Research map / 3D vision / Depth and stereo
Optical Model-Driven Sharpness Mapping for Autofocus in Small Depth-of-Field and Severe Defocus Scenarios
Chen-Liang Fan, Mingpei Cao, Chih Chien Hung, Yuesheng Zhu
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 U2-Net instead of U-Net for sharpness regression (ablation) | sharpness network · 800 focal stacks | MAE: 0.0582 -> 0.057 | Improved |
| use U2-Net instead of ViT for sharpness regression (ablation) | sharpness network · 800 focal stacks | MAE: 0.0639 -> 0.057 | Improved |
| use U2-Net instead of Res-Unet (ablation) | sharpness network · 800 focal stacks | MAE: 0.06 -> 0.057 | Improved |
| use U2-Net instead of Dense-Unet (ablation) | sharpness network · 800 focal stacks | MAE: 0.0607 -> 0.057 | Improved |
| replace direct distance estimation with optical-model sharpness plus adaptive adjustment (vs Herrmann et al., Anodized) | autofocus pipeline · AOI Anodized, 5 iterations | in-focus rate: 0.455 -> 0.998 | Improved |
| replace depth-from-defocus model (Nazir et al.) with sharpness mapping, Paper texture | autofocus pipeline · AOI Paper, 5 iterations | in-focus rate: 0.759 -> 0.947 | Improved |
| replace coarse-to-fine peak search with sharpness mapping (Modified DCT best conventional) | autofocus pipeline · AOI severe defocus | in-focus rate: 0.93 (10 steps) -> 0.912 (3 steps) | Improved |
| wrong initial direction (take one more step to resolve ambiguity) | Ours · AOI severe defocus | in-focus rate: 0.912 (3 steps) vs 0.962 (5 steps) | Mixed |
| bright illumination instead of normal | Ours · metal samples, step 3 | in-focus rate: 0.958 -> 0.754 | Hurt |
| dark illumination instead of normal | Ours · metal samples, step 3 | in-focus rate: 0.958 -> 0.923 | No clear effect |
| printed paper texture instead of anodized | Ours · step 3 | in-focus rate: 0.996 -> 0.719 | Hurt |
| severe defocus start vs all-range start | Ours · step 5, initial direction correct | in-focus rate: 0.967 -> 0.975 | No clear effect |
Closest papers by meaning
- Lens Parameter Estimation for Realistic Depth of Field Modeling (ICCV 2023 · similarity 0.87)
- Spatially-Varying Autofocus (ICCV 2025 · similarity 0.86)
- Learning to Autofocus (CVPR 2020 · similarity 0.86)
- User-Instructed Disparity-aware Defocus Control (NeurIPS 2025 · similarity 0.85)
- Multi-Focus Image Fusion via Explicit Defocus Blur Modelling (AAAI 2025 · similarity 0.85)
- DualFocus: Depth from Focus with Spatio-Focal Dual Variational Constraints (NeurIPS 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.
