Real Cameras are Radiometric: new paper lifts any geometric camera model with aperture, focus surface, vignetting
ssh4net · x · 2026-11-21
The paper "Real Cameras are Radiometric" (Ties Harder et al.) proposes the radiometric camera, lifting any central geometric camera model to a radiometric one by adding a finite aperture, a free-form focus surface, and a generic vignetting function while keeping the original projection.
Key points:
- Calibration is done via inverse rendering directly on sensor relative irradiance, using the same calibration-target images as geometric calibration — no features needed. Since geometric calibration absorbs feature-detection bias, the radiometric approach achieves lower reprojection errors on held-out test sets.
- The focus surface extends defocus modeling from the thin lens to any central camera, already used in a first multi-view application.
- A comparison table shows prior camera models/calibration methods cover only subsets; this is the first to jointly calibrate arbitrary central projection, aperture, focus surface, and vignetting.
- Figures demonstrate downstream tasks: vignetting correction for panorama stitching and texture learning in inverse rendering, outperforming pinhole models.
More from Research
- DatologyAI open-sources Zephon, a data loader with elastic determinism for LLM training — josh_wills · 2026-11-21
- Meta compression researcher Guangxuan Xiao says gradual pruning won't recover performance faster — teortaxesTex · 2026-11-21
- USC's Heave-Ho trains Unitree G1 humanoids to collaborate, scales 2 to 10 robots — CyberRobooo · 2026-11-21
- Kaomoji emotion probe validated: 0.68 correlation with valence direction, 0.79 when trained — Sauers_ · 2026-11-21
- PhysLDM reconstructs deformable trajectories at 2.48mm precision with 78x token compression — jiqizhixin · 2026-11-21
- Hand-crafting single MLP neurons to steer model behavior — no optimizer needed — voooooogel · 2026-11-21