guiding Diffusion Models Towards Unsupervised Training-free Real-world Low-light Image Enhancement (AAAI 2025 · similarity 0.86)
- UltraLED: Learning to See Everything in Ultra-High Dynamic Range Scenes (NeurIPS 2025 · similarity 0.86)
- CLIP-RestoreX: Restore Image Structure and Perception in Exposure Correction (AAAI 2025 · similarity 0.86)
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.
