Research map / 3D vision / Depth and stereo
Learning Lens Blur Fields
Esther Y. H. Lin, Zhecheng Wang, Rebecca Lin, Daniel Miau, Florian Kainz, Jiawen Chen, Xuaner Zhang, David B. Lindell, Kiriakos N. Kutulakos
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 proposed MLP blur field instead of Mannan & Langer | PSF estimation · Synthetic, validation lens positions | PSNR: 15.291 -> 30.462 | Improved |
| use proposed MLP blur field instead of Joshi et al. | PSF estimation · Synthetic, validation lens positions, validation patterns | PSNR: 28.964 -> 30.396 | Improved |
| use proposed MLP blur field instead of Joshi et al. | PSF estimation · Synthetic, validation lens positions, training patterns | PSNR: 30.797 -> 30.462 | Hurt |
| use proposed MLP blur field instead of Joshi et al. | PSF estimation · Synthetic, training lens positions, training patterns | SSIM: 0.950 -> 0.929 | Hurt |
| use MLP interpolation instead of linear interpolation of MLP PSFs (ablation) | lens blur field · Synthetic, validation lens positions, training patterns | PSNR: 29.394 -> 30.462 | Improved |
| use MLP interpolation instead of linear interpolation of MLP PSFs (ablation) | lens blur field · Synthetic, validation lens positions, validation patterns | SSIM: 0.887 -> 0.919 | Improved |
| use blur field PSFs instead of Seidel model PSFs (ablation) | image restoration · Siemens Star restoration | PSNR: 7.415 -> 8.900 | Improved |
| use blur field PSFs instead of Seidel model PSFs (ablation) | image restoration · Siemens Star restoration | SSIM: 0.127 -> 0.293 | Improved |
| use blur field PSFs instead of Seidel model PSFs (ablation) | image restoration · Synthetic blurry image (Blue channel) | PSNR: 17.754 -> 17.700 | Hurt |
| use MLP representation instead of discrete 5D PSF optimization | PSF acquisition · 5D PSF, iPhone 12 Pro | optimization time: 34 days -> 14 hours | Improved |
| use MLP representation instead of discrete 5D PSF storage | PSF acquisition · 5D PSF, iPhone 12 Pro | memory: 9.3GB -> 19.1MB | Improved |
| add target distance dimension (6D instead of 5D blur field) | lens blur field · iPhone 12 Pro wide | PSF variation | No clear effect |
Closest papers by meaning
- Learning Depth from Focus in the Wild (ECCV 2022 · similarity 0.87)
- A Physics-Informed Blur Learning Framework for Imaging Systems (CVPR 2025 · similarity 0.85)
- Multi-Focus Image Fusion via Explicit Defocus Blur Modelling (AAAI 2025 · similarity 0.85)
- LDP: Language-driven Dual-Pixel Image Defocus Deblurring Network (CVPR 2024 · similarity 0.84)
- A Physics-informed Low-rank Deep Neural Network for Blind and Universal Lens Aberration Correction (CVPR 2024 · similarity 0.84)
- BLURD: Benchmarking and Learning using a Unified Rendering and Diffusion Model (NeurIPS 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.
