Research map / 3D vision / Depth and stereo
Non-Line-of-Sight Surface Reconstruction Using the Directional Light-Cone Transform
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 D-LCT to jointly recover albedo and surface normals instead of LCT depth followed by normal estimation | NLOS reconstruction · ZNLOS Bunny | depth RMSE: 5.97cm -> 4.96cm | Helped |
| Use D-LCT to jointly recover albedo and surface normals instead of LCT depth followed by normal estimation | NLOS reconstruction · ZNLOS Bunny | depth MAE: 1.87cm -> 1.59cm | Helped |
| Use D-LCT to jointly recover albedo and surface normals instead of estimating normals from LCT depth | NLOS reconstruction · ZNLOS Bunny | surface normal end-point RMSE: 0.91cm -> 0.52cm | Helped |
| Use D-LCT to jointly recover albedo and surface normals instead of estimating normals from LCT depth | NLOS reconstruction · ZNLOS Bunny | surface normal MAE: 0.61cm -> 0.38cm | Helped |
| Use D-LCT rather than Fermat flow | NLOS surface reconstruction · Bunny, 64 × 64 spatial resolution | runtime: 1.6s -> 5.2s | Hurt |
| Use D-LCT rather than LCT | Bunny reconstruction · Bunny, 512 × 512 spatial resolution | runtime: 30.4s -> 370.0s | Hurt |
| Use D-LCT rather than Fermat flow | NLOS surface reconstruction · Bunny, 512 × 512 spatial resolution | runtime: 86.5s -> 370.0s | Hurt |
| Use D-LCT rather than Heide et al. joint albedo-normal optimization | NLOS surface reconstruction · Bunny, 64 × 64 spatial resolution | runtime: >10h vs 5.2s | Helped |
| Use D-LCT rather than Tsai et al. inverse rendering | NLOS surface reconstruction · Bunny, 64 × 64 spatial resolution | runtime: 7h vs 5.2s | Helped |
| Use joint albedo-normal D-LCT rather than methods specialized to surface reconstruction | NLOS reconstruction on captured Stanford data · SU, Discobolus, and Dragon | surface reconstruction quality | Helped |
| Use the linearized cosine fall-off model rather than zero cosine fall-off | NLOS surface reconstruction · ZNLOS scenes | surface geometry accuracy | Helped |
Closest papers by meaning
- Dual-branch Graph Feature Learning for NLOS Imaging (AAAI 2025 · similarity 0.88)
- Factored-NeuS: Reconstructing Surfaces, Illumination, and Materials of Possibly Glossy Objects (CVPR 2025 · similarity 0.88)
- Virtual Scanning: Unsupervised Non-line-of-sight Imaging from Irregularly Undersampled Transients (NeurIPS 2024 · similarity 0.86)
- 3D Gaussian Flats: Hybrid 2D/3D Photometric Scene Reconstruction (NeurIPS 2025 · similarity 0.86)
- RobIR: Robust Inverse Rendering for High-Illumination Scenes (NeurIPS 2024 · similarity 0.85)
- Multi-times Monte Carlo Rendering for Inter-reflection Reconstruction (NeurIPS 2024 · similarity 0.85)
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.